make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
//go:build !windows
package main
import (
"context"
cryptorand "crypto/rand"
"fmt"
"io/fs"
"net"
"os"
"os/signal"
"path/filepath"
"runtime"
"runtime/debug"
"strings"
"sync"
"syscall"
"time"
2023-12-05 15:35:58 +03:00
"golang.org/x/exp/slog"
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/client_golang/prometheus/promauto"
"github.com/mjl-/mox/dns"
"github.com/mjl-/mox/dnsbl"
"github.com/mjl-/mox/message"
"github.com/mjl-/mox/metrics"
"github.com/mjl-/mox/mlog"
"github.com/mjl-/mox/mox-"
"github.com/mjl-/mox/moxvar"
"github.com/mjl-/mox/store"
"github.com/mjl-/mox/updates"
)
2023-12-05 15:35:58 +03:00
func monitorDNSBL ( log mlog . Log ) {
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
defer func ( ) {
// On error, don't bring down the entire server.
x := recover ( )
if x != nil {
2023-12-05 15:35:58 +03:00
log . Error ( "monitordnsbl panic" , slog . Any ( "panic" , x ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
debug . PrintStack ( )
metrics . PanicInc ( metrics . Serve )
}
} ( )
l , ok := mox . Conf . Static . Listeners [ "public" ]
if ! ok {
log . Info ( "no listener named public, not monitoring our ips at dnsbls" )
return
}
var zones [ ] dns . Domain
for _ , zone := range l . SMTP . DNSBLs {
d , err := dns . ParseDomain ( zone )
if err != nil {
2023-12-05 15:35:58 +03:00
log . Fatalx ( "parsing dnsbls zone" , err , slog . Any ( "zone" , zone ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
zones = append ( zones , d )
}
if len ( zones ) == 0 {
return
}
type key struct {
zone dns . Domain
ip string
}
metrics := map [ key ] prometheus . GaugeFunc { }
var statusMutex sync . Mutex
statuses := map [ key ] bool { }
resolver := dns . StrictResolver { Pkg : "dnsblmonitor" }
var sleep time . Duration // No sleep on first iteration.
for {
time . Sleep ( sleep )
sleep = 3 * time . Hour
ips , err := mox . IPs ( mox . Context , false )
if err != nil {
log . Errorx ( "listing ips for dnsbl monitor" , err )
continue
}
for _ , ip := range ips {
if ip . IsLoopback ( ) || ip . IsPrivate ( ) {
continue
}
for _ , zone := range zones {
2023-12-05 15:35:58 +03:00
status , expl , err := dnsbl . Lookup ( mox . Context , log . Logger , resolver , zone , ip )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
if err != nil {
2023-12-05 15:35:58 +03:00
log . Errorx ( "dnsbl monitor lookup" , err , slog . Any ( "ip" , ip ) , slog . Any ( "zone" , zone ) , slog . String ( "expl" , expl ) , slog . Any ( "status" , status ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
k := key { zone , ip . String ( ) }
statusMutex . Lock ( )
statuses [ k ] = status == dnsbl . StatusPass
statusMutex . Unlock ( )
if _ , ok := metrics [ k ] ; ! ok {
metrics [ k ] = promauto . NewGaugeFunc (
prometheus . GaugeOpts {
Name : "mox_dnsbl_ips_success" ,
Help : "DNSBL lookups to configured DNSBLs of our IPs." ,
ConstLabels : prometheus . Labels {
"zone" : zone . LogString ( ) ,
"ip" : k . ip ,
} ,
} ,
func ( ) float64 {
statusMutex . Lock ( )
defer statusMutex . Unlock ( )
if statuses [ k ] {
return 1
}
return 0
} ,
)
}
time . Sleep ( time . Second )
}
}
}
}
// also see localserve.go, code is similar or even shared.
func cmdServe ( c * cmd ) {
c . help = ` Start mox , serving SMTP / IMAP / HTTPS .
Incoming email is accepted over SMTP . Email can be retrieved by users using
IMAP . HTTP listeners are started for the admin / account web interfaces , and for
automated TLS configuration . Missing essential TLS certificates are immediately
requested , other TLS certificates are requested on demand .
