On Mon, 15 Jun 2026 at 04:04, Mingyu Wang (Fujitsu) <[email protected]> wrote:
>
> Thank you for the suggestion. I tried patching out the : uint8_t part from
> enum jay_predication as you suggested, but that alone is not enough. After
> removing it, the build immediately fails with a static assertion:
>
> static_assert(sizeof(jay_inst) == 32 + (sizeof(uintptr_t) * 2), "packed");
> because the struct layout depends on the enum's underlying type.
You also need to adjust the type of the struct member, so that the
struct itself enforces the type:
@@ -634,7 +634,7 @@ typedef struct jay_inst {
bool decrement_dep:1;
unsigned padding :12;
- enum jay_predication predication;
+ uint8_t predication;
enum jay_conditional_mod conditional_mod;
jay_def cond_flag; /**< conditional flag */
Or just take the full patch from here :)
https://git.openembedded.org/openembedded-core-contrib/plain/meta/recipes-graphics/mesa/files/0001-src-intel-compiler-jay-jay_ir.h-do-not-used-typed-en.patch?h=akanavin/package-version-updates
I just confirmed that with the patch gcc-11 builds mesa-native successfully.
> Beyond that, there are additional errors in other files that appear one after
> another when compiled with GCC 11:
>
> src/imagination/pco/pco_nir_pvfio.c: function lower_pfo returns false but
> should return nir_def *, causing error: incompatible types when returning
> type '_Bool' but 'nir_def *' was expected
This (imagination) is not enabled by default in poky or oe-core, and
does not appear in my build logs for mesa-native. You should check and
adjust your configuration, or perform test builds using exact
unmodified poky or oe-core.
> src/intel/compiler/jay/jay_from_nir.c: missing return in a non-void function,
> leading to error: control reaches end of non-void function
This is the same error as the typed enum issue above.
> I also tried passing -Dc_std=gnu2x and -Dwerror=false to Meson, but they were
> not enough to get past all the failures, and the build never completed
> successfully. It's therefore unclear how many more such issues might still be
> hidden – we simply weren't able to reach a clean build with GCC 11.
I just succeeded in that :) so hope you can take the patch and make it work.
Alex
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#238830):
https://lists.openembedded.org/g/openembedded-core/message/238830
Mute This Topic: https://lists.openembedded.org/mt/119737187/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
