pcp
[Top] [All Lists]

Re: [pcp] URGENT potentially serious regression in 3.7.0

To: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Subject: Re: [pcp] URGENT potentially serious regression in 3.7.0
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon, 11 Mar 2013 05:07:41 -0400 (EDT)
Cc: pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <513D9A93.6080806@xxxxxxxxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Hi Ken,

----- Original Message -----
> On 10/03/13 07:21, Ken McDonell wrote:
> ...
> It is not a non-deterministic issue.  169 fails 20 out of 20
> attempts.

This is pretty bizarre then, but...

With current dev branch (and master now) for all of hosts I have
(RHEL, Debian, MacOSX) 169 passes both with/without NSS enabled.

> I have no clue why making the pmcd trace logs unbuffered leads to the
> correct behaviour ... some I/O, signal, select, ... snarfoo as result
> of recent changes in the whole socket/ipc layer?
> 
> I've attached both pmcd.logs ... ignore the DoDesc diags, I added
> these in desperation.
> 
> [...snip...]
> 
> OK, this one is a diagnostic only issue when PM_DBG_APPL0 is set, so
> it needs some rework but is not a release blocker.
> 

Release was tagged earlier today - the remaining issues here will have
to be resolved in a follow-up 3.7.1 (once I understand how to reproduce).

thanks!

--
Nathan

<Prev in Thread] Current Thread [Next in Thread>