KDE/kdevelop/lib/plugins/vcs/interfaces

[Matt Rogers]

On Thursday 31 May 2007 16:26, Andreas Pakulat wrote:
> On 31.05.07 15:20:42, Matt Rogers wrote:
> > On Thursday 31 May 2007 14:58, Andreas Pakulat wrote:
> > > On 31.05.07 14:32:16, Matt Rogers wrote:
> > > > On Tuesday 29 May 2007 11:32, Matthew Woehlke wrote:
> > > > > (I tried to reply to your commit log but don't see the message
> > > > > yet... might be moderated or blocked because I sent from a
> > > > > different address. So apologies if it shows up later and duplicates
> > > > > this :-).)
> > > > >
> > > > > Andreas Pakulat wrote:
> > > > > > On 27.05.07 21:04:33, Matt Rogers wrote:
> > > > > >> On Sunday 27 May 2007 20:53, Andreas Pakulat wrote:
> > > > > >>> SVN commit 668891 by apaku:
> > > > > >>>
> > > > > >>> add exec() which runs the job in a synchronous way
> > > > > >>
> > > > > >> Why? Under what circumstances would we need the synchronous way
> > > > > >> of doing things?
> > > > > >
> > > > > > For simple scripts for example, those don't have signal/slot
> > > > > > stuff set up necessarily.
> > > > >
> > > > > Right. Btw, whatever happened to wait()? Should we have exec(),
> > > > > wait() or both? (Note: exec() == start()+wait() if we have wait().)
> > > > >
> > > > > To answer Matt: scripts generally run sequentially, it is I think
> > > > > much easier to write a script that way than to all the time tie the
> > > > > finished() signal to starting the next step of the script.
> > > >
> > > > How bout for right now we don't care about scripts and instead focus
> > > > on actually making a usable IDE? If somebody is using the IDE, why
> > > > would they write scripts?
> > >
> > > Because they want to automate tasks, also look at the recent wishlist
> > > report about "open foo.cpp" and have it show up in KDevelop.
> >
> > My objection is that we have no one using this functionality already and
> > it overlaps with already existing functionality (aka command line
> > clients). It seems like we're just adding it because we think it would be
> > cool. If people want to write scripts to interact with version control
> > systems, they should use the command line version of said version control
> > system. The only version control system that doesn't have an command line
> > version is Visual SourceSafe and I don't want to support that from within
> > KDevelop anyways.
>
> So we should just drop what we have and revert to the interface we have
> implemented already?
>

No. Perhaps my problem is that I don't understand why we need the distinction 
between showLog and just log (and don't say it's for scripts) when, IMHO, 
we , as users of the interface, should just call a function, and have it do 
the right thing (either show a gui or not based on the information available) 
and give us the information we want, but perhaps this is too difficult or 
magical. 

> Also quite some of the functionality may be used by other plugins, think
> about getting a diff to head and sending that through teamwork to
> somebody who doesn't have access to the vcs (for whatever reason). Its
> not simply about scripts, its about exposing part of the functionality
> of the VCS in a way that is usable by other code and not just displayed
> to the user.
>
> Andreas, who will continue to work on new interfaces and help the VCS
> authors to implement them

Yes, we need to be able to implement functionality like this. If this is the 
one of the reason for the difference between showLog and log, then that's 
fine and I withdraw my objections.
-- 
Matt