Re: Support for VCS operations

[Yavor Doganov]

В чт, 2007-12-27 в 23:16 +0200, Yavor Doganov написа:
> On Thu, Dec 27, 2007 at 11:11:33PM +0200, Kaloian Doganov wrote:
> > By issuing `make' twice we do not lose information about object 
> > files. But we can effectively lose information about which files are
> > already under version control (if we do not use `cvs status' on 
> > given file and just assume that if file is in the working copy in 
> > the beginning of the current make run, then it should be under 
> > version control).
> 
> I have addressed this quite valid concern in the implementation 

Apparently the ideas I had in my mind did not work fully precisely
because of the reasons you quote.  When a a target is up-to-date, make
doesn't run any commands so it is really hard to adjust the behaviour as
I'd want.  To accomplish this, I think I'll implement your suggestion
partially.

It will still be useful.  `make' by default will not do any VCS
operation, `make VCS=yes' will always do the right thing provided it is
run in the same working copy in "daemon" mode (i.e. that will be the
option for the fencepost script), and `make VCS=always' will be slower,
but still OK in some unusual situations.

P.S.  I regret that we started the project in Subversion and now we have
to make various workarounds just to make it build.  Due to the internal
implementation differences between CVS and Subversion we cannot be
absolutely sure now that GNUN is the GNUN we think it is.