This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: i386 floating point test case integration with gdb.reverse test-suite


Hi,

please find my comments below.

1) There are some tests that look like this:
   gdb_test "info register st1" "st1 *1\t.*"
(with a tab in the regular expression).

I like the ones with the tab better -- its more exact, and it
makes certain that the value is really "1", and not eg. "1.1".
Could you make them all like that?  Just the single-digit values.
I'm not worried about the multi-digit ones.

Oza : I will correct this in next patch.

2) The file names should reflect the Intel architecture, since
eventually we will have process record for other architectures.
In fact, maybe these tests should go in gdb.arch instead of in
gdb.reverse, since all the other tests in gdb.reverse are
architecture independent.

How about testsuite/gdb.arch/i387-stack-reverse.exp, etc?

Oza : I will change the names.

3) I'm not an Intel expert, but you had mentioned a few
other registers in an earlier post, namely:
  fstat
  ftag
  fiseg
  fioff
  foseg
  fooff
  fop

Oza: second test case test some of above registers, such as fstat and ftag..these two registers are most frequently used as trackkers by intel.
In my opinion, If we test above it is sufficient as it gives us gurantee
of I387 FPU is working and recorded as expected. other registers are not of direct interest to us.

If you still feel then we can include one test.

Regards,
Oza.

--- On Mon, 8/17/09, Michael Snyder <msnyder@vmware.com> wrote:

> From: Michael Snyder <msnyder@vmware.com>
> Subject: Re: i386 floating point test case integration with gdb.reverse test-suite
> To: "paawan oza" <paawan1982@yahoo.com>
> Cc: "Hui Zhu" <teawater@gmail.com>, "gdb-patches@sourceware.org" <gdb-patches@sourceware.org>
> Date: Monday, August 17, 2009, 3:08 AM
> paawan oza wrote:
> > Hi Michael,
> > 
> > please find the test case patch attached. Mainly the
> patch is intending to test floating point patch which I have
> submitted earlier.
> > 
> > tests are as follws.
> > 
> > 1) first test tests basic testing of FPU stack
> (st0-st7) and its restoring while reverse execution,
> > 2) second test is testing FPU environment specially
> fstatus and ftag register.
> > 
> > please find the patch attached. (I am not pasting
> patch in email text because of space and tab issue)
> > 
> > please review and let me know your comments.
> > 
> > Regards,
> > Oza.
> 
> These are very good!? Congratulations on getting the
> hang of dejagnu so quickly.
> 
> A few pretty minor suggestions...
> 
> 1) There are some tests that look like this:
> ???gdb_test "info register st2" "st2 *1.*"
> and some that look like this:
> ???gdb_test "info register st1" "st1
> *1\t.*"
> (with a tab in the regular expression).
> 
> I like the ones with the tab better -- its more exact, and
> it
> makes certain that the value is really "1", and not eg.
> "1.1".
> Could you make them all like that?? Just the
> single-digit values.
> I'm not worried about the multi-digit ones.
> 
> 2) The file names should reflect the Intel architecture,
> since
> eventually we will have process record for other
> architectures.
> In fact, maybe these tests should go in gdb.arch instead of
> in
> gdb.reverse, since all the other tests in gdb.reverse are
> architecture independent.
> 
> How about testsuite/gdb.arch/i387-stack-reverse.exp, etc?
> 
> 3) I'm not an Intel expert, but you had mentioned a few
> other registers in an earlier post, namely:
> ? fstat
> ? ftag
> ? fiseg
> ? fioff
> ? foseg
> ? fooff
> ? fop
> 
> Are all of those handled now?? If so, could we include
> them in the test?
> 
> Thanks again for your hard work and patience.
> Michael
> 




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