Il giorno mar, 19/05/2020 alle 18.43 +0200, Dan Horák ha scritto:
> On Tue, 19 May 2020 17:42:29 +0200
> Guido Aulisi <guido.aul...@gmail.com> wrote:
> 
> > Il giorno mar, 19/05/2020 alle 12.11 +0200, Florian Weimer ha
> > scritto:
> > > * Guido Aulisi:
> > > 
> > > > I'm getting some strange errors from some packages built for
> > > > f32
> > > > with
> > > > gcc 10.0 [0].
> > > > Building with g++ 10.1 ardourd5 seems fine...
> > > > 
> > > > It seems GCC 10.0 had some bugs that could be discovered only
> > > > at
> > > > runtime. Did you have any similar problems?
> > > > 
> > > > Ciao
> > > > Guido
> > > > FAS: tartina
> > > > 
> > > > [0]: https://bugzilla.redhat.com/show_bug.cgi?id=1837089
> > > 
> > > Comment 12 on that bug suggests that it's the known underlinking
> > > issue
> > > in Ardour (“it seems related to a missing symbol (__atan2_finite)
> > > in
> > > libvampplugins”).  If true, this is not related to GCC 10 at all.
> > 
> > Yes, that really seems a link and/or symbol versioning problem, but
> > it
> > wasn't the only one.
> > I had some other crashes in packages, usually not x86 arches.
> > As an example sord [0] was crashing on power64 when compiled with
> > -O2,
> > an exit condition in a for loop was optimized out until it crashed
> > the
> > loop. I had to use -O1 [1], and I have not tried with gcc 10.1 yet.
> 
> usually it's useful to report a bug against gcc if such things
> happen.
> Also there is https://gcc.gnu.org/wiki/FAQ#misoptimization the should
> help to localize the root cause. My blog
> (https://gcc.gnu.org/wiki/FAQ#misoptimization) has also some
> pointers.
> 
> 
>               Dan

Yes, I was trying to report a bug, but it was really hard to repeat all
the steps I made to get it; at that time I thought it was a program
bug, not some compiler optimization.
Next time (I hope never!) I will try to file a bug...

Ciao
Guido

Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org

Reply via email to