[vlc-devel] Re: vlc: svn commit r17986 (md)

[Rémi Denis-Courmont]

Le vendredi 24 novembre 2006 00:31, Olivier Teulière a écrit :
> > So proposing a working and fast solution (fixing/upgrading the
> > compiler) is stupid, and forward porting a *non-working* patch
> > without even testing it is intelligent.
>
> I find it rather funny that you criticize a non-tested patch, when
> you broke the _compilation_ at least twice in the last 3 days. And
> I'm not even talking about testing...

Compare the number and size of commits.

> Now don't get me wrong, I think you are doing a very good job on VLC.
> But once again, everyone can make mistakes. Yes, even you. You may
> not realize it, but you are very arrogant on the mailing-list (and
> sometimes on IRC as well), and this can really hurt people.
               ^^^
At least I am not mocking people gratuitiously there as some third 
parties do there.

> For example, you can very well disagree with a commit, or simply want
> to 'challenge' it. But instead of reverting it immediately, you could
> consider asking for the reason of the commit first...

I did not revert that one. Last commit I reverted was a revert of my own 
commit, made without even trying to understand the logic. I am not one 
of those revert-any-commit-that-break-my-tree people (and yes, I am not 
the only one breaking someone's else tree).

> Or, to take the example of another recent thread, it's not because
> the Windows network stack implementation is completely buggy that we
> can accept regressions compared with previous versions. 

There is no regression that I can see. The last nightly branch works 
fine on XP. And it is fixed finally on Linux.

> VLC is
> portable, and saying "it works for me" is not enough.

That's not MY point. That's the point of people who revert (or ask to do 
so) any commit that breaks *their* tree and introduces temporary 
regression in "their" features. I have no corporate obligation wrt VLC, 
so I am not one of these.

> It must work
> still on as many platforms as possible, and one must be careful not
> to break stuff for the others as well. I can understand that you
> broke the Windows behaviour since you work on Linux, or that md broke
> the Linux one since he works on Windows. But now that you have
> identified the issues, why not cooperating to find a solution working
> for both systems?

Md did not break anything. The bug has ever been there.

> This may imply to find a dirty workaround for
> Windows, but nobody said that being a developer is always fun and
> easy... :)

It's not by telling people to revert any mildly disruptive commit that 
you can reach that kind of end. Unfortunately, I am not the one who 
needs to learn most on that _particular_ issue.

> As a side note, maybe we could decide that a commit only comes after
> full compilation + testing (if possible/relevant). I had the feeling
> that it was an implicit rule for a long time, but it may be useful to
> make it explicit, as a reminder.

I am afraid we cannot improve much. VLC is so slow to build that devs 
will always be tempted to commit without full build. And even then, 
breakage typically occurs on another tree/platform/etc :(

-- 
Rémi Denis-Courmont
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20061124/6e32ecf2/attachment.sig>