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]

Re: crashes building on Darwin using GCC trunk

"Marcus G. Daniels" wrote:
> I have made it through a bootstrap before on Darwin, but for the last
> week or two I've been seeing crashes like below.
> [...]
> /src/gcc/gcc/reload1.c:7317: Internal compiler error in reload_cse_simplify_operands, at reload1.c:8350

This is a known problem that I've been working on for the past several
days.  It's completely specific to Darwin, and is connected to Mach-O
PIC codegen, where in certain cases r0 should not be used for reloads.
The problem has been around since at least the beginning of April; I
think it's always been there, but a combination of other improvements
along with changed sources triggered the failure (I have a test case
cut down from reload1.c, but it's still pretty large; anything smaller
and the problem goes away.  Nothing in the testsuite triggers this

Anyway, there's no workaround for this, so I have to keep plugging away,
but since so far it's only been observed with GCC's own sources, my
advice to experimenters is to not worry about it when compiling your
own code.  And if anybody has any insights about how to keep the rs6000
backend from using r0 in addsi3_high patterns, I would appreciate any


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