This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug bootstrap/41771] Bootstrap with Sun Studio 12.1 fails
- From: "uweigand at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 5 Nov 2009 13:18:08 -0000
- Subject: [Bug bootstrap/41771] Bootstrap with Sun Studio 12.1 fails
- References: <bug-41771-279@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #10 from uweigand at gcc dot gnu dot org 2009-11-05 13:18 -------
(In reply to comment #9)
> IMHO, flags.h should not include real.h if things like gcc.c will include it
> and not link against real.o and -lgmp etc. But I don't know which if any
> targets now depend on this include. I tried x86_64-linux-gnu and I was able to
> link xgcc and cc1. But some other target may have a problem.
>
> Perhaps Ulrich can explain why he chose to add that include so we can see if
> there's another way to accomplish the goal.
flags.h contains macros like HONOR_NANS which are defined in terms of
macros like MODE_HAS_NANS. The MODE_HAS_... macros used to be defined
in target headers (or defaults.h) before my patch.
That patch changed the MODE_HAS_... macros to be defined in terms of
properties of the floating-point format structure (struct real_format)
instead of requiring the target to provide this information in a
redundant way via special target macros.
In order to be able to access struct real_format, those new MODE_HAS_...
macros needed to be defined in real.h. In order to keep the definition
of the HONOR_... macros in flags.h working, I added the include of real.h
to flags.h.
If this is problematic, I guess an alternative might be to move the
HONOR_... macros from flags.h to real.h as well, and update all files
that use these macros to also include real.h.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41771