make install watcher

didbaba didbaba at free.fr
Wed Jan 10 08:15:27 PST 2001


Le Tue, Jan 09, 2001 at 08:56:43PM -0500 , Gerard Beekmans écrivait un message.
> On January  9, 2001 09:36 am, Bryan Dumm wrote:
> > Howdy,
> >
> > Ron885 brought this link up...
> >
> > http://mayams.net/~izto/checkinstall-en.html
> >
> > I read about the program before, but lost it my mess called
> > "archives". Anyways check it out, as it might be useful for
> > package management things.....
> 
> It required installwatch, another binary on the system. A simple script can 
> do the same and I hve grown quite fond of the following script based on the 
> find-ideas by Jesse Tie Ten Quee. What it does is this:
> 
> you run it, it will collect all files current on the system (excluding a few 
> directories like tarball directrory, proc, dev, mnt). It will then pause so 
> you can run your make install process. Then you hit a key and it will collect 
> all files currently on the system. It then compares the two outputs (using 
> diff) and writes the diff to a file. Then modify this file (to remove those > 
> symbols diff puts in there) a bit and done. This file can then be piped 
> through xargs to remove it (read a line from the output file, run 'rm' on it. 
> Read next line, run 'rm' on it, until EOF).

I think that there is a major problem with the find script : He do no see if there is a file that is rewritten. I know that some package erase some file, and put their file instead. Don't ask what package, I do not remember. And if you re install a package over an old one you see nothing.
( I can find the name of this package if you have a command that look for the same name in all my list-package-file)

To change that I use touch to create a file who indicate the time, and I use find -newer ... But this is not the correct solution, that there is some package that cp file with a wrong time (yes, it's true but I don't remember which one, maybe bzip2 ???).

Maybe, a solution is to mix the two scripts.

-- 
didbaba
powered by LinuxFromScratch 2.4





More information about the alfs-discuss mailing list