Re: [PATCH] Add xgetbv/xsetbv instructions support for precord.

[Hui Zhu <teawater@gmail.com>]

Thanks Jilin, Michael and Joel.

I think this patch is very good.  We can find the "xgetbv" and
"xsetbv" in "Intel® 64 and IA-32 Architectures Software Developer’s
Manual Volume 2B: Instruction Set Reference, N-Z" Vol. 2B 4-509 and
Vol. 2B 4-529.
And I test this patch with reverse test, everything is OK.

BTW, I think the prec code that handle "0f 01" is not very well.
Jilin, do you interest with make this code better?

Thanks,
Hui



On Sat, Oct 10, 2009 at 10:33, Joel Brobecker <brobecker@adacore.com> wrote:
>> Sorry, but I don't know how to verify it, please help me with it. When
>> I ran 'make check', it displayed some 'FAIL' messages even before my
>> patch, and I noticed runtest uses
>> /usr/share/dejagnu/baseboards/unix.exp, which is not given by
>> http://sourceware.org/gdb/wiki/ReversibleDebugging.  How to use
>> runtest to test precord.exp only?
>
> I will let Michael and Hui explain to you how to use the special
> board file precord.exp.
>
> The FAILs are expected, as is simply impossible to mark all failures
> for all platforms.  There are too many variables, such as your compiler,
> your kernel, etc etc etc.  So what you need to do is run the testsuite
> before and after your patch, saving the contents of gdb.sum after the
> first run, and then do a comparison of the two files, to make sure
> that your patch did not cause any unexpected change.  Some tests are
> a little unpredictable, and cause you to see some apparent regressions
> that may not be caused by your patch. You'll have to use your judgement
> in these cases.
>
>> No, I don't have an FSF  assignment on file, but I'd like to. How to
>> have an FSF assignment on file?
>
> I'll send you the form to fill in and send to the FSF immediately.
>
> --
> Joel
>