Re: [PATCH] libreadline buildworld breakage.

On 2004-10-22 03:10, Andrey Chernov <ache_at_nagual.pp.ru> wrote:
> On Thu, Oct 21, 2004 at 12:01:15PM -0700, David O'Brien wrote:
> > What are you talking about??  src/contrib/libreadline/complete is off the
> > vendor branch.  After you import a new version of libreadline you have to
> > make a commit to HEAD to make the new version of
> > src/contrib/libreadline/complete.c show up.  Unless you commit *EXACTLY*
> > rev 1.1.1.9 to HEAD (which would be rev 1.10) without any change; rev
> > 1.10 should have a $FreeBSD$.
>
> 1) complete.c is not shown in conflics while importing (you can check
> commit message).
> 2) "cvs co contrib_libradline" extracts v1.1.1.9 and NOT v1.9
> 3) cvsup sends 1.1.1.9 diffs and NOT leave v1.9 inplace.

Step 2) seems weird.  It's probably because of the `branch: 1.1.1;' line
in the RCS header of the complete.c,v file.  Has this been added
manually to force complete.c back into the vendor branch?

> After my import I commit only files for which conflicts are shown (which
> is readline.h alone).
>
> Maybe you are right but something wrong happens inside, beyound of my
> knowledge.

When a file is already off the vendor branch and you do an import that
modifies it, a conflict might not show up but a cvs update -j VENDOR and
a subsequent commit might still be required to pull up changes from the
vendor branch into HEAD.  Some of those changes might not conflict with
earlier HEAD changes.  A local RCS id line like $FreeBSD$ will not
necessarily conflict with vendor changes, if they are more than a few
lines apart.  I think this is what David was referring to.

- Giorgos