Re: Problem building libnotify

[Robert Nestor <rnestor%mac.com@localhost>]

Reading thru the information in PR 51266 I was left with the feeling that between the time of successful builds and unsuccessful ones there were one or two updated to GCC.  Is it possible that newer versions of GCC are emitting code used to access shared data in glib/gthread that doesn’t guarantee exclusive access that worked with previous versions of GCC?  That might explain the random nature of the problem where builds work for some sometimes but not for others.

Is there some easy way to test this?  Installing an old version of GCC, disable threading, force running with only one CPU on a system, etc?

-bob

On Apr 3, 2017, at 4:11 PM, Patrick Welche <prlw1%cam.ac.uk@localhost> wrote:

> On Mon, Apr 03, 2017 at 09:35:19AM +0200, Martin Husemann wrote:
>> The gobject-introspection core dump is the now no longer infamous PR 51266,
>> which happens more or less randomly and is currently assumed to be a bug
>> in glib2.
> 
> Then why can I repeatably successfully build glib-introspection and libnotify?
> And why is it showing up now when the glib/gthread* files haven't changed
> in a long while, apart for a change to use pthread_setname_np() which
> I presume is harmless? (on -current/amd64 - don't know about i386)
> 
> Cheers,
> 
> Patrick