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 (and question): PowerPC -vs- const section


On Wed, Dec 13, 2000 at 02:26:20PM -0700, Tom Tromey wrote:
> I believe the comment that I deleted in the appended patch is simply
> wrong.  Constant objects with pointers in libgcj don't get put into
> .data with this patch.  Relocations and shared libraries (seem to)
> continue to work.

Just because it worked doesn't mean it's right.  Use
objdump -p and see if there is a DT_TEXTREL entry.  If
so, startup perfomance will suffer.

That said, the definition you are changing _is_ incorrect.
Any such modifications to which section is chosen should
be done with rs6000_select_section.

> My question is: what sort of testing would be sufficient?  A
> no-regression run against the gcc test suite?  Something more?

I'm not sure exactly how to test the -mrelocatable stuff.
Someone more familiar with that ppc hack should answer.


r~

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