pcp
[Top] [All Lists]

Re: pmdapapi buglets from Coverity scan

To: Lukas Berk <lberk@xxxxxxxxxx>
Subject: Re: pmdapapi buglets from Coverity scan
From: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Date: Thu, 26 Feb 2015 06:30:52 -0500
Cc: Nathan Scott <nathans@xxxxxxxxxx>, pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <87mw41s46o.fsf@xxxxxxxxxx>
References: <1730959537.2690842.1423612674835.JavaMail.zimbra@xxxxxxxxxx> <8761atz99h.fsf@xxxxxxxxxx> <y0mtwyaer3q.fsf@xxxxxxxx> <87mw41s46o.fsf@xxxxxxxxxx>
User-agent: Mutt/1.4.2.2i
Hi, Lukas -


> [...]  How about the following small patch in addition to yours?  It
> removes any warnings on startup, as we don't enable all the metrics
> we probe initially, and then returns the default to 120 seconds.

The effect of that patch would be not only to just disable auto-enable
during the initial pmda install, but on every subsequent pmcd restart.
That's harsh.


> The qa would need updating again, which I can provide.  I would also
> like to remove the phrasing that warnings are reasonable in the
> testsuite, imo in this case, they aren't.

The warnings in this case are apropos: some counter values are just
not available for the asking.  By turning auto-enable off during
install, as in your patch, this fact is obscured from the
pmprobe/user, because then no counters are even attempted to be used!
If the point of the pmdaInstall "pmprobe -i" is to exercise the pmda,
then let's exercise it.


> On a side note, I'm not sure what the changes in pmdapapi.1 are
> supposed to mean?  Are they mistakes? "#true -- ..." ?

They're just comments, explaining each command.  They're executable,
in case a user copy&pastes that section of the man page into a shell.


- FChE

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