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]

Re: [PATCH 2.95.3] Fix powerpc-wrs-vxworks eabi backwards incompatibi lity.


	I do not think that it is useful to duplicate most of the
gcc-2.7.2 SVR4/eABI ABI logic in gcc-2.95 or gcc-3.0.  Additionally
gcc-3.0 specifically is a binary incompatible break.

	gcc-2.7 included preliminary PowerPC support.  Most GCC users were
tracking the fixes in the development sources.  At this point in time I
feel that it will do more harm than good to burden the public repository
with maintaining the old, incorrect code which has not been utilized by
developers using the GCC sources for many years.  This seems to be a WRS
and possibly Red Hat support issue.

	I am happy to re-examine the longcall proposal for gcc-3.0/3.1 if
this patch is cleaner than the previous one.

	Also, the FP_ARGS portion needs some investigation.  Your patch
reverses both the setting of the cookie and the interpretation of the
cookie in the call patterns, so there should be no effect on the actual
code generated, right?

	I think that it would be great for you to provide correct versions
of your patch.  I also think that WRS should consider using it for a
compatible and improved compiler for use in their old release.  I am
surprised that WRS did not pursue this type of approach a long time ago.
This is exactly the flexibility that Free Software provides to developers
and users.

David

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