Fix FTBFS on ppc64el with gcc-15
authorDebian Science Team <debian-science-maintainers@lists.alioth.debian.org>
Sun, 14 Sep 2025 09:34:16 +0000 (11:34 +0200)
committerSébastien Villemot <sebastien@debian.org>
Sun, 14 Sep 2025 09:34:16 +0000 (11:34 +0200)
Origin: upstream, https://github.com/OpenMathLib/OpenBLAS/pull/5409/commits/a3b9c933c59809027937c7708e1d996988c112f2
Bug: https://github.com/OpenMathLib/OpenBLAS/issues/5372
Bug-Debian: https://bugs.debian.org/1114016
Reviewed-by: Sébastien Villemot <sebastien@debian.org>
Last-Update: 2025-09-14

The test failures are caused by incorrect clobber rules in asm inline
statements.
This patch is only a temporary workaround until a permanent fix is found.
Last-Update: 2025-09-14
Gbp-Pq: Name ppc64el-gcc-15.patch

kernel/power/dgemv_n.c

index ac365b3b2af1b22de2b17281e821cb2731eac854..303d9600fba9d754c1a45eb11d6c7960cf3840a4 100644 (file)
@@ -147,7 +147,7 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha, FLOAT *a, BLASLO
        BLASLONG m3;
        BLASLONG n2;
        BLASLONG lda4 =  lda << 2;
-       FLOAT xbuffer[8] __attribute__ ((aligned (16)));
+       volatile FLOAT xbuffer[8] __attribute__ ((aligned (16)));
        FLOAT *ybuffer;
 
         if ( m < 1 ) return(0);