This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: PowerPC64 __tls_get_addr_opt stub .eh_frame fix
- From: Alan Modra <amodra at gmail dot com>
- To: binutils at sourceware dot org
- Date: Wed, 1 Aug 2018 19:09:57 +0930
- Subject: Re: PowerPC64 __tls_get_addr_opt stub .eh_frame fix
- References: <20180801025226.GF26457@bubble.grove.modra.org>
The modified test failed on some powerpc targets due to differences in
default hash style. If the default hash style is both, then more
sections are created, bumping section ids. Section id is used in stub
symbols and although the test is careful to not depend on id in
labels, the stub hash traversal order changes when stub names change.
That lead to the stubs being emitted in a different order and thus not
matching expected output.
* testsuite/ld-powerpc/powerpc.exp: Run tlsopt5 with hash-style
specified.
diff --git a/ld/testsuite/ld-powerpc/powerpc.exp b/ld/testsuite/ld-powerpc/powerpc.exp
index ed7bc74cf4..ac563eba14 100644
--- a/ld/testsuite/ld-powerpc/powerpc.exp
+++ b/ld/testsuite/ld-powerpc/powerpc.exp
@@ -224,7 +224,7 @@ set ppc64elftests {
"tlsopt4"}
{"TLS DLL" "-shared -melf64ppc --version-script tlsdll.ver" "" "-a64" {tlsdll.s}
{} "tlsdll.so"}
- {"TLS opt 5" "-melf64ppc -shared --gc-sections --no-plt-localentry tmpdir/tlsdll.so" "" "-a64" {tlsopt5.s}
+ {"TLS opt 5" "-melf64ppc -shared --hash-style=both --gc-sections --no-plt-localentry tmpdir/tlsdll.so" "" "-a64" {tlsopt5.s}
{{objdump -dr tlsopt5.d} {readelf -wf tlsopt5.wf}}
"tlsopt5"}
{"sym@tocbase" "-shared -melf64ppc" "" "-a64" {symtocbase-1.s symtocbase-2.s}
--
Alan Modra
Australia Development Lab, IBM