Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 29 Apr 2021 18:03:25 +0000
From:      bugzilla-noreply@freebsd.org
To:        python@FreeBSD.org
Subject:   [Bug 255445] lang/python 3.8/3.9 SIGSEV core dumps in libthr TrueNAS
Message-ID:  <bug-255445-21822-whZ5DF3hXy@https.bugs.freebsd.org/bugzilla/>
In-Reply-To: <bug-255445-21822@https.bugs.freebsd.org/bugzilla/>
References:  <bug-255445-21822@https.bugs.freebsd.org/bugzilla/>

next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D255445

--- Comment #13 from yocalebo@gmail.com <yocalebo@gmail.com> ---
Maybe some progress....not sure but I'm cautiously optimistic.

I analyzed 7 core dumps again to see if I spotted anything interesting. 6 of
them were py3.8 and 1 was py3.9.

100% of them have a common pattern. The threads *tstate has
lxml.etree._ParserDictionaryContext in the frames. Either the thread that w=
as
on CPU when it core dumped or some other thread. Since this is seemingly me=
mory
corruption, lxml has become my suspect since it uses it's own C bindings for
the libxml2 and libxslt libraries.

I've instrumented a somewhat complicated script that daemonizes, creates a
concurrent.futures._base.Executor class and calls methods that use lxml lib=
rary
to parse geom xml information in a while true loop.

Am I looking at a red herring?? Idk but this grabbed my attention so I'm
running the script to see if it'll tickle the problem. Maybe the version of
py-lxml we're using has a subtle issue with py3.8+. (Queue my overly dramat=
ic
rant about py3.8 changes to PyGC_Head struct.)

--=20
You are receiving this mail because:
You are the assignee for the bug.=



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-255445-21822-whZ5DF3hXy>