pcp
[Top] [All Lists]

Re: pmRegisterDerived return values

To: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Subject: Re: pmRegisterDerived return values
From: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Date: Wed, 20 Jan 2016 21:27:54 -0500
Cc: "'Nathan Scott'" <nathans@xxxxxxxxxx>, "'Marko Myllynen'" <myllynen@xxxxxxxxxx>, pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <00e201d153f1$f78cab30$e6a60190$@internode.on.net>
References: <569CB025.4070603@xxxxxxxxxx> <569CB7C5.7030803@xxxxxxxxxxxxxxxx> <1305297915.11201184.1453261568406.JavaMail.zimbra@xxxxxxxxxx> <y0mio2nzofj.fsf@xxxxxxxx> <00e201d153f1$f78cab30$e6a60190$@internode.on.net>
User-agent: Mutt/1.4.2.2i
Hi -

> > Could the error message go into the __pmContext structure instead?
> 
> Probably not ... pmRegisterDerived() can be (and most often is) called prior
> to any PMAPI context being created ... which is why it is thread-private in
> the current implementation.

Hm, could that early initialization be deferred until the first
pmNewContext type call?  Before that point, there's no way to do use
the info anyway.  (Maybe something like the event.* metric cleanup you
did a few weeks back?)  Or in the early-init non-context case, drop or
print the error message and not store it?

- FChE

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