Only implemented on unix systems , not Windows .
`
args := c . Parse ( )
if len ( args ) != 0 {
c . Usage ( )
}
// Set debug logging until config is fully loaded.
mlog . Logfmt = true
mox . Conf . Log [ "" ] = mlog . LevelDebug
mlog . SetConfig ( mox . Conf . Log )
checkACMEHosts := os . Getuid ( ) != 0
2023-12-05 15:35:58 +03:00
log := c . log
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
if os . Getuid ( ) == 0 {
mox . MustLoadConfig ( true , checkACMEHosts )
// No need to potentially start and keep multiple processes. As root, we just need
// to start the child process.
runtime . GOMAXPROCS ( 1 )
moxconf , err := filepath . Abs ( mox . ConfigStaticPath )
log . Check ( err , "finding absolute mox.conf path" )
domainsconf , err := filepath . Abs ( mox . ConfigDynamicPath )
log . Check ( err , "finding absolute domains.conf path" )
2023-12-05 15:35:58 +03:00
log . Print ( "starting as root, initializing network listeners" , slog . String ( "version" , moxvar . Version ) , slog . Any ( "pid" , os . Getpid ( ) ) , slog . String ( "moxconf" , moxconf ) , slog . String ( "domainsconf" , domainsconf ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
if os . Getenv ( "MOX_SOCKETS" ) != "" {
log . Fatal ( "refusing to start as root with $MOX_SOCKETS set" )
}
if os . Getenv ( "MOX_FILES" ) != "" {
log . Fatal ( "refusing to start as root with $MOX_FILES set" )
}
if ! mox . Conf . Static . NoFixPermissions {
// Fix permissions now that we have privilege to do so. Useful for update of v0.0.1
// that was running directly as mox-user.
workdir , err := os . Getwd ( )
if err != nil {
log . Printx ( "get working dir, continuing without potentially fixing up permissions" , err )
} else {
configdir := filepath . Dir ( mox . ConfigStaticPath )
datadir := mox . DataDirPath ( "." )
err := fixperms ( log , workdir , configdir , datadir , mox . Conf . Static . UID , mox . Conf . Static . GID )
if err != nil {
log . Fatalx ( "fixing permissions" , err )
}
}
}
} else {
mox . RestorePassedFiles ( )
mox . MustLoadConfig ( true , checkACMEHosts )
2023-12-05 15:35:58 +03:00
log . Print ( "starting as unprivileged user" , slog . String ( "user" , mox . Conf . Static . User ) , slog . Any ( "uid" , mox . Conf . Static . UID ) , slog . Any ( "gid" , mox . Conf . Static . GID ) , slog . Any ( "pid" , os . Getpid ( ) ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
syscall . Umask ( syscall . Umask ( 007 ) | 007 )
// Initialize key and random buffer for creating opaque SMTP
// transaction IDs based on "cid"s.
recvidpath := mox . DataDirPath ( "receivedid.key" )
recvidbuf , err := os . ReadFile ( recvidpath )
if err != nil || len ( recvidbuf ) != 16 + 8 {
recvidbuf = make ( [ ] byte , 16 + 8 )
if _ , err := cryptorand . Read ( recvidbuf ) ; err != nil {
log . Fatalx ( "reading random recvid data" , err )
}
if err := os . WriteFile ( recvidpath , recvidbuf , 0660 ) ; err != nil {
2023-12-05 15:35:58 +03:00
log . Fatalx ( "writing recvidpath" , err , slog . String ( "path" , recvidpath ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
err := os . Chown ( recvidpath , int ( mox . Conf . Static . UID ) , 0 )
2023-12-05 15:35:58 +03:00
log . Check ( err , "chown receveidid.key" , slog . String ( "path" , recvidpath ) , slog . Any ( "uid" , mox . Conf . Static . UID ) , slog . Any ( "gid" , 0 ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
err = os . Chmod ( recvidpath , 0640 )
2023-12-05 15:35:58 +03:00
log . Check ( err , "chmod receveidid.key to 0640" , slog . String ( "path" , recvidpath ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
if err := mox . ReceivedIDInit ( recvidbuf [ : 16 ] , recvidbuf [ 16 : ] ) ; err != nil {
log . Fatalx ( "init receivedid" , err )
}
// Start mox. If running as root, this will bind/listen on network sockets, and
// fork and exec itself as unprivileged user, then waits for the child to stop and
// exit. When running as root, this function never returns. But the new
// unprivileged user will get here again, with network sockets prepared.
