RE: Make cvs ignore some changes in files

[Frederic Brehm]

At 07:12 AM 8/22/2002 -0700, Noel Yap wrote:
> I don't understand, if Makefile.cache has to be in
> synch with the other files:
> 1. Why are there systems that are able to commit, but
> can't build Makefile.cache?

There are no such systems. I'm not sure where my text misled you into 
thinking that, but that's not what I meant to say.


> 2. Why would you not want to commit Makefile.cache
> once it's regenerated?

You cannot commit it until the other corresponding source files are 
committed. If you commit new or modified source files, then you must commit 
the modified Makefile.cache.

Yes, there have been a couple of times where everything wasn't properly 
committed and it broke the build for other developers. That's no worse than 
other typical errors like forgetting to cvs add a new file and committing 
other modified files that depend on the new file.


> 3. Why can't the build system just generate
> Makefile.cache without it being versioned?

Building Makefile.cache from scratch requires running a GUI tool. The GUI 
cannot be run without manual intervention. The build system is used for 
automated (read "no manual intervention") builds to create a release. It's 
the middle part of "cvs export; build; package".


Just for the record, I dislike this GUI tool and the extra conditions it 
puts on the build and the extra work I had to do to get the builds to be 
repeatable. This method was the best solution I could find.

Fred
_____________________________________________________________________
Fred Brehm, Sarnoff Corporation, address@hidden
http://www.sarnoff.com/digital_video_informatics/vision_technology/index.asp