Makes -S output more easily preprocessable -- otherwise, the __i686 in
__i686.get_pc_think.reg chokes things.
bootstrapped x86_64-linux for C/C++ native on Ubunutu Lucid (x86-64),
no diff in testsuite output before/after.
(c, c++, libgomp, libmudflap, libstdc++ tested.)
Also manually tested:
char *foo() {
return "foo";
}
with "cc -fPIC -S test.c -o - -m32" to make sure that the new name was
used as expected.
OK for google/main branch?
(I don't *think* there's a need to push this to google/integration,
it's not *that* critical. And frankly while I think the standard
naming is stupid, it's long-standing and I *assume* that this issue
has come up before and thought not-important-enough-to-change -- even
though AFAIK there's ~no down-side to changing since these thunks
aren't exposed. So, I'm not really proposing this for trunk, but if
somebody wants it... great! 8-)
chris
---
[gcc/ChangeLog.google-main]
2011-04-30 Chris Demetriou <c...@google.com>
* config/i386/i386.c (get_pc_thunk_name): Make 32-bit
thunk prefix be __x86.get_pc_thunk.
[gcc/ChangeLog.google-main]
2011-04-30 Chris Demetriou <c...@google.com>
* config/i386/i386.c (get_pc_thunk_name): Make 32-bit
thunk prefix be __x86.get_pc_thunk.
Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c (revision 173204)
+++ config/i386/i386.c (working copy)
@@ -8751,7 +8751,7 @@
gcc_assert (!TARGET_64BIT);
if (USE_HIDDEN_LINKONCE)
- sprintf (name, "__i686.get_pc_thunk.%s", reg_names[regno]);
+ sprintf (name, "__x86.get_pc_thunk.%s", reg_names[regno]);
else
ASM_GENERATE_INTERNAL_LABEL (name, "LPR", regno);
}
