[vlc-devel] [PATCH] mmdevice: don't hard crash when CoInitializeEx fails on Start, show the reason

[Steve Lhomme]

On Wed, Mar 18, 2015 at 11:56 AM, Rémi Denis-Courmont <remi at remlab.net> wrote:
> Le 2015-03-18 13:26, Steve Lhomme a écrit :
>>
>> On Wed, Mar 18, 2015 at 11:24 AM, Rémi Denis-Courmont
>> <remi at remlab.net> wrote:
>>>
>>> Le 2015-03-18 13:13, Steve Lhomme a écrit :
>>>>
>>>>
>>>> Maybe you can tell us why ?
>>>
>>>
>>>
>>> It's obvious. We don't return errors on bugs, we abort.
>>
>>
>> And it is wrong.
>
>
> Intrinsically, returning an error on UB makes no sense. I don't get to
> decide on that, nor do you.
>
> We could just ignore the error. But abort() makes it easier to debug.

In my debugger (gdb in QtCreator on Windows) it did not catch the
crash at all. Also when a user reports a bug, it's much easier to use
a log than a stack trace. And as a more general remark, crashing is
the worst user experience, it should be avoid whenever possible.

>>> Plus your attitude is not really conductive of motivating detailed
>>> reviews.
>>
>>
>> Irrelevant.
>
>
> Oh it is very relevant. It shows once again how you have zero considerations
> for volunteer(s) and the time they waste reviewing your patches. This is
> unacceptable especially from a paid developper. Maybe you would be more
> considerate if I charged 80€ the hour payable in advance for code review
> "consulting services"?


If you have personal issues with me, you can talk to me on IRC, you
can start a thread in the mailing list. Hijacking a patch proposal,
providing fragments of answers and then complaining because I'm not
happy enough with what you provide is not the right way. It would be
better if you would restrict to technical masters in technical posts.

I do appreciate your contributions though. You know the code far
better than me, so for now I'm relying on you for things that are
outside of my main expertise.

Steve