Re: PATCH: other/20731: contrib/gcc_update hard code -rgcc-3_4-branch

[Richard Earnshaw <rearnsha at gcc dot gnu dot org>]

On Mon, 2005-04-04 at 17:40, H. J. Lu wrote:
> On Mon, Apr 04, 2005 at 12:38:39PM +0100, Richard Earnshaw wrote:
> > On Sat, 2005-04-02 at 17:44, H. J. Lu wrote:
> > > This patch will make sure cvs update picks up the branch from CVS/Tag.
> > > I tested it on mainline, gcc 3.4 and gcc 3.4 rhl.
> > > 
> > 
> > I think this is a bad idea.  If I explicitly check out a single file
> > with a specific version, then this patch will cause gcc_update to
> > silently undo that and reset the file to the head of the branch.
> > 
> > It would probably be ok if this were wrapped by some sort of flag (such
> > as --force-latest), but I don't think it should be the default.
> > 
> 
> Are you saying this patch
> 
> http://gcc.gnu.org/ml/gcc-patches/2005-01/msg00751.html
> 
> should be backed out?

If the sequence of commands

gcc_update -r branch_tag
cvs -q update -r 1.1 foo.c
gcc_update

does not leave foo.c at revision 1.1 then yes, that patch is broken too
and should be backed out IMO.

Setting revisions is supposed to be sticky.  If the patch breaks CVS
stickiness then the patch is broken.

R.