[p4] Is there a way to audit a workspace?

Richard Kistruck rhsk at ravenbrook.com
Mon Aug 17 09:25:22 PDT 2009


On 2009-08-17Mon, at 16:13, Roy Smith wrote:
> We have a very large workspace where we keep our entire build  
> toolchain (compilers, tools like perl and python, libraries, etc).   
> Our build team wants to run "p4 sync -f" every day because they  
> don't trust "p4 sync" (without the -f) to keep it up to date.  I'm  
> trying to convince them that this is silly.  It's certainly very  
> expensive, as it takes several hours to complete.  Worse, the act of  
> doing a "sync -f" touches the files on disk, so any build that's  
> currently running can be impacted as files change out from  
> underneath it.

...and "sync -f" is not sufficient anyway!  (See below).


> Is there some way to have p4 audit the files on disk and confirm  
> that what's there is what's supposed to be there, without actually  
> deleting and re-creating every file?

As far as I know, the five ways a file can differ from what it 'should  
be' are:

1. The local file is open for modification.
2. The local file is sync'd to the wrong revision.
3. The local file is not there -- it has been deleted.
4. The local file is there, unopened, but it has nonetheless been  
'edited' (corrupted).
5. The local file is extraneous, and should not be there.


The commands to check for these five, in sequence, are:

1.  p4 opened ...
2.  p4 sync -n ... at NNN
3.  p4 diff -t -sd ... at NNN
4.  p4 diff -t -se ... at NNN
5.  find . -type f -print | p4 -x - add -n | grep " - opened for add$"

(Where "@NNN" is the changelist number you want your local files to  
match).


If your local files do indeed match ... at NNN, then the results you will  
see are:
1.  "... - file(s) not opened on this client."
2.  "... - file(s) up-to-date."
3, 4, 5: no output.

This is the procedure I use.  I hope it is correct; corrections,  
improvements, or pointers to an extant TechNote very much welcomed :-)

Richard Kistruck
Ravenbrook Limited



More information about the perforce-user mailing list