[Indic-computing-devel] Re: debian toolchain question.

[<alo...@so...>]

Hi Cherry, 

This is interesting, particularly since I haven't been able to do the 
handbook build successfully on RH8 since the past month. 


> My suggestion is that we should move this virtual
> package ( indic-doc-toolchain.deb ) to the debian

Would this indic-doc-toolchain.deb file be such that deb2rpm works on it? 

> 
> II) _______The_src_makefiles_______ 
> 
> The Makefiles within doc/share/mk assume a FreeBSD
> directory layout. This is quite different from the
> debian Filesystem Heirarchy Standard [2] 
> 
> There are two solutions: 
> 
> 1) Provide an alternate cvs branch forked off the
> current main branch, and periodically merge from the
> main branch.
> 2) Provide OS specific hooks within the current branch
> and make the current build system portable. 
> 

> trying to make it portable would bloat the Makefile
> code a lot (writing wrapper variables, #ifdefs etc. ) 
> 

What is the cost of having a huge makefile? Does it
o increase build time?
o reduce readability ? 

> On the other hand, if we have a maintainer <->
> upstream relationship with the indic-port, where the
> FreeBSD branch, being the branch where all the action
> happens is treated as the upstream code, and the
> individual OS port is treated as maintainer ports,
> each OS port will have responsible owners and be
> tweaked and debugged to its best. 
> 

The way this currently happens is - Koshy makes the changes, as and when 
required, for everything to work on FreeBSD. And the rest(if any) follow, 
making the corresponding changes for their installations. I'm sure(or at 
least hoping) that wouldn't be the case in the long run. People working on 
various installations would be making changes to various parts.
Now, imagine a situation where developers are working on various sections of 
the handbook, on different installations. They may make some changes to the 
makefiles. Under the current system, other users would then be *forced* to 
make appropriate changes to make it work on their installations the next 
time they do a build. 

If we have separate makefiles for different installations, the changes would 
have to be done manually. Is there a way to automate them? Actually, yes, by 
the cvs commit logs one can make out what changes would be relevant for the 
other branches. So this could be workable. The questions that arise are, 

o how many branches do we keep? One for debian, another for RH, one more for 
mandrake, not to forget win*/cygwin
o how do the changes in one branch(not necessarily the bsd one) get merged 
to the others? 

I'm new to cvs so many questions may have obvious answers. 

Alok
-- 
                       alok_kumar ऍट softhome डॉट net
                     http://devanaagarii.net/hi/alok/blog
                   Can't see Hindi? http://devanaagarii.net
             http://groups.yahoo.com/group/linux-bangalore-hindi/
      Discuss devanagari at http://groups.yahoo.com/group/devanaagarii/