On Thu, 2023-11-16 at 14:10 +0100, Yoann CONGAL wrote:
> Hi Richard,
>
> Le jeu. 16 nov. 2023 à 13:13, Richard Purdie
> <[email protected]> a écrit :
> > On Thu, 2023-09-28 at 23:56 +0200, Yoann Congal wrote:
> > > From: Fawzi KHABER <[email protected]>
> > >
> > > Remove superfluous DEV_PKG_DEPENDENCY = "" previously used to bypass
> > > ${PN}-dev package RDEPENDS on empty&non-built ${PN}. DEV_PKG_DEPENDENCY
> > > applies RRECOMMENDS now, all workarounds are not needed anymore.
> > >
> > > Related to [YOCTO #6839] and [YOCTO #8222]
> > >
> > > Signed-off-by: Yoann CONGAL <[email protected]>
> > > Signed-off-by: Fawzi KHABER <[email protected]>
> > > ---
> > > meta/recipes-connectivity/bind/bind_9.18.19.bb | 1 -
> > > meta/recipes-core/musl/bsd-headers.bb | 1 -
> > > meta/recipes-core/musl/libssp-nonshared.bb | 1 -
> > > meta/recipes-core/newlib/newlib_git.bb | 3 ---
> > > meta/recipes-devtools/gcc/libgcc-common.inc | 4 ----
> > > meta/recipes-devtools/python/python3_3.11.5.bb | 1 -
> > > meta/recipes-graphics/mesa/libglu_9.0.3.bb | 3 ---
> > > meta/recipes-graphics/mesa/mesa.inc | 3 ---
> > > meta/recipes-graphics/xorg-lib/libpthread-stubs_0.5.bb | 1 -
> > > meta/recipes-graphics/xorg-lib/xtrans_1.5.0.bb | 1 -
> > > meta/recipes-graphics/xorg-proto/xcb-proto_1.16.0.bb | 1 -
> > > meta/recipes-graphics/xorg-proto/xorgproto_2023.2.bb | 3 +--
> > > meta/recipes-graphics/xorg-util/util-macros_1.20.0.bb | 3 +--
> > > meta/recipes-kernel/linux-libc-headers/linux-libc-headers.inc | 1 -
> > > meta/recipes-kernel/make-mod-scripts/make-mod-scripts_1.0.bb | 1 -
> > > meta/recipes-support/argp-standalone/argp-standalone_1.4.1.bb | 1 -
> > > 16 files changed, 2 insertions(+), 27 deletions(-)
> >
> > Firstly, sorry for taking as long to get to this.
>
> That is fine! This was sent really late in the dev cycle wrt the release...
>
> > The patch 1/2 looks
> > good and I'm happy to merge that. It does set us on the patch I
> > outlined originally so I think we're good there.
>
> Awesome, thanks!
>
> > I'm less sure about this second one. It appears to remove variables
> > which do actively change things and once they are removed, warnings
> > from the previous patch show up.
> >
> > I therefore think the second patch is incorrect and doesn't move us in
> > the direction we want?
>
> I would need to look into it to to confirm this but from your plan :
> https://bugzilla.yoctoproject.org/show_bug.cgi?id=6839#c7
> a) was to make warnings appear
> b) fixing the warnings
>
> We are now between a) and b) : warnings but no fix.
>
> Since the patch I finally integrated the fact that the CI must be warning
> free.
> So I guess, I'll come back with a series with this patch (a) _and the
> fixes (b)_.
> Does that sound good?
Yes. I think the changes in 2/2 actually create warnings where we don't
currently have any and you'd need to revert 2/2 to help with the fixes
in step b) above.
I did also do a bit more testing of 1/2 and it also needs a bit more
work unfortunately, it is the right direction but not quite there yet.
Firstly, it shows warnings like:
perl-modules rrecommends perl-module-warnings but perl-module-warnings won't be
built
which is because it is an RPROVIDES:
recipes-devtools/perl/perl_5.38.0.bb:RPROVIDES:${PN} += "perl-module-strict
perl-module-vars perl-module-config perl-module-warnings \
so we need to teach the check how to resolve and ignore those.
Secondly, it is only handling "self referencing" rrecommends.
That can be changed by updating the constraint to something like:
if not rrec_data or "FILES_INFO" not in rrec_data or
(rrec_data["FILES_INFO"] == "{}"
and not bb.utils.to_boolean(rrec_data.get("ALLOW_EMPTY", "0"),
False)):
oe.qa.add_message(messages, "rrecommends-non-existent", "{0}
rrecommends {1} but {1} won't be built".format(pkg, rrecommend))
but then a lot more warnings appear.
I think a lot of the new warnings would be fixed if the RPROVIDES issue
is accounted for.
So things are heading the right way but need a little work. I'd be
happy to merge 1/2 when it is ready since we can enable the warnings on
a per layer basis now once we fix the issues. 1/2 isn't quite ready yet
though, we need to fix the RPROVIDES issue.
Cheers,
Richard
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#190796):
https://lists.openembedded.org/g/openembedded-core/message/190796
Mute This Topic: https://lists.openembedded.org/mt/101647340/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
