Re: [DevHelp] Some requests

[Biswapesh Chattopadhyay <biswapesh_chatterjee tcscal co in>]

On Mon, 2003-03-31 at 16:45, Mikael Hallendal wrote:
> mån 2003-03-31 klockan 12.10 skrev Biswapesh Chattopadhyay:
> 
> > O.K - here goes:
> >
> > 1. There never was a merge. There was an announcement that the projects
> > will merge, but nothing came off it in terms of actual code merge. We
> > just continued on our seperate ways. This was probably because people
> > from both sides realized that it was easier said than done :-(
> 
> Ok, do you think that this can be announced to the community? And be
> made clear to everyone wanting to start helping on creating an IDE for
> GNOME?

This is not really my call - it is up to the creators of anjuta (Naba)
and gide (Dave Camp ?) to clarify the issue publicly - I am just a
developer and a branch maintainer. But yes, probably an announcement of
sorts is necessary to avoid further confusion. Actually, I wanted to go
through the GIde codebase thoroughly and understand the issues from a
neutral POV before jumping to any conclusions, but that will take some
time (sorry Jeroen - not yet sgtarted) :-(

> 
> > 2. At the time of announcement, anjuta and gide had a mutually exclusive
> > set of developers - this continues to be the case till date, mostly due
> > to the huge difference in architecture and emphasis of the two projects.
> > Simply put, no one had the time to read and understand both the
> > codebases thoroughly and then port stuff over from anjuta1 to anjuta2.
> > Anyway, the vast differences in architecture meant that any such effort
> > would have ended up being a total rewrite anyway :-(
> 
> Yeah, that is what surprised me when the merge was announced. I figured
> that the Anjuta1 developers wanted a clean start but it seems that
> wasn't the case.

I wan't that actively involved in anjuta development at that time, so
I'm not sure what Naba and Dave might have thought. Maybe that was the
original thought but somehow it never happened, probably due to develop
intertia ?

> 
> > 3. There is no one who is able to devote more than a few hours a week to
> > either of these projects. IMHO, with this kind of scarcity of resources,
> > it is virtually impossible to port all features of anjuta1 over to
> > anjuta2 without a total stagnation of new development for a prolonged
> > period of time - and we're talking years here. If it were not the case,
> > it would have been done.
> 
> Very true.
> 
> > > We need a good IDE *BADLY* and by having this situation really doesn't
> > > help. If you can convince the anjuta2 hackers to drop that and join you
> > > guys in anjuta1 that would also be better than this situation.
> >
> > The main reasons people think that all development should switch to
> > anjuta2 seems to be that:
> >     a) It's got a '2' after it's name
> 
> Which needs to be fixed if there is never to be a merge. They shouldn't
> be named the same, it's the same problem as 'gtkhtml' and 'gtkhtml2'.
> 

Well, as far as I am aware, there is no clear-cut decision that a merge
will *never* take place, so maybe there is no hurry to immediate change
the name ?

> > Personally, I'd really like to see a merge - but, the other way round.
> > I'd like to cleanup anjuta's codebase to a more component-based
> > architecture by taking ideas and code from the anjuta2 codebase
> > *gradually*, without throwing away existing working code. For example,
> > we can take the build and wizards design, the GDL docking stuff, and
> > probably most of the plugin code from anjuta2 (these are the places
> > where the existing anjuta code sucks most) and move, one by one, the
> > anjuta components to use these features. This doesn't take that much of
> > dedicated effort and IMHO is the path of least resistance.
> 
> Yeah, without having looked at either code bases, if it's possible it's
> almost always better to never rewrite. However in some cases the work of
> cleaning up and replacing the parts needed will be a larger work than
> rewriting from scratch. This was the way I got the feeling it was in
> Anjuta1 (since it was announced that that code base was to be dropped).

It (the anjuta1 code base) not beyond salvage IMO. People are always
tempted to throw away code and rewrite it - it's always so much more fun
;-) Sadly, it often proves more work and the rewrite somehow never gets
finished 90% of the time, esp. in open source projects, because the new
developer doing the rewrite loses interest, or no longer has enough
time, or moves on to other things, or maybe realizes that it was a wrong
decision in the first place !

The anjuta1 design is actually quite powerful and lends itself to a lot
of configurabuility and flexibility through it's properties interface.
You can virtually design your own IDE by overriding most of the
build/vcs/debugger/etc. commands through tools.

> 
> If it's not I agree that it's probably better to work from that code
> base and clean it up and move towards a better architecture part by
> part. This is less fun for hackers though (since it's always more fun to
> innovate).

Absolutely - see above ;-)

> 
> Btw, is the Anjuta1 port for GNOME 2 buildable from CVS somewhere?

jhbuild anjuta should give you a pretty stable and working HEAD build
without too much trouble. To get an idea about what I mean about the
configurabiluity and flexibility, check out the tool editor and advanced
search/replace in 1.0.2 or HEAD. As far as I'm aware, HEAD is pretty
stable except for the messages window, which still uses CList and is
therefore probably quite slow. Also, some parts of the code uses ZVT and
others VTE, which is a bit problematic. Otherwise, things are pretty
good I think, but I've been too busy with 1.0.2 release and sourcebaseto
help Naba out with HEAD :-(

-- 
Biswa.