[p4] Merge into target from a prior changelist?

[Jeff Jensen]

We ran and re-ran integrates a number of times and a few different ways.
This took many hours, and we were under pressures to get the release
prepped.  In the end the brute force worked (since we wanted to do a "copy"
merge anyway), and we stopped looking into it to finish the real task at
hand.
 
I know the need to "integrate twice" to see any error messages on the second
one, but I can't remember specifically doing it this time.
 
Haven't contacted support, as I was thinking we just were "missing
something", such as extra rev info - sometimes large integrates get a little
picky with rev info when not just taking #head on either side of the view.
 
When I get a chance, I will do some tests by trying to reproduce.  If I get
the same consistent behavior, then I will email support of it.
 
So from your experience, we weren't missing any steps (other than possibly
doing a second merge to see if any files remained due to errors) or rev info
for that particular merge types?
 

  _____  

From: Stephen Vance [mailto:steve at vance.com] 
Sent: Sunday, June 04, 2006 7:14 AM
To: Jeff Jensen
Cc: perforce-user at perforce.com
Subject: Re: [p4] Merge into target from a prior changelist?


Have you verified that there is nothing else needing integration by running
the command again with -n?  If so, at this point it sounds buggy to me. Have
you contacted support on this?

Steve

Jeff Jensen wrote: 

Hi Steve, thanks.
 
No, we spot checked many files, and in some the contents were different,
some the same (as in merged correctly) and some the same (as in not changed
between the two codelines, added to pending changelist via -f).
 
Shouldn't need a -i, as the target was originally populated from the same
source codeline.
 
No -d issues for the spot-checked files.
 

  _____  

From: Stephen Vance [mailto:steve at vance.com] 
Sent: Friday, June 02, 2006 8:52 PM
To: Jeff Jensen
Cc: perforce-user at perforce.com
Subject: Re: [p4] Merge into target from a prior changelist?


At Jeff's request, this is going back to the whole list.

What were the nature of the diffs between the target and source at 11111? Was
it keyword expansion? Was it things that could need a -i or a -d?

Steve

Jeff Jensen wrote: 

The delete all and integrate -f gave us a new branch ;-).  We briefly
mentioned

that option, but team wanted to keep in same place - it's the staging area
for

releases to integration testing.  We needed to "go back" to a more stable

version, and not take so much newer stuff.



Yes, did the -f and resolve -at.  When we diffed the target to source at 11111,

there were still differences - this was the concerning part, and raises the

question "what don't we understand?!" :-).  Also, it kept all files in the

changelist (even those that didn't change).





Quoting Stephen Vance  <mailto:steve at vance.com> <steve at vance.com>:



  

At this point I would suggest creating a new branch.



But if that's not what you want to do, did you try



p4 integ -f //abc/main/... at 11111 //abc/release/integration/...

p4 resolve -at



Steve



Jeff Jensen wrote:

    

Hi,



We're confused by results at our attempts to merge to a prior changelist.

      

What

    

is the correct sequence to accomplish that?



All the Googling and Perforce manual and tech tips reading we found were

      

always

    

about *forward* movement of merging; we didn't find anything on our

      

situation.

    

Here is the scenario:



We integrated from //abc/main/... to //abc/release/integration/...

Obviously, this was a merge done at #head.  At the time, the changelist was

      

some

    

number, say 12345.



Now we need to make the target the same as a previous changelist, say

      

11111.

    

So we tried:

p4 integ //abc/main/... at 11111 //abc/release/integration/...



But that did not update all the files correctly in target.  What else is

      

needed?

    

My thinking is that Perforce saw that nearly all the changes from the

      

sources

    

were already integrated to the targets, so only picked a couple of things

      

(e.g.

    

adds of delete files).  Not sure about getting around that.



We tried -f, but that caused it to put all files again on the changelist,

      

which

    

didn't seem right as many files had not changed (total file count was about

8200; less than 1000 had changed).  Resolving and then "revert unchanged"

      

had

    

no effect, yet doing a diff on many files yielded "identical".



In the end, they deleted the target files and reintegrated.  Brute force,

      

but

    

obtained the correct results!  :-(



So we're obviously missing something here...

Please enlighten!  :-)

_______________________________________________

perforce-user mailing list  -  perforce-user at perforce.com

http://maillist.perforce.com/mailman/listinfo/perforce-user