Subject: Re: libpthread -- document libpthread debugging technique?
To: None <,>
From: None <>
List: tech-kern
Date: 06/24/2003 12:19:09
On Friday 20 June 2003 16:59, Jason Thorpe wrote:
> No, I don't.  While I don't expect everyone to be an expert, I do 
> expect a little more effort that I seem to be seeing.
> The pthread library has a number of debugging options.  There is a 
> debug log that you can run a dmesg-like program on to get pretty 
> detailed information on every action the library takes.  There are also 
> various application debugging options (note that a fair number of 
> things that were blamed on libpthread at first turned out to be bugs in 
> the applications!).

Okay, I have a (hopefully helpful) suggestion here.

Whenever I see a problem with a libpthread assert, or a coredump, or some 
weird set of processes just sitting there unkillable, I would love nothing 
more than to dive in with a debugger and try to figure out what's going on. 
Unfortunately what many of us eager bucks lack is the technique to delve.

I knew there was a debugging library, attempted to use it a few times, but 
never saw any log, nor any dmesg-like output.

So here's the suggestion: How do I do it? Run me (and people like me) through 
a few debugging sessions and show us how *you* debug a threaded app problem 
(even if it's just capturing the next problem you run through) and believe 
me, more of us will jump on the bandwagon. I did list a few problems to the 
mailing list, but I got the strong impression I wasn't helping matters 
because I was unable to define precisely what the problem was or where it was 
occurring in the software I was crashing.

Anyway, believe us, we want to help more. We just need a leg up. :) Book 
recommendations would be terrific also. :)


> When people see failures, they should use the tools at their disposal 
> to get as much information as possible about the failure.  You don't 
> have to be an expert in how the SA/pthreads system works to do this.

I agree!

> I know of one person (Bill Studenmund, an expert in a number of areas, 
> but whom I suspect will not claim to be an expert in how SA/pthreads 
> works :-) who has had great success at getting pthreads bugs fixed by 
> providing very thorough debugging information to Nathan when he has a 
> problem.

Got it!