Re: Undesired Sticky Tags With cvs update

[Eric Siegerman]

On Mon, Feb 25, 2002 at 06:54:17PM +0000, Chuck Taylor wrote:
> I discovered last week that if I run
> 
>     cvs update -f -r <branch_tag>
> 
> in a working directory, then files that don't have <branch_tag>
> defined are checked out from the main trunk, as the -f option directs.
> However, CVS still applies the <branch_tag> sticky tag to these files,
> so subsequent cvs status commands result in errors on files that don't
> have the tag:
> 
>     cvs status: <file> is no longer in the repository
>     ...
>     File:  <file>    Status: Entry Invalid

IMO that's a bug.  CVS shouldn't be able to put its own data
structures into a state that it considers invalid.

That said, why are you only tagging some of the files?  Just tag
everything; then it won't be a problem.

This was discussed in another thread only a week or two ago, but
what it comes down to is: in CVS, a tag marks the state of the
*entire project* at a given point in time.  By tagging only some
files (I presume, those you either have changed or intend to
change), you're explicitly telling CVS, "at this moment, the
project consists of only these few files" -- which is very wrong,
of course.  When you update to your tag, CVS gives you only the
tagged files -- which isn't what you want, but is just what you
told it to do.  And using -f to override that, though of course
it shouldn't put CVS's data structures into an invalid state,
also shouldn't be expected to make up for the initial "garbage
in".

> For what it's worth, I've seen this behavior in the SGI-supplied port
> of CVS 1.10 and in CVS 1.10.7 on Linux (Red Hat 6.2).

Those are *ancient*.  They should be upgraded.  (Not that doing
so will change the behaviour under discussion, but it'll fix many
other problems.)

--

|  | /\
|-_|/  >   Eric Siegerman, Toronto, Ont.        address@hidden
|  |  /
One ring to rule the mall.
        - Movie review headline, eye Magazine