https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106813
Rainer Orth <ro at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |ro at gcc dot gnu.org
--- Comment #1 from Rainer Orth <ro at gcc dot gnu.org> ---
FWIW I had a similar patch in my local tree for years, but neglected to submit
it (among others because it lacked dumpregs support.
While going over the remaining go.test failures recently, I noted some that
might be related, so I completed the patch (attached). A few notes:
* I've used the customary __sun__ && __svr4__ guard for Solaris-specific code,
not __sun && __SVR4 which is only used in runtime/go-libmain.c.
* I've reused the Linux/x86_64 code in dumpregs for Solaris. Only a few
adjustments were necessary, but this still seemed better than replicating
the whole section.
* I've added SPARC support, too. Since gregs[] is almost identical between
32 and 64-bit, I've just used different formats for printing 32 and 64-bit
registers, again to avoid massive duplication.
* When testing on Debian/sparc64, the 64-bit version there reguired an ugly
variation: the general-purpose member of mcontext_t is called mc_gregs
instead of gregs, and the indices are named MC_* instead of REG_* (although
the names are identical with one exception). I have no idea what rid them
to do this, but at least my code does compile and run there.
The patch has been tested on i386-pc-solaris2.11, sparc-sun-solaris2.11,
x86_64-pc-linux-gnu, and sparc64-unknown-linux-gnu (32 and 64-bit each).
There were no regressions, but compared to a vanilla tree there where no new
PASSes either.
