[Xapian-devel] xapian csharp bindings crash

[Greg]

On Sun, Jun 26, 2011 at 8:28 AM, Olly Betts <olly at survex.com> wrote:
> On Sat, Jun 25, 2011 at 05:37:09PM +0200, Greg wrote:
>> FAULTING_IP:
>> _XapianSharp!CSharp_TermGenerator_SetFlags__SWIG_1+f2ca5
>> 000007fe`e3d8533e 8b440208        mov     eax,dword ptr [rdx+rax+8]
>>
>> EXCEPTION_RECORD:  ffffffffffffffff -- (.exr 0xffffffffffffffff)
>> ExceptionAddress: 000007fee3d8533e
>> (_XapianSharp!CSharp_TermGenerator_SetFlags__SWIG_1+0x00000000000f2ca5)
>>    ExceptionCode: c0000005 (Access violation)
>>   ExceptionFlags: 00000000
>> NumberParameters: 2
>>    Parameter[0]: 0000000000000000
>>    Parameter[1]: 000000001dcb0058
>> Attempt to read from address 000000001dcb0058
>
> What version of Xapian are you using here?

The version is 1.2.6, but it was crashing with 1.2.5 as well I just
didn't capture the dumps then.

> This sounds like Parameter[1] is the "this" pointer and Parameter[0] is
> the flags being set.  The other way round doesn't make sense, since the
> C++ code doesn't try to read the flags as an address, and also that's
> not a sensible flags value.
>
> The generated C++ code there looks fine to me, at least in the latest
> version.  So it looks like the pointer to the C++ object which the C#
> object is holding is invalid.  I'm not sure how that might have happened
> though.  Do you know if that's a plausible pointer value to a C++ heap
> allocated object in your environment?

I don't know what do you mean by that, how should I check if the
pointer is valid?

>> I still have the dump files I could upload if they'll help, I'll also
>> upload the latest compiled binaries to our production so that next
>> crash will have proper pdb-s. In any case it looks to be an issue with
>> pointer access in CSharp_TermGenerator_SetFlags__SWIG_1.
>
> The dump files wouldn't help me - perhaps somebody else would know
> what to do with them though.
>
> Cheers,
>    Olly
>
I'm not certain as to what seems to be the problem, I've printed out
the stack and the search that's being done isn't a special case trying
to call it again doesn't reproduce the problem, is there anything else
I could do to find the problem?

Thanks.