Re: [PATCH] mips/o32: fix internal_syscall5/6/7

["Maciej W. Rozycki" <macro at imgtec dot com>]

On Tue, 22 Aug 2017, Aurelien Jarno wrote:

> >  This passes o32 regular MIPS regression testing, however my GCC 8 binary 
> > seems to miscompile support_test_main.c in the MIPS16 mode, causing all 
> > the relevant test cases to crash with SIGILL (in the `support_set_test_dir 
> > (test_dir)' invocation it's `test_dir' that is jumped to rather than 
> > `support_set_test_dir'; I'll file a bug separately).  I'll find a way to 
> > run MIPS16 testing eventually, however meanwhile I will appreciate if you 
> > do it for me.
> 
> Thanks, this looks fine for me. I don't have more comments than the ones
> already done by Adhemerval. I have tested it on MIPS O32 BE and LE, I
> don't see any regression.

 Great, thanks!  I'll yet make the adjustments Adhemerval suggested.

> I can do a test build for MIPS16 and run basic testing under QEMU, but I
> don't have the hardware to do more.

 Thanks.  I have now identified the cause of my code generation problem 
(it was actually a register encoding bug in LD relaxation I have 
implemented in preparation to handle PR ld/21375, and which I thought I 
had removed from the binutils build used for this verification), so I will 
be able to run MIPS16 testing myself once I have the updated patch ready.

  Maciej