Re: edit document in several branches simultaneously

[Derek R. Price]

I'm moving this discussion to address@hidden  If you reply to this 
email, please remove address@hidden from the address list.

Paul Gelderblom (ptok) wrote:

> I checked with Matthew Ogilvie, who earlier published a version of the
> advisory locks patch including documentation and test cases.
> Matthew writes:
>  
>
> > I just posted an updated version of the patch on sourceforge: 
> > http://sourceforge.net/tracker/index.php?func=detail&aid=822306&group_id=20011&atid=320011
> >    
> >
> > Or: The "patches" link under http://sourceforge.net/projects/cvsenhancements/
> >    
>
> Derek: Could you indeed review this patch and see wheter it can be included
> or needs more work?
> If needed and insofar as Matthew is not doing it, I'm volunteering to do the
> additional work needed to get this patch included in the official CVS source
> tree.
>  

Thanks Paul!  I asked Matthew to seperate the patches into more easily 
consumable chunks a long time ago and I didn't realize he had actually 
done so.  Thanks for the link!

Anyhow, here's what I've reviewed so far.  I'm going through the file 
roughly in order:

> #####################################################
> #####################################################
> #####################################################
> ## makeComboPatch-contributor: cvs-sanityCvsrcFix
> This fixes one case when the server was using the users .cvsrc file,
> sometimes causing artificially failed tests.
>  

I've committed something like this patch to both stable and feature.

> #####################################################
> #####################################################
> #####################################################
> ## makeComboPatch-contributor: cvs-sanityCwdFix
> This subpatch fixes a sanity.sh problem with Linux 2.2 with bash 2.05b.
> Without this fix, an earlier workaround ("cd .." from a removed
> dirctory) does not work in my environment.

Is this really a common problem?  Bash 2.05b is a development release 
(as in not stable) of Bash, and I'm reluctant to put support into the 
test suite for a bug that only shows up in pre-releases of Bash.

> #####################################################
> #####################################################
> #####################################################
> ## makeComboPatch-contributor: cvs-releaseFix
> This subpatch fixes a problem with "cvs release ..." where it will
> screw up ./CVS/Entries when you try to release a subdirectory (it
> was modifiing ./CVS/Entries when it thought it was modifiing
> subdir/CVS/Entries).
>  

I've committed something like this patch.  The original patch was way 
overkill.  I've committed a revised version of it to both stable and 
feature along with a cleanup patch for some issues I discovered in 
client.c while reviewing this issue.  Feel free to take a look at the 
commit logs for more information.

Reviewing this patch caused me to note that the way release() 
continuously calls various functions and then restores the initial 
directory repeatedly seems slightly disorganized.  Is there a good 
reason that the functions it calls are not responsible for restoring 
directories themselves?  If noone else has a feel for this I'll probably 
try and answer this question myself at some point in the future.

Derek