extragear/sdk/kdevelop/app

Andreas Pakulat apaku at gmx.de
Sun Apr 18 19:07:52 UTC 2010


On 18.04.10 20:41:38, David Nolden wrote:
> 2010/4/18 Andreas Pakulat <apaku at gmx.de>:
> > On 18.04.10 11:29:31, David Nolden wrote:
> >> The lock within notify only "locks" within the main thread (that's the
> >> expression within the parens), and since the lock is recursive, this cannot
> >> lead to a deadlock.
> >
> > Just changed that to not use a global static (yet again). But my
> > question was not wether it can or can not, but wether you tested it? The
> > other thing is that the lock adds useless code which has 0 benefit over
> > the non-lock version unless you start adding code into the background
> > thread that uses it. And that must not happen for 4.0.0 (and IMHO also not
> > for 4.0.x)
> I tested it in the sense that KDevelop still works, yes. But now that
> we have it, we might start using it to fix specific bugs. I don't get
> why it should be more dangerous than any other way of fixing a bug, as
> any change is at first "untested". Then you test it, and if it works,
> it's fine.

Well, there's a difference on the impact the various fixes can have.
This one can potentially lock up (once used from the background)
kdevelop completely, making it unusable. Yes I understand there are way
to prevent them, but you'll have to make sure that this is done properly
at all places. If you miss one (for whatever reason) people will run
into problems.

> However, if we only want KDevelop to work with kdelibs 4.4, I guess we
> don't need the lock (yet), and can also revert it. Generally, given
> all the changes that seem to be going on in kate, I think we should
> completely disallow the startup of KDevelop 4.0 together with kdelibs
> 4.5.X, as I'm pretty sure it will be a sub-par experience.

No, that would mean not getting bugreports about the problems.

> We can tell the user to update KDevelop, as we may suppoert kdelibs
> 4.5.x in a later release. Else we'll get tons of bug-reports because
> of kdevelop + KDE 4.5 bugs.

And rightfully so.

Anyway, I'm tired of these discussions, I'll let you do whatever you
want and will do the same too. Which means focussing on other things I
have on my todo list.

Andreas

-- 
It may or may not be worthwhile, but it still has to be done.




More information about the KDevelop-devel mailing list