[vlc-devel] commit: FIX: DSHOW "none" device open (#2201). Some cleaning (basos g )

[basos g]

2009/3/18 xxcv <xxcv07 at gmail.com>:
> xxcv wrote:
>>
>> Hello,
>>
>> I just noticed there's something wrong with my build currently,
>> when I test the NB today from buildbot it doesn't crash.
>>
>> This issue might be caused by qt 4.5.0 which I'm experimenting
>> with vlc-1.0.0-git.
>>
>> [0xc4c3384] dshow demux error: no video capture device was detected
>> [0xc4c3384] dshow demux error: can't open video device
>> ---
>> Crashed here...
>> ---
>> QObject::killTimers: timers cannot be stopped from another thread
>> ^^^ noticed this
>>
>> So, I'm in the process of switching back to qt 4.4.3 and retest.
>>
>> Regards.
>>
>>> basos g wrote:
>>>>
>>>> Sorry, you are not helping me. I would like to understand the state of
>>>> dshow plugin. For example is an audio device opened ? If not then
>>>> Common Close will be called else it will print just an error for video
>>>> and start capturing audio. And this might lead to clues about what is
>>>> deallocated ..
>>>>
>>>> Anyway i might found something about the destruction order (i don't
>>>> think there is race condition. CoInitialize is an atomic operation).
>>>>
>>>>  Could you please try with the attached patch ?
>>>>
>>
>>
> Once again my retests concludes my vlc-dshow issue isn't related to
> updating to Qt4.5.0.
>
> Anyways, FYI, I'm using gcc-3 to build vlc.
> However the race condition, heap free bug is still in the dshow.cpp code.
> (Introduced in the commit by your patch
> [576844b24d53c39bdc524af235095ed7939e91fd])
>
> I recommend someone to look at the code, and fix up some
> ancient mistakes, like missing parameter for an interaction
> function, which lived through the dark ages of invisibility been
> pass through developer's very eyes six times..., while been through some
> very useless commits like removing trailing spaces and tabs.
>
> One way to find it is, to use a Windows XP machine with VIVO and
> put vlc to run inside GDB and under qt4, Open Capture dialog, might
> trigger this bug.
>
> Conclusion:
> Below is attached debugging log file and result is from running
> NB buildbot's binary, on a Windows XP machine with VIVO.


I'm confused. You said that this is a vista issue. Now are you saying
that the crash exists also in XP ?
Please attach a full debug log (vlc -v2).
And what is a VIVO exactly? Should i have special hardware to be able to check ?

In my system (XPSP3) with a sweex usb video camera and a bt73 capture
device this bug is not observed.

What are do you talk about saying "like missing parameter for an interaction
function" ?