On Thu, Apr 13, 2023 at 9:06 AM Richard Purdie <[email protected]> wrote: > > On Thu, 2023-04-13 at 08:29 -0700, Khem Raj wrote: > > On Thu, Apr 13, 2023 at 6:44 AM Richard Purdie > > <[email protected]> wrote: > > > > > > On Thu, 2023-04-13 at 06:16 -0700, Khem Raj wrote: > > > > On Thu, Apr 13, 2023 at 3:14 AM Richard Purdie > > > > <[email protected]> wrote: > > > > > > > > > > On Wed, 2023-04-12 at 22:54 -0700, Khem Raj wrote: > > > > > > Signed-off-by: Khem Raj <[email protected]> > > > > > > --- > > > > > > ...0001-Provide-fixes-for-tests-on-musl.patch | 73 > > > > > > +++++++++++++++++++ > > > > > > meta/recipes-extended/gawk/gawk_5.2.1.bb | 7 +- > > > > > > 2 files changed, 77 insertions(+), 3 deletions(-) > > > > > > create mode 100644 > > > > > > meta/recipes-extended/gawk/gawk/0001-Provide-fixes-for-tests-on-musl.patch > > > > > > > > > > > > diff --git > > > > > > a/meta/recipes-extended/gawk/gawk/0001-Provide-fixes-for-tests-on-musl.patch > > > > > > > > > > > > b/meta/recipes-extended/gawk/gawk/0001-Provide-fixes-for-tests-on-musl.patch > > > > > > new file mode 100644 > > > > > > index 0000000000..4ff2466c0e > > > > > > --- /dev/null > > > > > > +++ > > > > > > b/meta/recipes-extended/gawk/gawk/0001-Provide-fixes-for-tests-on-musl.patch > > > > > > @@ -0,0 +1,73 @@ > > > > > > +From 95bcf4397b6f727208f3c627b7d00dacec406d02 Mon Sep 17 00:00:00 > > > > > > 2001 > > > > > > +From: Khem Raj <[email protected]> > > > > > > +Date: Wed, 12 Apr 2023 21:57:18 -0700 > > > > > > +Subject: [PATCH] Provide fixes for tests on musl > > > > > > + > > > > > > +Fixes rebt8b1 and regx8bit > > > > > > +Disable clos1way6 [1] > > > > > > + > > > > > > +Disable commas and backsmalls1, they do not work > > > > > > +with musl-locales somehow even though we do isntall > > > > > > +en_US.UTF-8 on target > > > > > > + > > > > > > +[1] > > > > > > https://lists.gnu.org/archive/html/bug-gawk/2021-02/msg00005.html > > > > > > + > > > > > > +Upstream-Status: Inappropriate [musl-specific] > > > > > > +Signed-off-by: Khem Raj <[email protected]> > > > > > > +--- > > > > > > + test/Maketests | 11 ++++------- > > > > > > + > > > > > > > > > > Carrying a patch like this makes me a bit nervous as: > > > > > > > > > > a) it is only applied for musl so it makes recipe upgrade testing a > > > > > nightmare > > > > > > > > > > b) there doesn't appear to be any path to upstream it, we'd just have > > > > > to carry it indefinitely. > > > > > > > > > > c) the reasons for the failure don't seem to be understood. > > > > > > > > reasons are clear, these tests need full locale support which we do > > > > not have with musl. > > > > > > The above says something quite different to what the patch says: > > > > > > "they do not work with musl-locales somehow" > > > > yeah the metadata patch subject should be improved to say disable tests > > That would be clearer, yes. > > > > > > > which sounds like nobody has looked into it. > > > > yes, and my localedata knowledge is limited to root cause this further > > sadly. > > Fair enough, we just need to be clear about that. > > > > > > Is there something we can do to avoid having to carry this > > > > > indefinitely? > > > > You can do sed operations from the recipe. I don't know how that will be > > better > > than patch. btw. we are already carrying another patch to disable other > > tests > > because I guess they fail on Autobuilder on stress but not always, > > because they worked > > fine on my ptest run. > > > > https://git.openembedded.org/openembedded-core/tree/meta/recipes-extended/gawk/gawk/remove-sensitive-tests.patch > > > > Do we have plans to address this with upstream as well ? > > The difference between that patch and the one proposed is that the one > above: > > a) is unconditional > b) is clear and simple, it just deletes the tests > > This means when the recipe is upgraded, the patch is obvious and > therefore straight forward to port. > > Ideas: > > Can we simplify the musl one to delete the tests in question? > > Could we get a patch upstream which allowed us to disable tests from > the commandline to improve both cases? > > I think we should be able to work something out but I just wanted to > make it clear the patch as it stands isn't clear enough and will cause > maintenance issues so we need to work out something better. >
if you run make check on musl/emux86-64 then we only get one failure clos1way6 and others pass. Which matches alpine. However, I think the way ptest is copying partially the test environment seems to be the cause. I tried to pass LANG=C etc but it did not help. Secondly, there are some surgeries the run-ptest is also doing which makes the test environment quite unique to yocto ptests. Its fine to ignore this patch. I have already drove into weeds trying to fix it, not worth it. > Cheers, > > Richard > > > > > > > Out of interest did alpine do anything for this? > > > > > > > > they have patches. > > > > > > The same as ours or different? Was there any discussion with upstream > > > on this? > > > > There is one failure which is common and I have quoted upstream discussion > > others are not. I dont have alpine install handy to try it out quickly > > to confirm. > > > > > > > > Cheers, > > > > > > Richard > > > > > >
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#179991): https://lists.openembedded.org/g/openembedded-core/message/179991 Mute This Topic: https://lists.openembedded.org/mt/98235523/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
