Message ID | 20240830081819.2635680-1-martin.jansa@gmail.com |
---|---|
State | Accepted |
Headers | show |
Series | [meta-oe] gcab: keep buildpaths QA issue as a warning | expand |
Mikko: can you please explain why meson embeds absolute paths in glib-mkenum filename on purpose? Or were you seeing any other buildpaths than these 2? lib32-gcab/1.6/package $ grep -R styhead . ./usr/src/debug/lib32-gcab/1.6/libgcab/gcab-enums.c:#include "WORKDIR/build/../gcab-1.6/libgcab/gcab-file.h" ./usr/src/debug/lib32-gcab/1.6/libgcab/gcab-enums.c:#include "WORKDIR/build/../gcab-1.6/libgcab/gcab-folder.h" glib-mkenum supports also @basename@ since 2.22 which would avoid this QA issue @filename@ name of current input file @basename@ base name of the current input file (Since: 2.22) if we update ./libgcab/gcab-enums.c.etemplate:#include "@filename@" Cheers, On Fri, Aug 30, 2024 at 10:18 AM Martin Jansa <martin.jansa@gmail.com> wrote: > > INSANE_SKIP:gcab-src doesn't work with multilib where the package is > named ${PN}-src (e.g. lib32-gcab-src) and it's better to just lower > buildpaths from ERROR_QA to WARN_QA instead of skipping it completely, > because it's still an issue which should be fixed (at least to improve > hashserv efficiency if you don't care about reproducibility itself) > > The backport to scarthgap from: > https://lists.openembedded.org/g/openembedded-devel/message/111739 > should IMHO be just reverted there as buildpaths QA issues > aren't fatal in scarthgap. > > Signed-off-by: Martin Jansa <martin.jansa@gmail.com> > --- > meta-oe/recipes-gnome/gcab/gcab_1.6.bb | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/meta-oe/recipes-gnome/gcab/gcab_1.6.bb b/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > index 49c64f0ba6..0d9410b4d5 100644 > --- a/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > +++ b/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > @@ -24,4 +24,6 @@ PACKAGECONFIG[tests] = "-Dtests=true -Dinstalled_tests=true,-Dtests=false -Dinst > BBCLASSEXTEND = "native" > > # meson embeds absolute paths to generated files on purpose > -INSANE_SKIP:gcab-src += "buildpaths" > +# ERROR: QA Issue: File /usr/src/debug/gcab/1.6/libgcab/gcab-enums.c in package gcab-src contains reference to TMPDIR [buildpaths] > +ERROR_QA:remove = "buildpaths" > +WARN_QA:append = " buildpaths" > -- > 2.46.0 >
Hi, On Fri, Aug 30, 2024 at 10:40:53AM +0200, Martin Jansa wrote: > Mikko: can you please explain why meson embeds absolute paths in > glib-mkenum filename on purpose? > > Or were you seeing any other buildpaths than these 2? > > lib32-gcab/1.6/package $ grep -R styhead . > ./usr/src/debug/lib32-gcab/1.6/libgcab/gcab-enums.c:#include > "WORKDIR/build/../gcab-1.6/libgcab/gcab-file.h" > ./usr/src/debug/lib32-gcab/1.6/libgcab/gcab-enums.c:#include > "WORKDIR/build/../gcab-1.6/libgcab/gcab-folder.h" > > glib-mkenum supports also @basename@ since 2.22 which would avoid this QA issue > @filename@ name of current input file > @basename@ base name of the current input file (Since: 2.22) > if we update > ./libgcab/gcab-enums.c.etemplate:#include "@filename@" I was googling for this and found MRs and issues in meson github which pointed to embedding build time absolute paths on purpose to fix something. Sadly my laptop crashes (Thinkpad T14 gen 3 crashes when plugging to power lid closed) and I lost the browser tabs before writing the commit message. Your proposal could be the correct fix, I hope it works. Cheers, -Mikko > Cheers, > > On Fri, Aug 30, 2024 at 10:18 AM Martin Jansa <martin.jansa@gmail.com> wrote: > > > > INSANE_SKIP:gcab-src doesn't work with multilib where the package is > > named ${PN}-src (e.g. lib32-gcab-src) and it's better to just lower > > buildpaths from ERROR_QA to WARN_QA instead of skipping it completely, > > because it's still an issue which should be fixed (at least to improve > > hashserv efficiency if you don't care about reproducibility itself) > > > > The backport to scarthgap from: > > https://lists.openembedded.org/g/openembedded-devel/message/111739 > > should IMHO be just reverted there as buildpaths QA issues > > aren't fatal in scarthgap. > > > > Signed-off-by: Martin Jansa <martin.jansa@gmail.com> > > --- > > meta-oe/recipes-gnome/gcab/gcab_1.6.bb | 4 +++- > > 1 file changed, 3 insertions(+), 1 deletion(-) > > > > diff --git a/meta-oe/recipes-gnome/gcab/gcab_1.6.bb b/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > > index 49c64f0ba6..0d9410b4d5 100644 > > --- a/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > > +++ b/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > > @@ -24,4 +24,6 @@ PACKAGECONFIG[tests] = "-Dtests=true -Dinstalled_tests=true,-Dtests=false -Dinst > > BBCLASSEXTEND = "native" > > > > # meson embeds absolute paths to generated files on purpose > > -INSANE_SKIP:gcab-src += "buildpaths" > > +# ERROR: QA Issue: File /usr/src/debug/gcab/1.6/libgcab/gcab-enums.c in package gcab-src contains reference to TMPDIR [buildpaths] > > +ERROR_QA:remove = "buildpaths" > > +WARN_QA:append = " buildpaths" > > -- > > 2.46.0 > >
