Dirty files

[Andreas Pakulat]

On 16.05.10 21:05:06, Milian Wolff wrote:
> On Sunday 16 May 2010 16:27:08 Andreas Pakulat wrote:
> > On 16.05.10 15:46:01, Esben Mose Hansen wrote:
> > > On Sunday 16 May 2010 13:45:03 Milian Wolff wrote:
> > > > The last step could be automatically done, David once did that for
> > > > project- internal files. Lets see whether we can do it for generated
> > > > files as well.
> > > 
> > > It should be easily enough we KDirWatcher, but I am not enough at home
> > > with KDevelop to be able to see which part should be responsible for
> > > this. What part has the dependency information? And what part can
> > > schedule files for reparsing?
> > 
> > You cannot use KDirWatch to watch all files or includes of a project, at
> > least you cannot expect that to work on all systems. The reason is that
> > the underlying notification systems usually have limits as to how many
> > files can be watched at the same time.
> 
> True, but this is not a reason for not using it. Of course it's not needed 
> here as it's "working" already (just one step missing, see below). But 
> KDirWatch is quite nice to use and if the file system supports it, it's imo the 
> best solution.

Sure, but one needs to know the limits too. Unfortunately KDirWatch is
not well documented wrt. its limits.

> > And seeing that KDirWatch is buggy and not threadsafe (see the various
> > bugreports we have for cmake) I don't think its a useful solution at
> > all.
> 
> Sorry, apaku - but I call FUD: The problems we had with it where due to the 
> inherent instability of our ProjectModel + Items. The generic manager uses it 
> and is quite stable.

Unless my memory serves me wrong you'll find evidence in the kdirwatch
svn log. I've seen quite some fixes over the last 6 or so months to it.

> > Apart from that I didn't have that problem yet at all here, when I
> > changed something in the .ui file the generated header was reparsed at
> > the latest when I started to type into the .cpp that used the header.
> 
> But the .cpp didn't get updated automatically. I.e. the change did not bubble 
> into opened documents. This _is_ done when you e.g. use an assistant to add 
> something to the header. So it should be possible to do here as well.

Sure, the .cpp got updated as soon as I started typing.  With the
habit of using a ui-member instead of subclassing it simply works.

Andreas

-- 
You will win success in whatever calling you adopt.