//
// We listen to the unix domain ctl socket afterwards, which we always remove
// before listening. We need to do that because we may not have cleaned up our
// control socket during unexpected shutdown. We don't want to remove and listen on
// the unix domain socket first. If we would, we would make the existing instance
// unreachable over its ctl socket, and then fail because the network addresses are
// taken.
const mtastsdbRefresher = true
const skipForkExec = false
implement outgoing tls reports
we were already accepting, processing and displaying incoming tls reports. now
we start tracking TLS connection and security-policy-related errors for
outgoing message deliveries as well. we send reports once a day, to the
reporting addresses specified in TLSRPT records (rua) of a policy domain. these
reports are about MTA-STS policies and/or DANE policies, and about
STARTTLS-related failures.
sending reports is enabled by default, but can be disabled through setting
NoOutgoingTLSReports in mox.conf.
only at the end of the implementation process came the realization that the
TLSRPT policy domain for DANE (MX) hosts are separate from the TLSRPT policy
for the recipient domain, and that MTA-STS and DANE TLS/policy results are
typically delivered in separate reports. so MX hosts need their own TLSRPT
policies.
config for the per-host TLSRPT policy should be added to mox.conf for existing
installs, in field HostTLSRPT. it is automatically configured by quickstart for
new installs. with a HostTLSRPT config, the "dns records" and "dns check" admin
pages now suggest the per-host TLSRPT record. by creating that record, you're
requesting TLS reports about your MX host.
gathering all the TLS/policy results is somewhat tricky. the tentacles go
throughout the code. the positive result is that the TLS/policy-related code
had to be cleaned up a bit. for example, the smtpclient TLS modes now reflect
reality better, with independent settings about whether PKIX and/or DANE
verification has to be done, and/or whether verification errors have to be
ignored (e.g. for tls-required: no header). also, cached mtasts policies of
mode "none" are now cleaned up once the MTA-STS DNS record goes away.
2023-11-09 19:40:46 +03:00
if err := start ( mtastsdbRefresher , ! mox . Conf . Static . NoOutgoingDMARCReports , ! mox . Conf . Static . NoOutgoingTLSReports , skipForkExec ) ; err != nil {
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
log . Fatalx ( "start" , err )
}
log . Print ( "ready to serve" )
if mox . Conf . Static . CheckUpdates {
checkUpdates := func ( ) time . Duration {
next := 24 * time . Hour
current , lastknown , mtime , err := mox . LastKnown ( )
if err != nil {
log . Infox ( "determining own version before checking for updates, trying again in 24h" , err )
return next
}
// We don't want to check for updates at every startup. So we sleep based on file
// mtime. But file won't exist initially.
