log user tracebacks in the kernel (INFO-level)

[minrk]

Should make debugging errors a little less painful

should close #2473

I don't know why I didn't issue this PR earlier, this fix is from months ago.

[ellisonbg]

I realize it is nice to have this be possible, but I think that putting these messages at debug level INFO is a bit to aggressive. I would prefer DEBUG. I have tested this and it does what is say it does. I run at INFO level somewhat regularly and wouldn't always want to see my regular user exceptions in the log.

[minrk]

I see your point, but I wanted to draw a line between user exceptions and the massive mostly useless barf that is debug output. Currently there are exactly two messages at the info level (abort and eventloop transition), so I think INFO is actually the right way to go.

[ellisonbg]

OK, I think that is fine, I think this is ready for merge then.

On Wed, Jan 23, 2013 at 10:43 AM, Min RK notifications@github.com wrote:

I see your point, but I wanted to draw a line between user exceptions and
the massive mostly useless barf that is debug output. Currently there are
exactly two messages at the info level (abort and eventloop transition), so
I think INFO is actually the right way to go.

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/2818#issuecomment-12615109.

Brian E. Granger
Cal Poly State University, San Luis Obispo
bgranger@calpoly.edu and ellisonbg@gmail.com

[ellisonbg]

Ahh yes, I forgot that the kernel truly barfs everyting in DEBUG mode. We should probably fix that as it is too verbose to be useful for anything. Good call.

[minrk]

debug output is a challenge - if we got rid of that information, we would have no access to it if we were having problems with messaging. It makes sense for there to be a way to see all of the message traffic, but perhaps we would need a separate debug_all_messages flag, rather than implying this simply with the debug log-level.