This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] start of C binary compatibility testsuite


On Thu, May 15, 2003 at 12:14:31PM -0700, Janis Johnson wrote:
> On Thu, May 15, 2003 at 06:46:26PM +0100, Joern Rennecke wrote:
> > >         * gcc.dg/compat/struct-by-value-5_main.c: New test file.
> > >         * gcc.dg/compat/struct-by-value-5_x.c: New test file.
> > >         * gcc.dg/compat/struct-by-value-5_y.c: New test file.
> > 
> > This one fails for my sh-elf test setup because I use 
> > -Wl,--defsym,_stack=0x7f000 (the default is even lower) and
> > the test is simply too large:
> > 
> > bash-2.05$ size gcc-dg-compat-struct-by-value-5-01
> >    text    data     bss     dec     hex filename
> >  475906    1984   43608  521498   7f51a gcc-dg-compat-struct-by-value-5-01
> > bash-2.05$ size c_compat_main_tst.o c_compat_x_tst.o c_compat_y_tst.o
> >    text    data     bss     dec     hex filename
> >      28       0       0      28      1c c_compat_main_tst.o
> >  349200       0       0  349200   55410 c_compat_x_tst.o
> >  117272       0       0  117272   1ca18 c_compat_y_tst.o
> > 
> > In the past, we have used STACK_SIZE to disable or modify tests
> > that are otherwise too memory-hungry for embedded targets.
> > OTOH, it appears that the test could split into multiple smaller
> > parts that would cause no problem.
> > Can we split this test into a float, a double and a long double
> > part?
> 
> Sure, I'll split it up and get the new tests in today or tomorrow.

It doesn't look as if splitting them up that way will reduce stack
requirements so I'll try other things, using an sh-elf cross compiler
for testing.  What's a reasonable stack limit for tests?

Janis


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]