On Fri, 2008-07-11 at 13:02 -0700, MARK CALLAGHAN wrote: > I will be happy with any tracing capability. One of the early messages > mentioned using a debugger in place of tracing, and that would be > awful for diagnosing behavior.
Personally, i'd almost prefer a valgrind module that records machine state... or, for something not as heavy, a profiler that dumped state so you got a trace that way. or... something like jam() in NDB. DBUG is rather heavyweight, jam() is really cheap (we deploy with it enabled). It's basically a circular buffer of module,linenumber tuple of every time there's a jam() in the source. a dump of the jam buffer (known as trace file in ndb talk) is incredibly useful when diagnosing crashes that have occurred in the wild. -- Stewart Smith ([EMAIL PROTECTED]) http://www.flamingspork.com/
signature.asc
Description: This is a digitally signed message part
_______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
