[Gluster-devel] Compilation with gcc 7.x

[Niels de Vos]

On Wed, Jul 05, 2017 at 02:26:11PM +0530, Amar Tumballi wrote:
> On Wed, Jul 5, 2017 at 2:09 PM, Niels de Vos <ndevos at redhat.com> wrote:
> 
> > On Wed, Jul 05, 2017 at 12:45:11PM +0530, Amar Tumballi wrote:
> > > Csaba,
> > >
> > > Thanks for looking into this.
> > >
> > > On Tue, Jul 4, 2017 at 5:30 PM, Csaba Henk <chenk at redhat.com> wrote:
> > >
> > > > Hi list,
> > > >
> > > > I've compiled glusterfs with gcc 7.x (to be precise, with 7.1.1),
> > > > which is soon to get its prime time as the C compiler of
> > > > Fedora 26.
> > > >
> > > > The Release Notes (https://gcc.gnu.org/gcc-7/changes.html)
> > > > give account of a broad list of new and improved warnings...
> > > > and that shows. While with gcc 6.x the only warning I had
> > > > is "lchmod is not implemented and will always fail", with
> > > > gcc 7.x I got 218 warnings alltogether. For reference, I
> > > > attach the excerpted warnings from the compilation output.
> > > >
> > > > Went through the logs, and I see it is in project's interest to fix
> > them.
> > >
> > > Technically, fixing these warnings would be good to reduce our coverity
> > > warnings too in many cases. I am all for it.
> > >
> > >
> > > > Are you aware of this? Is there any plan what to do about it?
> > > >
> > > >
> > > I was not aware of it. Thanks for pointing it out. I propose we to fix it
> > > before 4.0 release branch out, and start having  gcc7.x job compiling as
> > > part of smoke.
> > >
> > > We can keep the smoke job non-voting till some time, and can turn the
> > knob
> > > ON someday, say October 15th for start voting -1 on any warnings. Others,
> > > any comments?
> >
> > We have similar job that checks for certain string-format warnings [1].
> > I think it is enabled to vote, otherwise only very few will pay
> > attention to the test results.
> >
> > Instead of cluttering the patch reviews with non-voting, the gcc-7.1
> > compile can be sent to the mailinglist similar to the Coverity results?
> > Maybe with some simple statistics in the beginning of the email, showing
> > how many warnings/error have been detected? This can be a regular job,
> > just build the RPMs within a Fedora Rawhide mock environment. Once the
> > next gcc version is out, it will get used automatically too (and we'll
> > test building with the latest headers/libraries as well).
> >
> > Niels, Ack! Makes sense to add it to 8hour tests. But it will still have
> to not make the test state as 'Failed'. Mainly because this is known to be
> not fixed before October.
> 
> Nigel, Can we add gcc7.1 to our 8 hourly tests on master?

We can make it easier for Nigel if we provide a test-script that does
the work. It can be based on the strfmt_errors.sh script at
http://git.gluster.org/cgit/build-jobs.git/tree/build-gluster-org/scripts/strfmt_errors.sh

The main thing that needs adjusting is "mock -r fedora-rawhide-x86_64"
command and "grep -E ..." matching.

The job itself is a YAML file, but I'm still not sure yet what
options/values should be used for the different ways of executing.
http://git.gluster.org/cgit/build-jobs.git/tree/build-gluster-org/jobs/strfmt_errors.yml

It can probably be scraped together from other jobs in that directory.
Patches can be sent to Gerrit (use "git review") for the project listed
on https://review.gluster.org/#/admin/projects/build-jobs

HTH,
Niels


> 
> 
> > Thanks,
> > Niels
> >
> > 1. https://build.gluster.org/job/strfmt_errors/
> >
> 
> 
> 
> -- 
> Amar Tumballi (amarts)