x11-wm/muffin: added workaround for -fno-semantic-interposition

[asaparov]

Added -fno-semantic-interposition workaround for x11-wm/muffin to fix compilation error due to maybe-uninitialized warnings and -Werror. Earlier versions of muffin are unaffected.

[InBetweenNames]

No errors here. Could you post a build.log?

[asaparov]

This one I just retested and, interestingly, it fails with -O3 -fno-semantic-interposition but not with -O2 -fno-semantic-interposition. Attaching build.log:
build.log

[InBetweenNames]

Very interesting! It looks like you aren't actually passing -flto in your CFLAGS. Is this in your workarounds by any chance?

[asaparov]

Ah no I've disabled system-wide LTO for now. I have no other workarounds for muffin or graphene.

[asaparov]

After some more testing I found its the combination of -finline-functions and -fno-semantic-interposition that causes the error on my machine.

[InBetweenNames]

Just letting you know, I've encountered at least one other user getting -Werror-style errors for may be uninitialized. I'm looking into it.

[InBetweenNames]

Still not able to reproduce this. Closing for now -- please comment if you still need this workaround and I'll reopen.

[asaparov]

I'm not using muffin on any machine any more, but I just did try to re-emerge version 4.6.3 without the workaround and I get the compiler errors:

cogl-pipeline-layer-state.c: In function ‘cogl_pipeline_get_layer_min_filter’:
cogl-pipeline-layer-state.c:1546:10: error: ‘min_filter’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
 1546 |   return min_filter;
      |          ^~~~~~~~~~
cogl-pipeline-layer-state.c: In function ‘cogl_pipeline_get_layer_mag_filter’:
cogl-pipeline-layer-state.c:1558:10: error: ‘mag_filter’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
 1558 |   return mag_filter;
      |          ^~~~~~~~~~
...
cogl-pipeline.c: In function ‘_cogl_pipeline_prune_to_n_layers’:
cogl-pipeline.c:1689:10: error: ‘state.first_index_to_prune’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
 1689 |       if (layer->index >= state.first_index_to_prune)
      |          ^
cc1: some warnings being treated as errors

But this is still with Ofast and no LTO. I don't know if you think we should re-open this.