[PATCH] aarch64: Fix ICE happening in SET_TYPE_VECTOR_SUBPARTS with libgccjit

[Antoni Boucher]

The patch was approved by Richard Sandiford on the forge 
(https://forge.sourceware.org/gcc/gcc-TEST/pulls/29#issuecomment-490).

Le 2024-12-05 à 21 h 35, Antoni Boucher a écrit :
> Hi.
> This is a patch for the bug 117923.
> I'd like to know if there's a simpler fix for this.
> I tried keeping all the fields in the same struct and annotating the 
> scalar fields as in:
> 
> enum GTY((skip)) machine_mode mode;
> 
> but it didn't fix the issue.
> Any other ideas?
> 
> The following test suites passed for this patch:
> 
>   * The aarch64 tests.
>   * The aarch64 regression tests.
> 
> Here are the test results:
> 
>                  === gcc Summary ===
> 
> # of expected passes            382150
> # of unexpected failures        146
> # of unexpected successes       16
> # of expected failures          1946
> # of unresolved testcases       48
> # of unsupported tests          5465
> /home/antoyo/gcc-build/build/gcc/xgcc  version 15.0.0 20241203 
> (experimental) (GCC)
> 
> Same on the master branch:
> 
> # Comparing directories
> ## Dir1=/home/antoyo/tmp/gcc-compare/master/: 2 sum files
> ## Dir2=/home/antoyo/tmp/gcc-compare/fix-aarch64/: 2 sum files
> 
> # Comparing 2 common sum files
> ## /bin/sh ./contrib/compare_tests  /tmp/gxx-sum1.32352 /tmp/gxx-sum2.32352
> # No differences found in 2 common sum files
> 
> I also ran the jit tests on Aarch64 with another patch and there are 
> much less failures now.
> 
> Thanks for the review.