> -----Original Message----- > From: Richard Purdie <[email protected]> > Sent: Monday, September 28, 2026 3:04 PM > To: Marko, Peter (FT D EU SK BFS1) <[email protected]>; > [email protected] > Cc: Khem Raj <[email protected]>; Ross Burton <[email protected]>; > Freihofer, Adrian (SI B PRO TI EAC CCP) <[email protected]> > Subject: Re: [PATCH v2] toolchain/clang: prepend target host-sys to > HOST_CC_ARCH > > On Mon, 2026-09-28 at 12:35 +0000, Marko, Peter wrote: > > This patch is sitting in master-next for some time already and is > > probably spending a minute or two of everyone's time on each patch > > review call. > > That is true. > > > Let's try to do a decision whether to merge it or drop it. > > I (without asking or rewording via AI) think this: > > > > I see following ways how to approach this clang-tidy problem: > > - make clang-tidy architecture aware upstream > > - add these flags to tell it the architecture (what this patch > > proposes and what was present in scarthgap) > > - add it only to recipes which use clang-tidy (or other similar clang > > tools) > > > > clang-tidy (and this also applies to some other clang support tools) > > is in principle architecture independent. > > Linter should not care about build arch, correct?, it just needs to > > know the paths to sources. > > So imho making it architecture dependent would go in a wrong > > direction, probably wasting time of whoever develops this and > > upstream clang maintainers. > > > > Explicit set of architecture globally looks like a better choice. > > Yocto currently does not set the triplet in build flags, because it's > > hardcoded in toolchain and thus "it works" (however as we see here, > > only in 99.9% of cases) > > For clang compiler it's a no-op (the value is the same as value that > > it has hardcoded during its compilation). > > And it helps all "arch-independent" clang tools to find the right > > headers. > > > > Having a special handling for clang-tidy in all recipes using it is > > certainly possible, but seems to be sub-par. > > Yocto should support clang-tidy workflows natively. > > I think several of us all have mixed feelings on it. On the one hand it > fixes a problem. On the other, the rest of clang doesn't need this and > these kinds of flags, once added, are very very hard to remove again. > > I guess we lack enough data to make an informed choice and are worried > about sending this in the wrong direction if we reject it, equally, we > don't really want to make a change like this as it doesn't feel right. > We are therefore struggling to reach a decision :/.
I didn't want to cause too much effort with this patch. So feel free to drop this one, it can wait for more demand than a single project. It's trivial to keep this in our product layer without fear of getting broken. Peter > > Even trying to articulate that in a way you can take some action on is > hard, hence it is sitting there :/ > > Cheers, > > Richard
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#246787): https://lists.openembedded.org/g/openembedded-core/message/246787 Mute This Topic: https://lists.openembedded.org/mt/120852316/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
