Re: [harinath@cs.umn.edu: Re: Gnome-libs and libart_lgpl]

[Raja R Harinath <harinath cs umn edu>]

"Brandon S. Allbery" <allbery@ece.cmu.edu> writes:
> In message <d91zmuuh50.fsf@rawana.cs.umn.edu>, Raja R Harinath writes:
> +-----
> | > OK. That constitutes a good reason to use checkout all the time, since 
> | > one never really knows when such a thing will happen.
> | 
> | I hope not.  Each has its place.  `cvs update' is better than `cvs
> | checkout' for day to day updates since it sends deltas over the net in
> | the Remote CVS case.  `cvs checkout' appears to send out the whole
> | changed file.
> +--->8
> 
> I, and many others, do the update via cron, which means I'm not there
> to check out the modules file and figure out what needs to be "cvs
> checkout"ed.  So yes, "cvs checkout" is now what the cron script does.

IMHO that's overkill.  Virtual subdirs are added rarely enough.  I don't
think the set of virtual subdirs for `gnome-core', say, has changed in
the last six months (yep, the last change to that line was 09-Mar-1998,
that's 8 months ago).

Even `gnome-libs' went without change for a long time -- `libart_lgpl'
was added only recently (and there was a short time when `popt' was
added, then removed).  We are talking about timescales about once
every two months or so when you need to do a `cvs checkout' to get a
virtual subdir.

And, it is not as if you have to waste a humungous amount of time trying
to figure out you need to re-checkout.  

  1. Check the `SUBDIRS =' line in the Makefile.am (Makefile.in may be
     better) and verify that the set of subdirs listed there do exist.
     You can probably get away with looking only at the toplevel
     Makefile.

  2. If not, just to make sure, wait some time do a `cvs update -d', you
     may get that subdir.

  3. If not, cd .. && cvs co ...

  4. If that fails to get that subdir, something more is wrong

I think this can all be scripted (extracting a SUBDIRS= line out of the
Makefile.in should be a simple matter), and can even be verified before
compiling, so that you don't even have to wait till `make' complains.
Note that this doesn't need to look at the modules file for the
`checkout' decision.

Still, as long as you don't nuke the tree before checking it out, I'm
"happy". 

- Hari
-- 
Raja R Harinath ------------------------------ harinath@cs.umn.edu
"When all else fails, read the instructions."      -- Cahn's Axiom
"Our policy is, when in doubt, do the right thing."   -- Roy L Ash