Firstly, we adjust the test conditions to use non-PIC code for Darwin.
Secondly, we have to account for out-of-line GPR restores which gives
a false positive on one of the scan-assembler-not. Lastly, we make the
test a bit more specific for Darwin - that it looks for absence of
branches to local labels.
2019-06-23 Iain Sandoe <iain@sandoe.co.uk>
* gcc.target/powerpc/pr71785.c: For Darwin, make test non-PIC,
expect the out-of-line GPR restore, and test specifically for
absence of branches to local labels.
From-SVN: r272601
+2019-06-23 Iain Sandoe <iain@sandoe.co.uk>
+
+ * gcc.target/powerpc/pr71785.c: For Darwin, make test non-PIC,
+ expect the out-of-line GPR restore, and test specifically for
+ absence of branches to local labels.
+
2019-06-22 Jerry DeLisle <jvdelisle@gcc.gnu.org>
PR fortran/89782
/* { dg-do compile { target { powerpc*-*-* } } } */
/* { dg-options "-O2" } */
-/* { dg-final { scan-assembler-not {\mb\M} } } */
+/* We have to lose the default pic codegen on Darwin. */
+/* { dg-additional-options "-mdynamic-no-pic" { target powerpc*-*-darwin* } } */
+/* ... and account for the out-of-line GPR restore. */
+/* { dg-final { scan-assembler-times {\mb[ \t]*restGPR} 1 { target powerpc*-*-darwin* } } } */
+/* { dg-final { scan-assembler-not {\mb[ \t]L} { target powerpc*-*-darwin* } } } */
+/* { dg-final { scan-assembler-not {\mb\M} { target { ! powerpc*-*-darwin* } } } } */
/* Check that all computed gotos in this testcase end up unfactored completely.
If some is not there will be a unconditional jump left; if all works fine,