Re: How to modify algorithm that determines weather a target needs rebuild

[Bob Proulx]

Paul Smith wrote:
> Bob Proulx wrote:
> > This sounds like the X part of the X-Y problem.  Why will the
> > timestamp of the file be changing back and forth?
> > 
> > Is this because you have two competing builds running in the same
> > directory with one doing it one way and the other doing it another
> > way?
> 
> Another likely candidate is source code control tools.  Some tools
> (ClearCase is the obvious one) will do this when you change what version
> of code you are looking at in your view.

And there are other possibilities too.  But I still wonder if it isn't
better to address the issue in the X part of the problem rather than
the Y part.  It would at least be useful to know the X part of the
problem.

> That's one primary reason why ClearCase comes with its own build tool
> (clearmake), precisely to manage this situation.  Clearmake has a GNU
> make emulation mode, although last time I checked (3-4 years ago) it was
> missing significant GNU make features still.

You might also remember Glenn Fowler's AT&T nmake program too.  (Just
to head off comments from the mailing list, the AT&T 'nmake' has
nothing to do with the MS 'nmake' and predates it.  It is a name
collision.)  It has been a while since I have used it but as I recall
it used exact timestamps along with a state file (base.ms) to hold the
timestamp state.  Of course nmake uses a different makefile syntax.
But it does keep timestamps in a state file.

  http://www2.research.att.com/~gsf/nmake/nmake.html
  http://www2.research.att.com/~gsf/man/man1/nmake.html

Bob