Bug#935292: lintian: reporing-harness get stuck when lintian deadlocks
On Wed, 21 Aug 2019 11:35:00 +0000 Niels Thykier <niels@thykier.net> wrote:
> Niels Thykier:
> > Package: lintian
> > Version: 2.18.0
> > Severity: important
> >
> > Spotted on lindsay.d.o:
> >
> > The reporting framework runs lintian on a timer and can send it
> > SIGTERM if it gets stuck (usually when the processing time is over).
> > However, it seems that the reporting framework sometimes just
> > dies/disappeares after lintian is killed. Below is the log from where
> > lintian was manually killed by me on a stuck package - but there are
> > plenty of examples that suggest it also happens when the
> > reporting-lintian-harness automatically kills lintian on a timeout.
> >
> > """
> > [2019-08-21T10:26:24]: New lintian log at /srv/lintian.debian.org/logs/lintian.log-bPGtadL
> > [2019-08-21T10:29:05]: [lintian] error processing golang-1.11/1.11.13-1 (time: 122.898s)
> > [2019-08-21T10:29:06]: [lintian] processed golang-github-mcuadros-go-gin-prometheus/0.1.0+git20190723.c7374e9-2 successfully (time: 1.396s)
> > [2019-08-21T10:32:37]: [lintian] processed grass/7.8.0~rc1-1~exp1 successfully (time: 210.685s)
> > [2019-08-21T10:33:45]: [lintian] processed grpc/1.23.0-1 successfully (time: 68.446s)
> > [2019-08-21T10:33:50]: [lintian] processed guake/3.6.3-1 successfully (time: 5.118s)
> > [2019-08-21T10:33:56]: [lintian] processed intel-cmt-cat/3.1.0-2 successfully (time: 5.953s)
> > [2019-08-21T10:34:03]: [lintian] processed intel-processor-trace/2.0.1-1 successfully (time: 6.566s)
> > [2019-08-21T10:34:06]: [lintian] processed libbpp-qt/2.4.1-2 successfully (time: 3.384s)
> > [2019-08-21T10:34:13]: [lintian] processed libbpp-seq-omics/2.4.1-4 successfully (time: 7.356s)
> > [2019-08-21T10:34:17]: [lintian] processed libconfig-model-itself-perl/2.018-1 successfully (time: 3.495s)
> > [2019-08-21T10:34:21]: [lintian] processed libseqlib/1.1.2+dfsg-4 successfully (time: 3.913s)
> > [2019-08-21T10:55:19]: [lintian] processed linux-signed-i386/5.2.9+1 successfully (time: 1258.298s)
> > [2019-08-21T10:55:42]: [lintian] processed mandos/1.8.8-1 successfully (time: 22.420s)
> > [2019-08-21T11:15:26]: Signal SIGTERM acknowledged: disabled timed alarms
> > """
> >
> > After this, the harness.log simply "stops".
> >
> > [...]
>
> Correction, it died. I spotted a mail from cron in
> /var/spool/mail/lintian saying it could not send a main due to invalid
> cmd-line parameters to the mail-cmd in our crontab. I have "fixed" this
> setting MAILTO and removing the explicit mail-cmd - it has apparently
> never worked.
>
> Thanks,
> ~Niels
>
>
>
>
We are still experiencing this issue and have now almost 60 blacklisted
packages to work around this (and a few other issues). Sadly, mailing
does not work, so we do not get the actual error from
reporting-lintian-harness (so I do not have any additional debug
information).
This issue neuters the lintian processing as all results end up being
discarded due to this issue.
Thanks,
~Niels
Reply to: