[p4] Who deleted a client
steve at vance.com
Thu Oct 4 14:02:23 PDT 2001
At 08:00 AM 10/4/2001 -0700, Jeff A. Bowles wrote:
>At 09:04 AM 10/4/2001 -0500, you wrote:
>>if you have client specs that are complex and change often, then the
>>recreate approach can pick up those changes. Not that you couldn't update
>>your fixed client before each build too...
>Oh, certainly. But the question remains in my mind,
>what sorts of procedures would require constantly
>changing the client specs and thereby changing the
>contents of the workspace?
[ ... ]
>I've heard some people say "but my build system insists
>that the source is on c:\hardcodedpath!" - which points
>to an approach that, to me, feels foreign. (I try to make
>$SRCHEAD, or something like that, point to the top of
>a tree that's being built; if I have to run sed/perl/python
>against a MSDev-generated makefile to defeat its
>hardcoded idea of where a tree should live, I will.)
It's a bit like driving nails with a sledgehammer, but W2K's drive mounting
could help here.
>I'm not saying that a procedure that updates the client
>spec as part of development or part of an overnight
>build is always dumb. I do believe, however, that such
>a procedure invites further questions and inspection.
I agree with Jeff here. If changes to the client spec introduce this much
risk into the procedures that have been developed, an alternative solution
might be to determine the source of the need to change the spec and solve
that problem instead.
mailto:steve at vance.com
More information about the perforce-user