This is the mail archive of the 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: static-after-extern tests vs. mips explicit relocs

On Fri, May 09, 2003 at 08:53:48AM +0100, Richard Sandiford wrote:
> The patch below moves the problematic string-opt-* tests under the
> control of a new harness, builtins.exp.

I like the new harness.

>   - It removes a static qualifier from link_error() in string-opt-16.c.
>     (This hunk was in the last patch too.)

While this is ok, we still have 

  static int memcmp (const void *s1, const void *s2, size_t len)

below.  Seems best to move it to the new harness as well.

>   - It adds -mno-explicit-relocs to no-builtin-1.c for mips targets.
>     If this isn't acceptable, I can try adding a similar test to
>     builtins.exp.

Please do move it.  Seems trivial to do.

>   - It adds -mno-explicit-relocs for builtin-noret-*.c.  These tests
>     override _exit and _Exit, which I guess really have to be kept local.
>     And irix does need the _Exit definition.

Hmm.  I think these tests should be moved to gcc.dg where we
can specify { dg-do link }.  Then we don't have to care that
we are overriding both _exit and _Exit (and so the program
cannot run).

With that done, the functions won't have to be static, and
you won't have to have -mno-explicit-relocs added.


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