On 11/20/25 03:40, Khem Raj wrote:
>
>
> On Wed, Nov 19, 2025 at 10:17 AM Richard Purdie
> <[email protected]> wrote:
>
>     On Wed, 2025-11-19 at 09:29 -0800, Khem Raj via
>     lists.openembedded.org <http://lists.openembedded.org> wrote:
>     > On Wed, Nov 19, 2025 at 9:20 AM Gyorgy Sarvari
>     <[email protected]> wrote:
>     > > Hmm, I just noticed that this flag wasn't omitted accidentally
>     at all,
>     > > but because clang doesn't support it.
>     > >
>     > > Please ignore this patch, will try to come up with some
>     conditional
>     > > patch or something like that.
>     > >
>     >
>     >
>     >
>     > Right, the intention was to keep the DEBUG_PREFIX_MAP to be generic
>     > and I was hoping that packages won't need -fcanon-prefix-map but
>     that's
>     > not the case. In any case I think we now have toolchain classes
>     and these
>     > options are compiler specific so we could either move the whole
>     > assignment out of bitbake.conf and into toolchain classes
>     > or define addon variable in gcc.bbclass
>     >
>     > DEBUG_PREFIX_MAP_EXTRA = "-fcanon-prefix-map"
>     >
>     > and prepend it to DEBUG_PREFIX_MAP in bitbake.conf
>     >
>     > That will allow it to be overridden cleanly in recipes if needed.
>     >
>     > Third option is to keep it as such and add
>     >
>     > DEBUG_PREFIX_MAP =+
>     "-fcanon-prefix-map" meta/classes/toolchain/gcc.bbclass
>     >
>     > also keep in mind that this option is only available in gcc-13+
>     > so DEBUG_PREFIX_MAP if used on native recipes might fail for
>     hosts with
>     > gcc < 13
>     >
>
>     There is an additional issue in rust too, e.g.:
>
>     https://autobuilder.yoctoproject.org/valkyrie/#/builders/22/builds/2773
>
>
> Right, as I noted above in my reply, ibuild host is running debian-12
> and the gcc version on debian-12 is GCC-12 which does not
> have this option implemented. We need atleast gcc 13. Perhaps we
> should follow option 2 or 3, that will address the provlem
> surgically.

Yes, but I believe that we are also hitting this bug:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=15976

This log corroborates with this - bootstrapping is building some crates
for the host, it uses the host's compiler, but it adds the target flags
to it: "HOST_CFLAGS" uses no prefix-maps as prepared by
oe_cargo_fix_env(), but the failing call uses them. (And the failing
crate has cc-rs as a build dependency...)

This is getting into a rabbit hole :) Will stare at this a bit more.

>  
>
>     Cheers,
>
>     Richard
>

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#226604): 
https://lists.openembedded.org/g/openembedded-core/message/226604
Mute This Topic: https://lists.openembedded.org/mt/116371356/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to