[Bug go/64999] s390x libgo test failure in TestMemoryProfiler

[ian at airs dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64999

--- Comment #23 from Ian Lance Taylor <ian at airs dot com> ---
Yes, I do mean to change saveg in mprof.goc.

runtime_callers in general returns full file/line information, which is
required for full correctness when using gccgo.  When it devolves back to a
plain PC, you've correctly observed that it gets the wrong PC.  I'm suggesting
correcting the PC value wherever we devolve, which happens in runtime.Callers
and, as you've pointed out, in saveg in mprof.goc.

I looked for other calls to runtime_callers, and found one other case: in
runtime_sigprof in proc.c.  I think it's worth trying fixing those cases to see
if it fixes this bug.