[p4] P4EXP 2005.2

Thu Feb 9 15:07:16 PST 2006

This looks very much like a problem we've just noticed (we upgraded the 
server to 2005.2 last week). I've been working with Perforce support on 
this, but we haven't gotten to the bottom of it yet. Here's what we're 
seeing on a single file in a large integration: The only diffs are on 
theirs, but if you do an auto-accept-theirs, the client file will be 
incorrect (only taking in some of the changes on the source branch). If 
we then do a revert, and do the merge manually using p4merge, the 
version of the file listed as theirs is actually incorrect as well.

After reading the post below, I decided to try what happened on 
submitting the apparently corrupt file, and the server version was 
indeed correct, taking in all of the changes on the source branch. The 
client version on the machine that did the integrate, however, remains 
corrupt until you do a force sync. This is rather worrisome, and I'm 
hoping this gets resolved soon. In the meantime, I've recomended our 
developers refrain from integrating as much as possible.

If it is indeed a bug in the new server, it's pretty scary, so, 
personally, I'd advise against upgrading until more is known on this 
issue. Perforce support being as good as they are, I'm sure this will be 
cleared up pretty fast.

Frank.
---------------------------------------------------------------------
Frank Compagner                                       Guerrilla Games

Leo Zelevinsky wrote:
> Wow - that sounds bad! So auto-resolve with "safe automatic resolve"
> choice doesn't work right when only "theirs" is different, but only
> sometimes? Ouch. Maybe this is bad enough for us to hold off our
> update (which I hate to do because I had to fight to get them to agree
> to update it to begin with - we are currently on 2004.2). The reason
> why I really want 2005.2 is because of the "time-lapse view across
> integrations" feature.
> 
> I agree that it'd be nice to keep the p4win in the main installer -
> but that's not that big a deal for me.
> 
> Thanks a lot for relating this issue - has anyone else run into it?
> (I'm trying to understand how common a problem it is).
> 
> P.S. I am getting perforce-source-control bounces too... Did someone
> subscribe perforce-user to perforce-source-control or something?
> 
> On 2/9/06, Whitfield, Greg <gwhitfield at europe.ea.com> wrote:
> 
>>The big one is a merge/integration bug that causes corrupted files when
>>you auto merge some files where the only changes are on "theirs". The
>>merge corrupts the local client copy. If the user just submits, the
>>local copy stays corrupted but the server file is OK due to the lazy
>>copy mechanism. When first reported to me I just did not believe it
> 
> 
>>However - and this is the killer - if the user does notice, often by a
>>failed compilation, and then attempts to edit the file after the merge
>>and before submission, the version that gets submitted is corrupted.
>>
>>The odd thing is that not all files show this behaviour - just some. By
>>"corrupted" we get seemingly random symbol deletions and additions in
>>the files being merged.
>>
>>Note that Perforce support are actively looking at this, and have
>>managed to repeat it. I just don't yet have a fix date.
>>
>>They supplied a procedural work-around, but that does not work in the
>>case of individual change list integrations.
>>
>>Apart from that we have an odd verify error introduced by the server
>>update - also being looked at by support. There's a related memory
>>concern, but I think that has been reported in another thread, and maybe
>>a case of a change in behaviour rather than a bug.
>>
>>We also had a bug in P4Win (2005.2.90475) with reverted integrations,
>>but this was fixed by a new release on Jan 19th (2005.2.91908).
>>Although, interestingly, Perforce's download page did not appear to
>>report the later version - this may since have been updated.
>>
>>And, while not really a bug, or even 2005.2, I really wish P4Win was in
>>the main installer - this change has been an administrative niggle,
>>where users with old versions only install "Perforce.exe", and so end up
>>still using old P4Win's.
>>
>>God, this sounds like a giant moan!  It's really not meant to be -
>>Perforce support have been their usual helpful selves and the only
>>critical issue is the integration bug. I'm sure they'll fix it soon -
>>hopefully before my users lynch me!
>>
>>We're on Windows XP Server for P4s, and 95% of clients are WinXP based.
>>
>>BTW, my recommendation to not upgrade is purely a personal one borne out
>>of the above, and in no way reflects the views of my company.
>>
>>Greg
>>~~~~
>>
>>
>>-----Original Message-----
>>From: leo.zelevinsky at gmail.com [mailto:leo.zelevinsky at gmail.com] On
>>Behalf Of Leo Zelevinsky
>>Sent: 09 February 2006 16:05
>>To: Whitfield, Greg
>>Cc: Zarnes, Todd; perforce-user at perforce.com
>>Subject: Re: [p4] P4EXP 2005.2
>>
>>Would you mind listing some of these issues? We are considering
>>upgrading shortly and I'm disturbed by your statement.
>>
>>Also - are we talking about server or client performance?
>>
>>Thanks!
>>
>>On 2/9/06, Whitfield, Greg <gwhitfield at europe.ea.com> wrote:
>>
>>>I have had reports from a few of my users that things do appear slower
>>
>>>since the upgrade, to the extent that uninstalling that component has
>>>sometimes been necessary.
>>>
>>>We are experiencing a number of issues with 2005.2 that we did not
>>>have with 2005.1. I would not recommend the upgrade.
>>>
>>>Greg
>>>~~~~
>>>
>>>
>>>-----Original Message-----
>>>From: perforce-user-bounces at perforce.com
>>>[mailto:perforce-user-bounces at perforce.com] On Behalf Of Zarnes, Todd
>>>Sent: 09 February 2006 01:16
>>>To: perforce-user at perforce.com
>>>Subject: [p4] P4EXP 2005.2
>>>
>>>Anybody else seeing a slow down in Performance with the new 2005.2
>>>P4EXP?
>>>
>>>
>>>
>>>Todd  Zarnes
>>>
>>>Intuit Eclipse SCM
>>>
>>>303.938.8801 x1544
>>>
>>>
>>>
>>>_______________________________________________
>>>perforce-user mailing list  -  perforce-user at perforce.com
>>>http://maillist.perforce.com/mailman/listinfo/perforce-user
>>>
>>>_______________________________________________
>>>perforce-user mailing list  -  perforce-user at perforce.com
>>>http://maillist.perforce.com/mailman/listinfo/perforce-user
>>>
>>
> 
> _______________________________________________
> perforce-user mailing list  -  perforce-user at perforce.com
> http://maillist.perforce.com/mailman/listinfo/perforce-user
> 