On Fri, 2026-08-21 at 14:25 +0530, Sadineni, Harish via lists.openembedded.org
wrote:
> Even though the script filename includes ${RUST_TARGET_SYS}, for
> multilib SDKs both the 32-bit and 64-bit scripts end up in the same
> environment-setup directory.
>
> The SDK's environment-sourcing logic sources every '*.sh' file in
> that directory, so both scripts get sourced regardless of which
> target you're building for. Since 'CARGO_TARGET_<TRIPLE>_RUSTFLAGS'
> and 'CARGO_BUILD_TARGET' are set unconditionally in each script,
> whichever one is sourced last wins and that can silently override the
> 32-bit settings with the 64-bit ones (or vice versa). That's what
> causes the 32-bit Cargo build to fail.
Ok, can you point me to which "environment-sourcing logic" we're taking
about? Is this in the test code or within the SDK itself. I suspect we
should be fixing that issue...
>
> Putting each script in a target-specific subdirectory (environment-
> setup.d/${TARGET_SYS}/) ensures only the script matching the selected
> target gets sourced, so the two configurations no longer clobber each
> other.
>
Perhaps. I'd like to avoid the complexity if we can so I'd like to
understand more about where this is happening.
Cheers,
Richard
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#244503):
https://lists.openembedded.org/g/openembedded-core/message/244503
Mute This Topic: https://lists.openembedded.org/mt/120640029/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
