Re: CVS doesn't like its own checkouts???

[Jonathan Sergent <sergent+gnome kgb etla net>]

In message <367C5EE3.C2AE4B73@home.com>, Shane Michael Lenagh writes:
 ] This line generates a perfect (assuming your dep script is perfect :-) list 
 ]  of
 ] modules in proper COMPILE-TIME dependency order.

I have worked on it some more.
I shall put it up on the web rather than sending continual email about 
it.  I have it somewhat integrated with my build "script" (it's actually 
a makefile).

 ] Where gnome-deps is just your makefile script, modified slightly so that 
 ] each target just echoes its name, so each target entry in the "makefile" 
 ] looks like:
 ] 
 ] target:  dep1 dep2 dep3 ...
 ]         echo target
 ] 
 ] This seems to work great for generating a non-redundant, dependency-correct 
 ] list of modules for both checkout and compilation.  Furthermore, it is 
 ] paremetrized so that you can quite easily replace WORLD with any subset 
 ] or single target module that you like.  My question is:

The tsort based script also works fairly well rather than trying the
file with make.  I never tried doing that.  It's nice that it seems
to work.

 ] 1.) Is this dependency tree (attached) still accurate?

No, it had some omissions.

Please see 

	http://www.io.com/~sergent/gnome-make/ 

for what I'm using.

I'm able to do a straight-through "type make and let 'er rip" and
it builds (and runs for the most part).  There are a few modules
that I have had to do local diffs to (gtkicq and gwp as implied in
other threads; also grpm and yagirc have some vestiges of old 
gnome-libs).

 ] 
 ] 2.) Any idea why I might be getting caught in an infinite loop when trying t
 ]  o do
 ] a:
 ] 
 ] cvs -z3 -d $CVSROOT co glib

I don't know anything about that (your infinite loop).  Perhaps the
files in glib/CVS are confused.  I can think of any number of other


--jss (evidently a release engineer at heart).