Re: [Qemu-devel] qemu-system-arm system support for big endian BE8

[Peter Crosthwaite]

On Thu, Sep 24, 2015 at 3:48 AM, sridhar kulkarni
<address@hidden> wrote:
> The issue is mostly related to my application under test. When the
> application calls a function the PC is getting set up to a wrong address,
> and then qemu crashes by displaying "floating point exception(core dumped)"
> message.
> I am able to move ahead by resolving the issue.

Was this a QEMU bug or an issue in your program?

But interestingly whenever
> my app crashes it always displays the same "floating point exception"
> message. But I don't see any floating operations at the point code crashes.
> I don't see any dump of the processor registers also. It's always just a one
> line message as I described above.
>

Ok, are you unable to share the binary or source? Alternatively, can
you strip it down to a super-minimal program that replicates just this
one issue? Pasting us a GDB backtrace of the failure might help as
well.

Regards,
Peter

> Regards
> Sridhar
>
>
>
> On Thursday, September 24, 2015 8:23 AM, Peter Crosthwaite
> <address@hidden> wrote:
>
>
> On Wed, Sep 23, 2015 at 8:41 AM, Peter Maydell <address@hidden>
> wrote:
>> On 23 September 2015 at 03:48, sridhar kulkarni <address@hidden>
>> wrote:
>>> Hi Peter,
>>>
>>> I was able to progress well using the BE8 work in the branch that you
>>> pointed out. I am experiencing floating point issue. The qemu just exits,
>>> by
>>> putting a message that "floating point exception(core dumped)". I suppose
>>> QEMU do support floating point operations. I heard about hard floating
>>> point
>>> and soft floating point support. Is there any configuration option in
>>> QEMU
>>> for floating point?
>>
>> QEMU's floating point support for ARM is good and known to work.
>> If QEMU exits with a coredump then that is either:
>>  * your test binary is dumping core due to a bug in your test
>>    (assuming you're using linux-user mode)
>>  * a bug in QEMU (unlikely but not impossible)
>>
>> If you can provide a reproducible test case we can have a look at it.
>>
>
> Yes, so the thing stopping me upstreaming this was a reasonable test.
> Can I have a look at your reproducer?
>
> Regards,
>
> Peter
>
>> thanks
>> -- PMM
>
>
>