A better console.log

Webkit (and Firefox?) emits live objects via console.log, instead of strings. Nothing is more frustrating than tracking down odd behavior, only to realize way too late that you've had unreliable logs. That's why we don't use console.log; instead we've installed a global logger object which always prints real strings. As a bonus, it keeps a ring-buffer of log messages and on logger.error events, it posts that buffer to the server.

EDIT: Here's a Gist: https://gist.github.com/2210783

Don't we already have console.trace?

As a curiosity, the bult-in logging module from Python provides the same features (file name, line number): http://docs.python.org/library/logging.html#logging.LogRecor...

Please, people. If we mean "use" can't we just use the word "use"?

Sometimes I think there's a special place in hell reserved where you are fooled into thinking you can get out if you can add just... one... more... layer... of logging abstraction.

Or maybe this is a development phase all programmers go through where they fixate on the mechanism for finding bugs instead of the bugs.

I mean, sure: logging files and line numbers with messages is useful sometimes. I think we all do it when chasing bugs. But is this really something that needs to be abstracted and pickled for posterity? Imagine looking at dmesg or syslog output and seeing all those source files you'll never read.