[PATCH] arm-syscall record support [phase-3]

[oza Pawandeep]

I do remember the result, But I understood that the results are
failing mainly because of thumb2 and syscall missing support.
Chandra K was working on improving the test suite (but no hear from him).
The confusing part for me is;

1) How Do I run gdb.reverse ?
(I think I need to run them in gdb server mode, where host is x86 and
target is ARM, how did you run it ?)
[and along with that all the test code has to be cross compiled)

2)
Yao > "reverse/record doesn't consider software-single-step, because
i368 and amd64 doesn't need it.  I pointed out the similar problem
when reviewing your phase-2 patch, but it was not addressed."

Oza : I do not get this much; Can you please elaborate on it ?

3) If syscall and thumb2 test cases are falling in unexpected failure
then until we support them it might always be unexpected failures;
what do we do about that ?

4) I shall definitely work on "Process record does not support
instruction at  0xXXXXX" which must be resolved and I would like to
see those working.

5) What I would do is; instead of going for syscall patch; I shall fix
as many things as possible. then I shall integrate syscall record
support.

Thanks Yao; Waiting for your reply.

Regards,
Oza.
On Tue, Jun 19, 2012 at 4:09 PM,  <yao@codesourcery.com> wrote:
> On 06/08/2012 08:42 PM, oza Pawandeep wrote:
>> This patch provides arm-syscall record support. currently it supports
>> till gdb_sys_sched_getaffinity = 242.
>>
>
> Did you run the testsuite? and any improvements on tests in gdb.reverse?
>
> I run gdb.reverse test cases in arm mode (-marm) yesterday (without your
> patch), and got results below,
>
>                === gdb Summary ===
>
> # of expected passes            1137
> # of unexpected failures        1135
> # of expected failures          142
>
> It is worse than what I posted on Oct. last year
>
>  http://sourceware.org/ml/gdb-patches/2011-10/msg00435.html
>
> It is not surprise because we don't turn gdb.reverse on for ARM in
> default.  Although some fails are caused missing thumb-2 support, it
> looks like there are some bugs in basic functionality instead of missing
> features like signal recording.  For example,
>
>  1. gdb/infrun.c:1859: internal-error: resume: Assertion
> `!(singlestep_breakpoints_inserted_p && step)' failed.  Existing code on
> reverse/record doesn't consider software-single-step, because i368 and
> amd64 doesn't need it.  I pointed out the similar problem when reviewing
> your phase-2 patch, but it was not addressed.
>
>  http://www.cygwin.com/ml/gdb-patches/2011-10/msg00451.html
>
>  2. "Process record does not support instruction xe6bf3073 at address
> 0x840c" is what I saw in gdb.log.  Instruction on 0x840c is a correct
> ARM insn.
>
> (gdb) disassemble /r 0x840c,+10
> Dump of assembler code from 0x840c to 0x8416:
>   0x0000840c <main+12>:        01 30 40 e3     movt    r3, #1
>   0x00008410 <main+16>:        00 20 a0 e3     mov     r2, #0
>   0x00008414 <main+20>:        00 20 83 e5     str     r2, [r3]
>
> I didn't examine gdb.log carefully, so there may be other issues.
>
> Before we move on to the new feature, better to get existing bugs fixed,
> and fails in testsuite fixed as much as you can.  50% fail rate is not a
> good signal.
>
> --
> Yao (齐尧)
>