Jump to content


p4/v99999 (client in logs): what is this?

p4 client version logs p4/v99999

  • Please log in to reply
9 replies to this topic

#1 Miles O'Neal

Miles O'Neal

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 11 October 2017 - 10:28 PM

MYSTERY SOLVED (read comments):
I ran across a handful of log entries with an odd client software version: p4/v99999.
These all appear to be "user-login" or "user-login -a" and they all came to the master via a broker (or possibly a forwarding replica).
Is this just an oddity in the pass-through of this particular command?
We have 150K to 200K commands logged per day from the broker and replica these came through, and almost all show something I'd expect to see, such as "P4V/NTX64/2017.1/1491634/v81".
We have eleven user-logins via those two hosts that show up this way, and two that show up normally.

#2 Sambwise

Sambwise

    Advanced Member

  • Members
  • PipPipPip
  • 325 posts

Posted 12 October 2017 - 03:49 AM

99999 is the "api" level set by the command line client (it sets it to an arbitrarily high value because the CLI doesn't care if it sees "newer" message strings).  I'd thought that the CLI set a version string that included the release number, but checking the source it just sets the "prog" to "p4":

https://swarm.worksh...ientmain.cc#513

so I expect what you're seeing is just normal old command line clients.

#3 Miles O'Neal

Miles O'Neal

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 12 October 2017 - 04:12 PM

It may be, but in that case why are there so few that show up this way?
Yesterday's log found 412,075 p4 commands logged that had a full string (e.g., "p4/2015.1/LINUX26X86_64/1252202") and only 11 commands logged that said "p4/v99999". Of those, about half came from the broker or a replica, and half did not. All of the v99999 commands came through  broker or replica. This holds true across the past month's worth of logs (383 p4/v99999 commands logged).
Correlation does not equal causation, but I find it doubly curious because of that. I'd exext at least a couple to have come in directly with those numbers.

#4 Sambwise

Sambwise

    Advanced Member

  • Members
  • PipPipPip
  • 325 posts

Posted 12 October 2017 - 05:50 PM

Do you ever see instances of the same user connecting with different p4 versions (e.g. one user that has both "p4/v99999" and "p4/2017.1/etc")?  And have you checked with any of those users to just see what version of p4 they have installed on their machines?

The simplest explanation is that you've got a small handful of users with a particular p4 build (maybe it's just the open-source build that has the v99999 version string and the in-house version has the official release string to mark it as an officially supported build -- someone from Perforce support could verify this), and those same users are all connecting via a broker.  I can't imagine that the server is doing any special mangling/translation of the version string that the broker would be getting in the way of.

#5 Miles O'Neal

Miles O'Neal

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 12 October 2017 - 06:54 PM

It's 33 users, most of whom always connect from the same system per user. I haven't contacted anyone yet; I was waiting to see if anyone here already knew the answers.  Many of the engineers here just run the commands they were given, and don't know all the places Helix commands are invoked. The tools team doesn't know of anything odd.

#6 Sambwise

Sambwise

    Advanced Member

  • Members
  • PipPipPip
  • 325 posts

Posted 12 October 2017 - 09:12 PM

Looking back at that SetProg() line in clientmain.cc, I'm feeling pretty confident that in the "official" client that's doing
SetProg( ident.GetIdent() );
and what those 33 users have got is an open-source build.  Can't see how else "p4" would end up turning into something that resembles the normal ident string; that line must get changed during the open source push.

#7 Miles O'Neal

Miles O'Neal

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 12 October 2017 - 09:24 PM

Could be.
I did a web search; the only things I can find are from 2005 and 2007- the latter being Dan Bloch's paper from Google on large Perforce installations. So these may be antique versions hiding in the bowels of equally antiquated lab systems. Whether those were built from open source by the users I have no idea.

#8 Miles O'Neal

Miles O'Neal

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 12 October 2017 - 09:25 PM

We've given fair warning, and will be blocking everything but 2017 versions soon. It will be a moot point then. 8^)

#9 P4Reg

P4Reg

    Advanced Member

  • Staff Moderators
  • 40 posts

Posted 13 October 2017 - 01:57 PM

v99999 would be the nightly snapshot build (ftp://ftp.perforce.com/perforce/snapshot/p4v/) that customers can download, they are bleeding edge and only really intended for testing and if there is critical bugfix or feature that has been coded but not yet rolled out till the next release

#10 Miles O'Neal

Miles O'Neal

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 13 October 2017 - 03:37 PM

Aha! Thank you. Although apparently now it shows up as 9999.9 . But close enough!





Also tagged with one or more of these keywords: p4, client, version, logs, p4/v99999

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users