Are RCS ID's bad?

Are RCS ID's bad?

[Andrew Cagney]

Hello,

I'm considering deleting any line containing something like:

	$Id: .... $

($Date: ..$; ....) from the GDB source tree.  They make a right mess of 
merges, diffs, compares and the like.

Can anyone come up with a reason to retain these?

GDB has gdb/version.in as a revision identifier.  Going by recent 
e-mail's this is proving very effective - people are identifying GDB 
snapshots and checkouts by date (although sometimes the quoted dates are 
backwards :-)

	Andrew

Re: Are RCS ID's bad?

[Kevin Buettner]

On Aug 9,  2:59pm, Andrew Cagney wrote:

> I'm considering deleting any line containing something like:
> 
> 	$Id: .... $
> 
> ($Date: ..$; ....) from the GDB source tree.  They make a right mess of 
> merges, diffs, compares and the like.
> 
> Can anyone come up with a reason to retain these?

FWIW, I'm all in favor of getting rid of them.  I once did a merge (in
a different tree) between the trunk and a branch which had a lot of
conflicts.  Over ninety percent of them were due to RCS identifiers.

Kevin

RE: Are RCS ID's bad?

[Mcspadden, William C]

I find the RCS keywords useful (except for $Log$) and would not
recommend deleting them. 

For purposes of merges and diffs, keyword expansion can be suppressed
by cvs when you do a checkout. See chapter 12 of the CVS manual for
a discussion of this. 

This eliminated the many problems I had when doing merges caused by
the very thing you describe. However, it didn't eliminate the problems
with the $Log$ expansion. Just don't use $Log$.

Bill Mc.

> -----Original Message-----
> From: Andrew Cagney [ mailto:ac131313@cygnus.com ]
> Sent: Thursday, August 09, 2001 11:59 AM
> To: gdb@sources.redhat.com
> Subject: Are RCS ID's bad?
> 
> 
> Hello,
> 
> I'm considering deleting any line containing something like:
> 
> 	$Id: .... $
> 
> ($Date: ..$; ....) from the GDB source tree.  They make a 
> right mess of 
> merges, diffs, compares and the like.
> 
> Can anyone come up with a reason to retain these?
> 
> GDB has gdb/version.in as a revision identifier.  Going by recent 
> e-mail's this is proving very effective - people are identifying GDB 
> snapshots and checkouts by date (although sometimes the 
> quoted dates are 
> backwards :-)
> 
> 	Andrew
> 
> 

Re: Are RCS ID's bad?

[Andrew Cagney]

> I find the RCS keywords useful (except for $Log$) and would not
> recommend deleting them. 


In the case of GDB, any useful information those RCS keywords carried 
has long since been lost.  Remember the sources have been in and out of 
random CVS repositories.

A far more reliable mechanism for tracking change is the ChangeLog. 
However, if the below can be made to work ...


> For purposes of merges and diffs, keyword expansion can be suppressed
> by cvs when you do a checkout. See chapter 12 of the CVS manual for
> a discussion of this. 


Would you know how -kk and -kb interact?  A number of files are created 
with -kb.


> This eliminated the many problems I had when doing merges caused by
> the very thing you describe. However, it didn't eliminate the problems
> with the $Log$ expansion. Just don't use $Log$.


Thanks for the pointers.

	Andrew

Re: Are RCS ID's bad?

[Richard Earnshaw]

The NetBSD folks use $NetBSD: .. $ instead of $Id: $.  The great advantage 
of this is that if the file is imported into another repository, or if a 
foreign file is imported into NetBSD's repository, then the master Id 
numbers are preserved.

If you are going to keep using such tags, I'd strongly encourage you to 
consider a similar approach ... $GNU:...$?

R.