Re: Meson buildtype for ci jobs?

[Patrick Robb <probb@iol.unh.edu>]

[-- Attachment #1: Type: text/plain, Size: 4418 bytes --]

>
> On Wed, Mar 15, 2023 at 4:52 PM Patrick Robb <probb@iol.unh.edu> wrote:
>
> Are you sure about the default value?
>
> Afaics, meson selects by default a "release" buildtype (and I think I
> always saw this value in the past).
> I have this with meson 1.0.
>
> $ meson setup qsdlgfh
> ...
> $ meson configure qsdlgfh | grep buildtype
>   buildtype                      release
> [plain, debug, debugoptimized,   Build type to use
>
>
Thanks, you're right - I missed this in the meson.build file. So the
default is release, which is what we run at the lab since we never pass a
custom buildtype option in at meson setup.


> >
> > The reason I'm writing this email is that I'm wondering whether the
> buildtype decision made by those who wrote .ci/linux-build.sh for GHA was
> intentional and important. I know many of the people who have commits on
> that script follow this mailing list. Obviously if it's in some way more
> appropriate for CI purposes to run meson setup in this way, I'm happy to
> make that change at the lab and in the process that would free up bringing
> Alpine compile testing online. But, if not, then I think it's most
> appropriate to consider compile on Alpine as broken and avoid bringing
> coverage for Alpine online until that issue is resolved.
>
> The reason why we went with debugoptimized was primarly for the ABI
> checks, as by default, the debug symbols were missing (which would
> match with a "release" default buildtype).
> See 777014e56d07 ("devtools: add ABI checks").
>

Gotcha. Thanks that makes sense. Well, I guess what I had (unknowingly)
done initially with our Alpine container at the lab was a release build,
but I'll do a release build on GHA to sanity check again anyways, which
will presumably fail.

So, I guess the conclusion is if it doesn't build with the default
buildtype set by meson.build, it's broken and we will not bring Alpine
compile coverage online right now.

Thanks, let me know if you think I'm missing anything here! -Patrick

On Wed, Mar 15, 2023 at 12:08 PM David Marchand <david.marchand@redhat.com>
wrote:

> On Wed, Mar 15, 2023 at 4:52 PM Patrick Robb <probb@iol.unh.edu> wrote:
> >
> > Hello all,
> >
> > The lab recently received a request to re-enable Alpine compile jobs,
> which have been disabled for almost a year. In dry running the compile job,
> I noticed that it was failing. At the same time, David Marchand did an
> Alpine compile with Github Actions which was successful. It seems the
> source of the different behavior is the meson buildtype being used - the
> build script used by GHA sets meson buildtype to debugoptimized, whereas
> the script used by the community lab runs with buildtype debug (the meson
> default). I did my own Github Actions runs (with both buildtype options) to
> sanity check:
> https://github.com/PatrickRobbIOL/dpdk/actions/runs/4427160204/jobs/7764368640
>
> Are you sure about the default value?
>
> Afaics, meson selects by default a "release" buildtype (and I think I
> always saw this value in the past).
> I have this with meson 1.0.
>
> $ meson setup qsdlgfh
> ...
> $ meson configure qsdlgfh | grep buildtype
>   buildtype                      release
> [plain, debug, debugoptimized,   Build type to use
>
> >
> > The reason I'm writing this email is that I'm wondering whether the
> buildtype decision made by those who wrote .ci/linux-build.sh for GHA was
> intentional and important. I know many of the people who have commits on
> that script follow this mailing list. Obviously if it's in some way more
> appropriate for CI purposes to run meson setup in this way, I'm happy to
> make that change at the lab and in the process that would free up bringing
> Alpine compile testing online. But, if not, then I think it's most
> appropriate to consider compile on Alpine as broken and avoid bringing
> coverage for Alpine online until that issue is resolved.
>
> The reason why we went with debugoptimized was primarly for the ABI
> checks, as by default, the debug symbols were missing (which would
> match with a "release" default buildtype).
> See 777014e56d07 ("devtools: add ABI checks").
>
>
> --
> David Marchand
>
>

-- 

Patrick Robb

Technical Service Manager

UNH InterOperability Laboratory

21 Madbury Rd, Suite 100, Durham, NH 03824

www.iol.unh.edu

[-- Attachment #2: Type: text/html, Size: 7542 bytes --]