[patch testsuite]: Skip some tests for w64 target
Kai Tietz
Kai.Tietz@onevision.com
Wed Feb 4 14:47:00 GMT 2009
Di Dave,
Dave Korn <dave.korn.cygwin@googlemail.com> wrote on 04.02.2009 15:36:24:
> Kai Tietz wrote:
>
> > I get these errors also for i686 mingw and cygwin cross-compilers for
w64.
> > So I assume it is more a general issue for all windows based targets.
> > Maybe the skip option needs even to be extended. Better indeed would
be,
> > that the recursion deepth for gimplifier would be lowered, but this is
out
> > of scope for 4.4.
>
> Kai, how about we add a target-specific -Wl,--stack= option to the
> testsuite instead? I've found that those testcases sometimes pass when
run
> manually from the commandline, for example:
>
> /gnu/gcc/obj-dw2/gcc/testsuite/gcc $ /win/i/FSF-Gcc/obj-dw2/gcc/xgcc
> -B/win/i/FSF-Gcc/obj-dw2/gcc/ -O3 -g -w -c -o limits-declparen.o
> /gnu/gcc/gcc/gcc/testsuite/gcc.c-torture/compile/limits-declparen.c
>
> /gnu/gcc/obj-dw2/gcc/testsuite/gcc $ echo $?
> 0
>
> So I suspect they're not missing by a great amount. (Interesting;
the
> default stack size is 2MB on cygwin, for win64 you have 8MB and it
> still fails
> in the testsuite, but it passes on the command-line with only 2MB. This
may
> be some kind of artifact of the testsuite environment.)
Well, this would be a way for windows based targets, too. IMHO the stack
size for cygwin should be the same as for mingw32 (and it is at the moment
8MB). For w64 itself there is a bit more stack necessary, because the abi
itself consumes much more space (beside the 8 byte pc word).
For 32-bit windows hosts the smallest stackframe of a function call is 4
bytes (just the return address). For w64 it needs at least 48 bytes for
the smallest function call (32-bytes save-area, 8 bytes return address,
and 8 bytes for stack re-alignment).
Cheers,
Kai
| (\_/) This is Bunny. Copy and paste Bunny
| (='.'=) into your signature to help him gain
| (")_(") world domination.
More information about the Gcc-patches
mailing list