if ! mtime . IsZero ( ) && time . Since ( mtime ) < 24 * time . Hour {
d := 24 * time . Hour - time . Since ( mtime )
2023-12-05 15:35:58 +03:00
log . Debug ( "sleeping for next check for updates" , slog . Duration ( "sleep" , d ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
time . Sleep ( d )
next = 0
}
now := time . Now ( )
if err := os . Chtimes ( mox . DataDirPath ( "lastknownversion" ) , now , now ) ; err != nil {
if ! os . IsNotExist ( err ) {
log . Infox ( "setting mtime on lastknownversion file, continuing" , err )
}
}
2023-12-05 15:35:58 +03:00
log . Debug ( "checking for updates" , slog . Any ( "lastknown" , lastknown ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
updatesctx , updatescancel := context . WithTimeout ( mox . Context , time . Minute )
2023-12-05 15:35:58 +03:00
latest , _ , changelog , err := updates . Check ( updatesctx , log . Logger , dns . StrictResolver { Log : log . Logger } , dns . Domain { ASCII : changelogDomain } , lastknown , changelogURL , changelogPubKey )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
updatescancel ( )
if err != nil {
2023-12-05 15:35:58 +03:00
log . Infox ( "checking for updates" , err , slog . Any ( "latest" , latest ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
return next
}
if ! latest . After ( lastknown ) {
log . Debug ( "no new version available" )
return next
}
if len ( changelog . Changes ) == 0 {
2023-12-05 15:35:58 +03:00
log . Info ( "new version available, but changelog is empty, ignoring" , slog . Any ( "latest" , latest ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
return next
}
var cl string
for _ , c := range changelog . Changes {
cl += "----\n\n" + strings . TrimSpace ( c . Text ) + "\n\n"
}
cl += "----"
2023-12-05 15:35:58 +03:00
a , err := store . OpenAccount ( log , mox . Conf . Static . Postmaster . Account )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
if err != nil {
log . Infox ( "open account for postmaster changelog delivery" , err )
return next
}
defer func ( ) {
err := a . Close ( )
log . Check ( err , "closing account" )
} ( )
2023-12-05 15:35:58 +03:00
f , err := store . CreateMessageTemp ( log , "changelog" )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
if err != nil {
log . Infox ( "making temporary message file for changelog delivery" , err )
return next
}
2023-11-01 20:57:38 +03:00
defer store . CloseRemoveTempFile ( log , f , "message for changelog delivery" )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
m := & store . Message {
Received : time . Now ( ) ,
Flags : store . Flags { Flagged : true } ,
}
n , err := fmt . Fprintf ( f , "Date: %s\r\nSubject: mox %s available\r\nContent-Type: text/plain; charset=utf-8\r\nContent-Transfer-Encoding: 8-bit\r\n\r\nHi!\r\n\r\nVersion %s of mox is available, this install is at %s.\r\n\r\nChanges:\r\n\r\n%s\r\n\r\nRemember to make a backup with \"mox backup\" before upgrading.\r\nPlease report any issues at https://github.com/mjl-/mox, thanks!\r\n\r\nCheers,\r\nmox\r\n" , time . Now ( ) . Format ( message . RFC5322Z ) , latest , latest , current , strings . ReplaceAll ( cl , "\n" , "\r\n" ) )
if err != nil {
log . Infox ( "writing temporary message file for changelog delivery" , err )
return next
}
m . Size = int64 ( n )
if err := a . DeliverMailbox ( log , mox . Conf . Static . Postmaster . Mailbox , m , f ) ; err != nil {
log . Errorx ( "changelog delivery" , err )
return next
}
2023-12-05 15:35:58 +03:00
log . Info ( "delivered changelog" , slog . Any ( "current" , current ) , slog . Any ( "lastknown" , lastknown ) , slog . Any ( "latest" , latest ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
if err := mox . StoreLastKnown ( latest ) ; err != nil {
// This will be awkward, we'll keep notifying the postmaster once every 24h...
log . Infox ( "updating last known version" , err )
}
return next
}
go func ( ) {
for {
next := checkUpdates ( )
time . Sleep ( next )
}
} ( )
}
go monitorDNSBL ( log )
ctlpath := mox . DataDirPath ( "ctl" )
_ = os . Remove ( ctlpath )
ctl , err := net . Listen ( "unix" , ctlpath )
if err != nil {
log . Fatalx ( "listen on ctl unix domain socket" , err )
}
go func ( ) {
for {
conn , err := ctl . Accept ( )
if err != nil {
log . Printx ( "accept for ctl" , err )
continue
}
cid := mox . Cid ( )
ctx := context . WithValue ( mox . Context , mlog . CidKey , cid )
go servectl ( ctx , log . WithCid ( cid ) , conn , func ( ) { shutdown ( log ) } )
}
} ( )
// Remove old temporary files that somehow haven't been cleaned up.
