PATCH: bug in find_reloads_address_part

Ian Lance Taylor iant@google.com
Wed Aug 1 13:54:00 GMT 2007


Richard Sandiford <richard@codesourcery.com> writes:

> I thought that too when I first saw it, because I assumed that the
> second argument would be stored.  But as far as I can tell, that isn't
> true; the pointer never escapes.  It would be wrong to pass loc rather
> than "&tem" because loc is not yet a memory reference, whereas "*memrefloc"
> must be.

Ah, yes, you're right.

Sandra Loosemore <sandra@codesourcery.com> writes:

> Richard's understanding of this issue matches mine.  FWIW, here's the
> version of the patch I'm currently testing.  Assuming this passes, is
> it OK to commit?

Yes.

Thanks.

Ian



More information about the Gcc-patches mailing list