pcp
[Top] [All Lists]

Re: [pcp] pcp updates - pmdapapi update

To: Lukas Berk <lberk@xxxxxxxxxx>
Subject: Re: [pcp] pcp updates - pmdapapi update
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Fri, 14 Nov 2014 03:53:05 -0500 (EST)
Cc: pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <87oascow3f.fsf@xxxxxxxxxx>
References: <87oascow3f.fsf@xxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: khhgzDkZJ+KvbXTW3bNITYrHj5v7uA==
Thread-topic: pcp updates - pmdapapi update
Hi Lukas,

----- Original Message -----
> Hi,
> 
> The PAPI pmda has gone through a fair bit of work.  The changes can be
> found at: git://sourceware.org/git/pcpfans.git lberk/papi or gitweb:
> https://www.sourceware.org/git/gitweb.cgi?p=pcpfans.git;a=shortlog;h=refs/heads/lberk/papi
> 
> This iteration features much improved internals, additional qa,
> [...]

This is looking pretty good - I've updated some minor stuff (please do
a review) but otherwise its all merged now.  There's a few things that
would be good to still get in that I didn't tackle...

- I'm not following why the (static) metrictab has all those entries
  for event counters still?  Can those be removed now?

- The permissions_check() changed to only look for uid==0 - should we
  drop all references to gid* elsewhere?  I think so (slims down some
  data structures, simplifies the attributes callback, etc)

- As per earlier mail, there's no error handling for auto-fu counters.
  Simplest will be, I think, to give refresh_metric() a parameter that
  indicates whether its caller is going to ignore the return code, and
  in that case, handle_papi_error() should really log any errors into
  the pmdapapi.log file.  In the pmStore-enable/disable-metrics case,
  these need not be logged (the client tool will get notified about an
  error in that case, which is better - keep that pmDebug diagnostic
  for both cases though).

Thanks!

--
Nathan

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