tmpdir := mox . DataDirPath ( "tmp" )
os . MkdirAll ( tmpdir , 0770 )
tmps , err := os . ReadDir ( tmpdir )
if err != nil {
log . Errorx ( "listing files in tmpdir" , err )
} else {
now := time . Now ( )
for _ , e := range tmps {
if fi , err := e . Info ( ) ; err != nil {
2023-12-05 15:35:58 +03:00
log . Errorx ( "stat tmp file" , err , slog . String ( "filename" , e . Name ( ) ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
} else if now . Sub ( fi . ModTime ( ) ) > 7 * 24 * time . Hour && ! fi . IsDir ( ) {
p := filepath . Join ( tmpdir , e . Name ( ) )
if err := os . Remove ( p ) ; err != nil {
2023-12-05 15:35:58 +03:00
log . Errorx ( "removing stale temporary file" , err , slog . String ( "path" , p ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
} else {
2023-12-05 15:35:58 +03:00
log . Info ( "removed stale temporary file" , slog . String ( "path" , p ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
}
}
}
// Graceful shutdown.
sigc := make ( chan os . Signal , 1 )
signal . Notify ( sigc , os . Interrupt , syscall . SIGTERM )
sig := <- sigc
2023-12-05 15:35:58 +03:00
log . Print ( "shutting down, waiting max 3s for existing connections" , slog . Any ( "signal" , sig ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
shutdown ( log )
if num , ok := sig . ( syscall . Signal ) ; ok {
os . Exit ( int ( num ) )
} else {
os . Exit ( 1 )
}
}
// Set correct permissions for mox working directory, binary, config and data and service file.
//
// We require being able to stat the basic non-optional paths. Then we'll try to
// fix up permissions. If an error occurs when fixing permissions, we log and
// continue (could not be an actual problem).
2023-12-05 15:35:58 +03:00
func fixperms ( log mlog . Log , workdir , configdir , datadir string , moxuid , moxgid uint32 ) ( rerr error ) {
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
type fserr struct { Err error }
defer func ( ) {
x := recover ( )
if x == nil {
return
}
e , ok := x . ( fserr )
if ok {
rerr = e . Err
} else {
panic ( x )
}
} ( )
checkf := func ( err error , format string , args ... any ) {
if err != nil {
panic ( fserr { fmt . Errorf ( format , args ... ) } )
}
}
// Changes we have to make. We collect them first, then apply.
type change struct {
path string
uid , gid * uint32
olduid , oldgid uint32
mode * fs . FileMode
oldmode fs . FileMode
}
var changes [ ] change
ensure := func ( p string , uid , gid uint32 , perm fs . FileMode ) bool {
fi , err := os . Stat ( p )
checkf ( err , "stat %s" , p )
st , ok := fi . Sys ( ) . ( * syscall . Stat_t )
if ! ok {
checkf ( fmt . Errorf ( "got %T" , st ) , "stat sys, expected syscall.Stat_t" )
}
var ch change
if st . Uid != uid || st . Gid != gid {
ch . uid = & uid
ch . gid = & gid
ch . olduid = st . Uid
ch . oldgid = st . Gid
}
if perm != fi . Mode ( ) & ( fs . ModeSetgid | 0777 ) {
ch . mode = & perm
ch . oldmode = fi . Mode ( ) & ( fs . ModeSetgid | 0777 )
}
var zerochange change
if ch == zerochange {
return false
}
ch . path = p
changes = append ( changes , ch )
return true
}
xexists := func ( p string ) bool {
_ , err := os . Stat ( p )
if err != nil && ! os . IsNotExist ( err ) {
checkf ( err , "stat %s" , p )
}
return err == nil
}
// We ensure these permissions:
//
// $workdir root:mox 0751
// $configdir mox:root 0750 + setgid, and recursively (but files 0640)
// $datadir mox:root 0750 + setgid, and recursively (but files 0640)
// $workdir/mox (binary, optional) root:mox 0750
// $workdir/mox.service (systemd service file, optional) root:root 0644
const root = 0
ensure ( workdir , root , moxgid , 0751 )
fixconfig := ensure ( configdir , moxuid , 0 , fs . ModeSetgid | 0750 )
fixdata := ensure ( datadir , moxuid , 0 , fs . ModeSetgid | 0750 )
// Binary and systemd service file do not exist (there) when running under docker.
