| Message ID | 4EB98F59.7050600@gmail.com |
|---|---|
| State | New |
| Headers | show |
On 11/08/2011 12:21 PM, Patrick Marlier wrote:
> The other way could be to change the regex but we don't want that.
Why not?
r~
On 11/08/2011 04:29 PM, Richard Henderson wrote: > On 11/08/2011 12:21 PM, Patrick Marlier wrote: >> The other way could be to change the regex but we don't want that. > > Why not? My guess was that the purpose to have "long" in the testcase was to test the generation of another TM-builtin than _ITM_RU4. 3 possibilities to fix: - change g to "long long" - change g to "int" and match _ITM_RU4 for g - change the match for g to _ITM_RU[48] Patrick.
On 11/08/2011 02:08 PM, Patrick Marlier wrote:
> - change the match for g to _ITM_RU[48]
Change the match to [248].
r~
Index: testsuite/gcc.dg/tm/memopt-1.c =================================================================== --- testsuite/gcc.dg/tm/memopt-1.c (revision 181162) +++ testsuite/gcc.dg/tm/memopt-1.c (working copy) @@ -1,7 +1,7 @@ /* { dg-do compile } */ /* { dg-options "-fgnu-tm -O -fdump-tree-tmmemopt" } */ -long g, xxx, yyy; +long long g, xxx, yyy; extern george() __attribute__((transaction_callable)); extern ringo(long int); int i;
In x86/32bits, long is 4bytes, and the test checks for _ITM_RU8 (8bytes). I propose to change to long long (it is C99 but it should not be a problem, right?). The other way could be to change the regex but we don't want that. Tested on x86. Richard/Aldy, can you apply it to trunk and trans-mem? (of course if approved). It removes a unexpected failure on x86/32. Patrick. * testsuite/gcc.dg/tm/memopt-1.c: Use long long.