8 messages in com.perforce.revml[revml] Re: [p4] Why does p4 look at ...| From | Sent On | Attachments |
|---|---|---|
| barries | 16 May 2001 07:14 | |
| Stephen Vance | 16 May 2001 08:38 | |
| barries | 16 May 2001 08:57 | |
| Stephen Vance | 16 May 2001 09:00 | |
| Schaible, Jorg | 16 May 2001 09:03 | |
| barries | 16 May 2001 09:48 | |
| Maurice Meyer | 16 May 2001 09:49 | |
| barries | 16 May 2001 10:12 |
| Subject: | [revml] Re: [p4] Why does p4 look at PWD? |
|---|---|
| From: | barries (barr...@slaysys.com) |
| Date: | 05/16/2001 10:12:03 AM |
| List: | com.perforce.revml |
On Wed, May 16, 2001 at 09:50:10AM -0700, Maurice Meyer wrote:
I've got a call into Perforce about this since Solaris 8 cd seems to just use whatever you pass as the argument to set PWD.
Yeah, physical vs. logical paths is shell (and thus vendor) dependant. I remember wrestling with it on SCO and the -P/-L options to, IIRC, the cd and pwd shell builtins. It's probably a misfeature to take the word of PWD at all if it doesn't point to the same physical path as getcwd() returns, and as Steve Vance pointed out, best is probably to look at both and see if either can be made to work (and probably issue an error if you get to two different physical directories, each of which matches a portion of the client view).
I've found it necessary to have a wrapper that unsets PWD before running p4.
That's currently what I've done to see if it fixes Mathew Attaway's problem with vcp's test quite, I'll probably turn enable Perl's chdir to maintain it for me (see Cwd's docs in case anyone's curious).
On solaris 8, I'd consider this a mis-feature but only since it trusts something it should be able to which I consider broken.
I've started to think it's a buggy feature. If it did the arbitration Steve proposes, it would be a nice feature, working around the whole physical vs. logical path dilemma in most cases.
- Barrie