On Thu, 2017-12-14 at 13:22 -0800, Adam Williamson wrote:
> What I think is going on in several bugs like this is that there are
> many ways to produce a very similar traceback, because of this
> `_g_log_abort` function in glib. What that basically is - I went and
> looked at it - is a sort of 'special' logging function which, when
> called, is intended to log a message and then terminate the app. I
> think in many cases the same 3 or 4 frames appear in the trace around
> the `_g_log_abort` call - in this case it's these frames that seem to
> appear in all the tracebacks:
> It would probably be a good idea to get some abrt/libreport and glib
> folks together to figure out how abrt/libreport can handle these
> `_g_log_abort` cases (and any other similar cases) better.
> For now, I'll try and go through all the "dupes" of this bug and
> them up.
So I've now gone through all the dupes, tried to sort them into groups,
and posted a little summary:
A couple of bugs cropped up several times. One is
, identified by an
attempt to allocate (exactly) 18446744072098939136 bytes. There's a
matching upstream bug report with a patch currently under review:
It'd be great to get that patch reviewed and then backport it for F27
(and F26, if relevant, I'm not 100% sure).
The other one which just kept coming up is this:
I provided some references to external reports in comment #16,
including a Freedesktop.org
Wayland issue which suggests the underlying
problem is that creating and destroying temporary global objects is
inherently racy (AIUI). I found 8 reports of that one, suggesting it
could really do with fixing.
Aside from that the most commonly encountered ones seem to be a crash
triggered by a bug in the Multi Monitors extension, and one which is
*possibly* caused by trying to run the game Kerbal Space Program. There
are also several bugs which all ultimately seem to wind up crashing due
to "Creating pipes for GWakeup: Too many open files" - like
- but that one
kinda feels like I don't have the whole story figured out yet.
Thanks for any attention you can give to these, folks!
Sorry, totally forgot the second part of the post:
Meanwhile, I recalled we actually have a RHBZ and a github report for
the abrt false-dupe side of things already. I have updated those too:
They've made a change upstream to sort of blacklist (AIUI) _g_log_abort
and another function in the commonly-encountered trace, and sent it to
Rawhide, but not sent it out as an update for F26 and F27 yet; I asked
if they can do that. I also suggested another variant path to
_g_log_abort which may indicate another function or two to add to the
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net