pcp
[Top] [All Lists]

Re: PCP 2.3.0 dumps core on Linux system without module support

To: Mark Goodwin <markgw@xxxxxxx>
Subject: Re: PCP 2.3.0 dumps core on Linux system without module support
From: Mike Mason <mmlnx@xxxxxxxxxx>
Date: Fri, 06 Dec 2002 10:25:31 -0800
Cc: pcp@xxxxxxxxxxx
References: <Pine.LNX.4.44.0212052116070.7991-100000@sherman.melbourne.sgi.com>
Sender: pcp-bounce@xxxxxxxxxxx
User-agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.0.1) Gecko/20020823 Netscape/7.0
If PCP is adopted by the Carrier Grade Linux project (as it appears it will be) I think we'll see broader test coverage and at least a few more contributors get involved. Also, have you considered trying to get PCP included in United Linux? My group had some success in getting UL to adopt some of our projects, but it's been more difficult with other distros.

BTW, I still plan to submit a patch later today or at least by Monday.

Mark Goodwin wrote:
On Wed, 4 Dec 2002, Mike Mason wrote:

As more people are encouraged to contribute to PCP, perhaps we need to
move to the dual "stable" and "development" release track model.

I think a dual release track model is a good idea in the open source environment. I would suggest moving to it sooner rather than later. It allows us to utilize the collective resources of the "community" during testing. ...


understanding of how the open source community works.  I've been told
"release early and often", gather comments and bug reports, fix the
bugs, release again, and so on, until you become comfortable enough with
the code to make an official release.  Granted, we shouldn't knowingly
put out buggy code, at least not without caveats, but I don't think
finding bugs during testing is necessarily a bad thing.


well actually, this is the model I have been following by publishing the latest pcp src in the dev directory. And it seems to have worked well in terms of fleshing out bugs. I certainly agree with the "community effort" style of development wherever it works, and "release early and often" has worked well for many other projects, but in the case of PCP I'd seriously like to see more people contributing and testing (other than Mike and Todd for this release .. not to say nobody else has contributed over the years).

In any case, at least we nailed a few bugs by testing on the unified linux
release that would probably not have been found otherwise; the ksym code worked
fine on RH machines inside SGI.

Mike, do you have a patch for ksyms for the case with no module support
in the kernel and missing /proc/ksyms? (or will I work on that?)

Thanks
-- Mark





--
Mike Mason
IBM Linux Technology Center
Beaverton, OR, USA
mmlnx@xxxxxxxxxx
503-578-4123


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