We're using a very old version of it and the functions we
actually use are available in glibc anyway.
The only difference I found was that the result of
glibc's `strptime()`'s `%s` does *not* want an additional
`.to_local()` call anymore...
... which was weird anyway.
As a minor optimization I pass the format string as `&CStr`
to avoid the memcpy.
(The CString::new() call in `strptime()` only happens during
parameter parsing)
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Reviewed-By: Stoiko Ivanov <s.ivanov@proxmox.com>
Tested-By: Stoiko Ivanov <s.ivanov@proxmox.com>
[dropped left-over and fixed FIXME]
Signed-off-by: Stoiko Ivanov <s.ivanov@proxmox.com>
the changes introduced in 1cdbebe57b
changed the date handling to have current_month and current_year
represent the month and year at time of invocation.
This change was not carried over to the logfile collecting, which set
the current month to january.
Signed-off-by: Stoiko Ivanov <s.ivanov@proxmox.com>
Tested-by: Aaron Lauterer <a.lauterer@proxmox.com>
Reviewed-by: Mira Limbeck <m.limbeck@proxmox.com>
rather start with the current time at invocation and if the month in
the log is larger assume a year-wrap happened between logwriting and
invocation.
the old logic is flawed (but that can be said about parsing
traditional syslog timestamps (w/o year) in general) - it got worse
with the change in bullseye of rotating syslog weekly by default -
resulting in users losing one week of logs per day in the new year as
reported in https://forum.proxmox.com/threads/.102322/
Reviewed-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Signed-off-by: Stoiko Ivanov <s.ivanov@proxmox.com>
We only compared the filter `to` with the Noqueue `to` when both were
not empty, and in turn set the nq.dstatus to DStatus::Invalid when they
didn't match.
But in the case of an empty Noqueue `to`, we have to skip it as well if
the filter `to` is not empty.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
Because of the relay before lmtp issue we now add and remove QEntrys
based on their message-id. But if the cleanup line containing the
message-id appears twice in the log for the same QEntry it keeps a weak
reference to itself which leads to a panic because of a mutable borrow
while it is already mutably borrowed in the 'finalize' function.
To circumvent this we check after the lookup if it is the same QEntry
and if so, insert it again with that message-id in the lookup table and
don't add the weak reference.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
Tested-By: Aaron Lauterer <a.lauterer@proxmox.com>
We only match a QEntry with the relay Qentry in the after-queue case
when we get to the 'lmtp' line. This is because it contains a reference
from the first QEntry to the filter, and the filter contains a reference
to the relay QEntry.
If the relay log entries are finished before the 'lmtp' line, there's no
way to match the relay to the first QEntry and we still assume it is
before-queue filtered.
To fix this we try to match the QEntries via message-id and add weak
references to each other. Then we wait with finalizing either until the
other is also 'removed' (finished).
The message-id matching might break if the same message-id is used for
different entries, so this is rather a 'hack' than a nice solution, but
there's no other info in the logs we could use to match both QEntries.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
This is a best effort try to add the outgoing TLS connection information
to the output of pmg-log-tracker. The only thing we can match on is the
PID of the 'smtp' process. In the code we asumme that the TLS log entry
always happens before the actual smtp send entry that has a QID. This means
we save the TLS log entry in a map with the PID as key and then, once the
send entry happens, we look it up and add the log entry to the QEntry's
logs.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
We only have the month, day, hours, minutes and seconds available to us,
but not the year. Because of this we assume the logs are always relative
to the current year. This means when running the log tracker the CTIME
is calculated based on the current year and once a new year starts, all
old tests have to be updated to match the new year.
To work around this issue, use 'faketime' to set the time to 2020-12-31
23:59:59 as all tests assume 2020 as the year.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
Signed-off-by: Stoiko Ivanov <s.ivanov@proxmox.com>
With the rewrite from C to Rust the search string match was changed to
be case sensitive by accident. Fix this by comparing the lowercase values
of both the input and the search string.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
Reviewed-By: Stoiko Ivanov <s.ivanov@proxmox.com>
Tested-By: Stoiko Ivanov <s.ivanov@proxmox.com>
Messages like the following
'proxy-reject: END-OF-MESSAGE: 501 5.5.2 Syntax: MAIL FROM: <address>;'
can happen if an EHLO keyword is announced which is not handled by
pmg-smtp-filter (see #2795). This patch adds output to the log tracker
so this mail shows up as 'rejected' in the GUI instead of silently
ignoring it.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
Tested-By: Stoiko Ivanov <s.ivanov@proxmox.com>
Reviewed-By: Stoiko Ivanov <s.ivanov@proxmox.com>
When a 'to' filter is specified and a mail has multiple 'to's with at
least one being accepted, the others blocked or quarantined and the
'to' for the accepted mail does not match, the entries in the filter
will be ignored even if one of them would match.
To fix it the filter ToEntrys are now considered in the 'to' check.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
In before-queue filtering a filter can be set on the SEntry. When a
search-string is set, we have to check both the SEntry and the attached
filter for a match, not just the SEntry. Otherwise some entries are not shown.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
The testcase contains a mail that is accepted for one address, blocked
for another and quarantined for a third address.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
In the case of before-queue filtering we have the separation 'QEntry'
for accepted mails, as those are the only ones with a queue. And
'SEntry' for blocked and quarantined mails. But if a mail is accepted
for one address, but at the same time blocked or quarantined for another
address we previously discarded those entries. To fix it we now print it
in the 'QEntry' if the attached filter contains blocked and/or
quarantined mails.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
As a user had the problem that 'fa' of 'fatal' matched for a QID,
increase the number of characters that have to match. The QID always has
at least 5 characters for the microseconds, so increase it to 5.
See http://www.postfix.org/postconf.5.html#enable_long_queue_ids
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
Updates the other test outputs' line numbers so the tests don't fail
because of an addition in between.
Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
Currently the status of a before queue filtered mail is the status of the
downstream delivery in case the downstream rejects or defers a mail.
This information is too little to distinguish between a mail which went
through the rule-system, and a locally generated one
(e.g. bounce or spamreport).
By introducing 3 new distinct dstatus values we can use the information to
correctly display the status of a mail in the GUI.
Reviewed-By: Mira Limbeck <m.limbeck@proxmox.com>
Tested-By: Mira Limbeck <m.limbeck@proxmox.com>
Signed-off-by: Stoiko Ivanov <s.ivanov@proxmox.com>
