11:02 a.m.
Hi,
I kicked off a couple of builds last night which appear stuck on koji.
I had checked they build locally with a make mockbuild before
submitting and all was fine.
The builds are:
http://koji.fedoraproject.org/koji/buildinfo?buildID=147755
http://koji.fedoraproject.org/koji/buildinfo?buildID=147756
Should I just cancel and resubmit, or?
Cheers,
Jonathan
3:11 p.m.
On Sat, 19 Dec 2009 17:02:24 +0000
Jonathan Underwood <jonathan.underwood(a)gmail.com> wrote:
Hi,
I kicked off a couple of builds last night which appear stuck on koji.
I had checked they build locally with a make mockbuild before
submitting and all was fine.
The builds are:
http://koji.fedoraproject.org/koji/buildinfo?buildID=147755
http://koji.fedoraproject.org/koji/buildinfo?buildID=147756
Should I just cancel and resubmit, or?
I've seen this same thing in some other emacs packages builds.
Something just causes them to get stuck. ;(
I guess I would say try canceling and re-submitting, if that doesn't
work, try a local mock and see if you can see where it's getting stuck.
kevin
4:19 p.m.
On Sat, Dec 19, 2009 at 4:11 PM, Kevin Fenzi wrote:
On Sat, 19 Dec 2009 17:02:24 +0000
Jonathan Underwood wrote:
> Hi,
>
> I kicked off a couple of builds last night which appear stuck on koji.
> I had checked they build locally with a make mockbuild before
> submitting and all was fine.
>
> The builds are:
>
> http://koji.fedoraproject.org/koji/buildinfo?buildID=147755
> http://koji.fedoraproject.org/koji/buildinfo?buildID=147756
>
> Should I just cancel and resubmit, or?
I've seen this same thing in some other emacs packages builds.
Something just causes them to get stuck. ;(
There is a -O2 flag issue that causes some builds to hang. Maybe those
emacs hangs are related to this bug in gcc
It happened to 2 of my packages. Removing -O2 makes things compile in
my packages. (No I didn't do official builds without -O2 flag)
Someone filed a bug about this a while ago but unfortunately gcc
maintainers don't seem to care much. Anything we can do about this?
Can we use -O1 instead? The fact is these packages will not compile if
there will be a mass rebuild. And this needs to be fixed at some point
before F-13 is out.
Orcan
PS: Some references:
https://bugzilla.redhat.com/show_bug.cgi?id=548826
https://bugzilla.redhat.com/show_bug.cgi?id=531218
https://bugzilla.redhat.com/show_bug.cgi?id=538233
https://bugzilla.redhat.com/show_bug.cgi?id=539636
There are a couple more. Just search for gcc+hangs+slow etc.
5:30 p.m.
On Sat, 19 Dec 2009 17:19:37 -0500, Orcan wrote:
On Sat, Dec 19, 2009 at 4:11 PM, Kevin Fenzi wrote:
> On Sat, 19 Dec 2009 17:02:24 +0000
> Jonathan Underwood wrote:
>
>> Hi,
>>
>> I kicked off a couple of builds last night which appear stuck on koji.
>> I had checked they build locally with a make mockbuild before
>> submitting and all was fine.
>>
>> The builds are:
>>
>> http://koji.fedoraproject.org/koji/buildinfo?buildID=147755
>> http://koji.fedoraproject.org/koji/buildinfo?buildID=147756
>>
>> Should I just cancel and resubmit, or?
>
> I've seen this same thing in some other emacs packages builds.
> Something just causes them to get stuck. ;(
>
There is a -O2 flag issue that causes some builds to hang. Maybe those
emacs hangs are related to this bug in gcc
They just seem to hang.
It happened to 2 of my packages. Removing -O2 makes things compile
in
my packages. (No I didn't do official builds without -O2 flag)
Someone filed a bug about this a while ago but unfortunately gcc
maintainers don't seem to care much.
There has been activity in bugzilla.redhat.com related to this, however.
See e.g. http://bugz.fedoraproject.org/gcc and look for "eternity":
http://bugzilla.redhat.com/532763
The temporary work-around is to compile with
-fno-var-tracking-assignments
and that also works for lv2-c++-tools in the review queue, btw.
6:40 p.m.
On Sat, Dec 19, 2009 at 6:30 PM, Michael Schwendt wrote:
The temporary work-around is to compile with
-fno-var-tracking-assignments
and that also works for lv2-c++-tools in the review queue, btw.
Thanks! Yes, with that flag I was able to finish compiling muse and
lv2-c++-tools. Will that break the debuginfo package or anything else?
Orcan
4:23 a.m.
On Sat, 19 Dec 2009 19:40:36 -0500, Orcan wrote:
On Sat, Dec 19, 2009 at 6:30 PM, Michael Schwendt wrote:
>
> The temporary work-around is to compile with
>
> -fno-var-tracking-assignments
>
> and that also works for lv2-c++-tools in the review queue, btw.
>
Thanks! Yes, with that flag I was able to finish compiling muse and
lv2-c++-tools. Will that break the debuginfo package or anything else?
Well, it switches something off that is supposed to be an attempt at
improving debug information when using optimisations. How important
(read "how successful in practice") this feature is, I don't know,
and I could rather quit writing this mail and delete it instead. ;)
[Our default %optflags are just a compromise with regard to debug
information anyway.]
5:32 p.m.
On 09-12-19 17:19:37, Orcan Ogetbil wrote:
...
There is a -O2 flag issue that causes some builds to hang. Maybe
those emacs hangs are related to this bug in gcc
It happened to 2 of my packages. Removing -O2 makes things compile in
my packages. (No I didn't do official builds without -O2 flag)
Someone filed a bug about this a while ago but unfortunately gcc
maintainers don't seem to care much. Anything we can do about this?
Can we use -O1 instead? The fact is these packages will not compile
if there will be a mass rebuild. And this needs to be fixed at some
point before F-13 is out.
`man gcc` suggests that one can find out the differences between -On
levels with:
$ gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
$ gcc -c -Q -O1 --help=optimizers > /tmp/O1-opts
$ diff /tmp/O{1,2}-opts | grep enabled
(and the man page also lists differences).
Perhaps only one of the added optimizations is the problem and
disabling that one will fix it. A bug report about that particular
optimization on a particular file might get better results. I suppose
I'd disable half of the added optimizations at a time on a problem file
until I found the one that causes the hangs.
--
____________________________________________________________________
TonyN.:' <mailto:tonynelson@georgeanelson.com>
' <http://www.georgeanelson.com/>
4:40 a.m.
2009/12/19 Kevin Fenzi <kevin(a)scrye.com>:
I've seen this same thing in some other emacs packages builds.
Something just causes them to get stuck. ;(
I guess I would say try canceling and re-submitting, if that doesn't
work, try a local mock and see if you can see where it's getting stuck.
Well, strangely, I did a mockbuild locally before dropping the job
into koji and it completed fine. I see later posts refer to a gcc
problem, so I'll try playing with those - I think I'll need to rebuild
emacs locally as well though with different debug options to really
get to the bottom of it.
J
5279
days inactive
5280
days old
8 comments
6 participants
participants (6)
-
Jonathan Underwood -
Kevin Fenzi -
Kevin Kofler -
Michael Schwendt -
Orcan Ogetbil -
Tony Nelson