DRDT: breakpoint on data race?
Hi,
I just started to try DRDT, and it looks very promising. I understand that things like missing stack traces are due to the beta status of the tool. I am also a bit concerned about the performance: I used it with an application which is about to be transformed from single threaded to multithreaded, and I get about 800 data races. This needs, on a Blade-1500, more than five minutes until the list of races is displayed in rdt.
It is immediately apparent, when looking at these data race stack traces, that it would be VERY useful to have a feature like "breakpoint on data race" in dbx! That would probably mean that the event mechanism in dbx would have to be extended accordingly, and that this would only work with instrumented executables, but this could be an invaluable feature.
Purify has something like that with memory errors: they invoke a function "purify_stop_here" on errors they detect, and developers can set breakpoints on this function. This is extremely useful.
Regards
Dieter Ruppert