> Your proposal could be the correct fix, I hope it works. It builds without the QA issue, I haven't tested it in runtime as I don't use gcab anywhere it was just one of buildpaths failures in my world builds. On Fri, Aug 30, 2024 at 10:57 AM Mikko Rapeli <mikko.rapeli@linaro.org> wrote: > > Hi, > > On Fri, Aug 30, 2024 at 10:40:53AM +0200, Martin Jansa wrote: > > Mikko: can you please explain why meson embeds absolute paths in > > glib-mkenum filename on purpose? > > > > Or were you seeing any other buildpaths than these 2? > > > > lib32-gcab/1.6/package $ grep -R styhead . > > ./usr/src/debug/lib32-gcab/1.6/libgcab/gcab-enums.c:#include > > "WORKDIR/build/../gcab-1.6/libgcab/gcab-file.h" > > ./usr/src/debug/lib32-gcab/1.6/libgcab/gcab-enums.c:#include > > "WORKDIR/build/../gcab-1.6/libgcab/gcab-folder.h" > > > > glib-mkenum supports also @basename@ since 2.22 which would avoid this QA issue > > @filename@ name of current input file > > @basename@ base name of the current input file (Since: 2.22) > > if we update > > ./libgcab/gcab-enums.c.etemplate:#include "@filename@" > > I was googling for this and found MRs and issues in meson github > which pointed to embedding build time absolute paths on purpose to > fix something. > > Sadly my laptop crashes (Thinkpad T14 gen 3 crashes when plugging to power lid closed) > and I lost the browser tabs before writing the commit message. > > Your proposal could be the correct fix, I hope it works. > > Cheers, > > -Mikko > > > Cheers, > > > > On Fri, Aug 30, 2024 at 10:18 AM Martin Jansa <martin.jansa@gmail.com> wrote: > > > > > > INSANE_SKIP:gcab-src doesn't work with multilib where the package is > > > named ${PN}-src (e.g. lib32-gcab-src) and it's better to just lower > > > buildpaths from ERROR_QA to WARN_QA instead of skipping it completely, > > > because it's still an issue which should be fixed (at least to improve > > > hashserv efficiency if you don't care about reproducibility itself) > > > > > > The backport to scarthgap from: > > > https://lists.openembedded.org/g/openembedded-devel/message/111739 > > > should IMHO be just reverted there as buildpaths QA issues > > > aren't fatal in scarthgap. > > > > > > Signed-off-by: Martin Jansa <martin.jansa@gmail.com> > > > --- > > > meta-oe/recipes-gnome/gcab/gcab_1.6.bb | 4 +++- > > > 1 file changed, 3 insertions(+), 1 deletion(-) > > > > > > diff --git a/meta-oe/recipes-gnome/gcab/gcab_1.6.bb b/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > > > index 49c64f0ba6..0d9410b4d5 100644 > > > --- a/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > > > +++ b/meta-oe/recipes-gnome/gcab/gcab_1.6.bb > > > @@ -24,4 +24,6 @@ PACKAGECONFIG[tests] = "-Dtests=true -Dinstalled_tests=true,-Dtests=false -Dinst > > > BBCLASSEXTEND = "native" > > > > > > # meson embeds absolute paths to generated files on purpose > > > -INSANE_SKIP:gcab-src += "buildpaths" > > > +# ERROR: QA Issue: File /usr/src/debug/gcab/1.6/libgcab/gcab-enums.c in package gcab-src contains reference to TMPDIR [buildpaths] > > > +ERROR_QA:remove = "buildpaths" > > > +WARN_QA:append = " buildpaths" > > > -- > > > 2.46.0 > > >
diff --git a/meta-oe/recipes-gnome/gcab/gcab_1.6.bb b/meta-oe/recipes-gnome/gcab/gcab_1.6.bb index 49c64f0ba6..0d9410b4d5 100644 --- a/meta-oe/recipes-gnome/gcab/gcab_1.6.bb +++ b/meta-oe/recipes-gnome/gcab/gcab_1.6.bb @@ -24,4 +24,6 @@ PACKAGECONFIG[tests] = "-Dtests=true -Dinstalled_tests=true,-Dtests=false -Dinst BBCLASSEXTEND = "native" # meson embeds absolute paths to generated files on purpose -INSANE_SKIP:gcab-src += "buildpaths" +# ERROR: QA Issue: File /usr/src/debug/gcab/1.6/libgcab/gcab-enums.c in package gcab-src contains reference to TMPDIR [buildpaths] +ERROR_QA:remove = "buildpaths" +WARN_QA:append = " buildpaths"
INSANE_SKIP:gcab-src doesn't work with multilib where the package is named ${PN}-src (e.g. lib32-gcab-src) and it's better to just lower buildpaths from ERROR_QA to WARN_QA instead of skipping it completely, because it's still an issue which should be fixed (at least to improve hashserv efficiency if you don't care about reproducibility itself) The backport to scarthgap from: https://lists.openembedded.org/g/openembedded-devel/message/111739 should IMHO be just reverted there as buildpaths QA issues aren't fatal in scarthgap. Signed-off-by: Martin Jansa <martin.jansa@gmail.com> --- meta-oe/recipes-gnome/gcab/gcab_1.6.bb | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)