binary := filepath . Join ( workdir , "mox" )
if xexists ( binary ) {
ensure ( binary , root , moxgid , 0750 )
}
svc := filepath . Join ( workdir , "mox.service" )
if xexists ( svc ) {
ensure ( svc , root , root , 0644 )
}
if len ( changes ) == 0 {
return
}
// Apply changes.
log . Print ( "fixing up permissions, will continue on errors" )
for _ , ch := range changes {
if ch . uid != nil {
err := os . Chown ( ch . path , int ( * ch . uid ) , int ( * ch . gid ) )
2023-12-05 15:35:58 +03:00
log . Printx ( "chown, fixing uid/gid" , err , slog . String ( "path" , ch . path ) , slog . Any ( "olduid" , ch . olduid ) , slog . Any ( "oldgid" , ch . oldgid ) , slog . Any ( "newuid" , * ch . uid ) , slog . Any ( "newgid" , * ch . gid ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
if ch . mode != nil {
err := os . Chmod ( ch . path , * ch . mode )
2023-12-05 15:35:58 +03:00
log . Printx ( "chmod, fixing permissions" , err , slog . String ( "path" , ch . path ) , slog . Any ( "oldmode" , fmt . Sprintf ( "%03o" , ch . oldmode ) ) , slog . Any ( "newmode" , fmt . Sprintf ( "%03o" , * ch . mode ) ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
}
walkchange := func ( dir string ) {
err := filepath . WalkDir ( dir , func ( path string , d fs . DirEntry , err error ) error {
if err != nil {
2023-12-05 15:35:58 +03:00
log . Printx ( "walk error, continuing" , err , slog . String ( "path" , path ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
return nil
}
fi , err := d . Info ( )
if err != nil {
2023-12-05 15:35:58 +03:00
log . Printx ( "stat during walk, continuing" , err , slog . String ( "path" , path ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
return nil
}
st , ok := fi . Sys ( ) . ( * syscall . Stat_t )
if ! ok {
2023-12-05 15:35:58 +03:00
log . Printx ( "syscall stat during walk, continuing" , err , slog . String ( "path" , path ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
return nil
}
if st . Uid != moxuid || st . Gid != root {
err := os . Chown ( path , int ( moxuid ) , root )
2023-12-05 15:35:58 +03:00
log . Printx ( "walk chown, fixing uid/gid" , err , slog . String ( "path" , path ) , slog . Any ( "olduid" , st . Uid ) , slog . Any ( "oldgid" , st . Gid ) , slog . Any ( "newuid" , moxuid ) , slog . Any ( "newgid" , root ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
omode := fi . Mode ( ) & ( fs . ModeSetgid | 0777 )
var nmode fs . FileMode
if fi . IsDir ( ) {
nmode = fs . ModeSetgid | 0750
} else {
nmode = 0640
}
if omode != nmode {
err := os . Chmod ( path , nmode )
2023-12-05 15:35:58 +03:00
log . Printx ( "walk chmod, fixing permissions" , err , slog . String ( "path" , path ) , slog . Any ( "oldmode" , fmt . Sprintf ( "%03o" , omode ) ) , slog . Any ( "newmode" , fmt . Sprintf ( "%03o" , nmode ) ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
return nil
} )
2023-12-05 15:35:58 +03:00
log . Check ( err , "walking dir to fix permissions" , slog . String ( "dir" , dir ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
}
// If config or data dir needed fixing, also set uid/gid and mode and files/dirs
// inside, recursively. We don't always recurse, data probably contains many files.
if fixconfig {
2023-12-05 15:35:58 +03:00
log . Print ( "fixing permissions in config dir" , slog . String ( "configdir" , configdir ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
walkchange ( configdir )
}
if fixdata {
2023-12-05 15:35:58 +03:00
log . Print ( "fixing permissions in data dir" , slog . String ( "configdir" , configdir ) )
make mox compile on windows, without "mox serve" but with working "mox localserve"
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
2023-10-14 11:54:07 +03:00
walkchange ( datadir )
}
return nil
}