Re: CVS problem

[Eric Siegerman]

On Thu, May 10, 2001 at 10:37:22AM +1200, Chris Cameron wrote:
> That was my first reaction as well.  However IMHO this theory is incorrect.
> 
> 1. I check out a file.
> 2. I build the .o file
> 3. I make changes to the file.
> 4. I realise I made a mistake in the changes, so I remove the file and do an
> update
> 5. make will now do an unnecessary rebuild of my .o file

While unnecessary, (5) produces correct results.  If there was
another step in your recipe:
  3b. "make"

and if (4) preserved the file's timestamp, then (5) would not
happen, and the results would be *in*correct.

> Only I can make decision as to whether the make needs to rebuild the .o file

True.  But at least CVS fails safe.  Your way, it wouldn't.

> and the best way for me to make the decision is to manually remove the .o
> file if it needs rebuilding!

Which you might well forget to do, and then spend a *long* time
trying to find an "I thought I fixed that!" bug.  Probably a lot
longer than you would have spent waiting for the superfluous
recompile.

> This rebuild could have knock on effects
> throughout the rest of the developers sandbox.

Alas, true.  Perhaps "update" could have a preserve-timestamp
option, to let you tell the program to do this usually-unsafe
thing on occasions when you're sure it's ok (something like
"make -t" in intent, though pretty much opposite in what it
actually does :-).  But making this the default behaviour would,
IMO, be an *extremely* bad idea.

--

|  | /\
|-_|/  >   Eric Siegerman, Toronto, Ont.        address@hidden
|  |  /
With sufficient thrust, pigs fly just fine. However, this is not
necessarily a good idea.
        - RFC 1925 (quoting an unnamed source)