On Thu, 11 Dec 2025 at 18:56, Ross Burton via lists.openembedded.org
<[email protected]> wrote:
> -# Needed to set up qemu wrapper below
> -export STAGING_DIR_HOST
...
> -    qemu_binary="${@qemu_wrapper_cmdline(d, '$STAGING_DIR_HOST', 
> ['$STAGING_DIR_HOST/${libdir}','$STAGING_DIR_HOST/${base_libdir}'])}"
> +    qemu_binary="${@qemu_wrapper_cmdline(d, '${STAGING_DIR_HOST}', 
> ['${STAGING_LIBDIR}','${STAGING_BASELIBDIR}'])}"
>      cat > ${WORKDIR}/meson-qemuwrapper << EOF

I didn't want to take time in the patch review, but I went and checked
how this export appeared in the first place. It happened in 2016, when
using qemu to run target binaries was introtroduced to enable
cross-built gobject introspection:

https://git.openembedded.org/openembedded-core/commit/?id=c54c5fc63c05d416f3733a742fc94f7a169adc8b

I think Alex-2016 simply didn't realize that bitbake can substitute
its own variables in shell functions directly, and everything in there
must be 'exported' into the shell and deferred until shell runs the
script. And then the proven-to-work wrapper code was mindlessly copy
pasted around to places like gtk-doc or meson. So similar cleanups can
probably happen in gtk-doc class and g-i recipe as well.

Alex
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#227752): 
https://lists.openembedded.org/g/openembedded-core/message/227752
Mute This Topic: https://lists.openembedded.org/mt/116733177/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to