> (switching to regular email... these nested conversations via reviewboard
> get awkward IMO)
> I didn't say I won't complain ;-).  However, if we're going to switch, let's
> be serious about it and not go halfway.
Sounds reasonable.

> It didn't occur to me before that there are several other arguments that get
> affected: --trace-help, --trace-start, --trace-file, and --trace-ignore.
>  Are you planning on just doing s/trace/debug/ on all of these?  I guess
> that's OK; the only one I worry about is --debug-help, which seems like it's
> asking for a lot more than just listing flags.  I don't think we need error
> messages on all of them (since none of them make sense w/o --trace-flags
> anyway).
I agree. I'd only put the error message on --trace-flags.

Instead of --debug-help, we could just do --debug-flags-help

> The other interesting thing is that --debug-break now fits in with this
> group, for better or for worse.  I like that it's not so lonely anymore, but
> it's the only one that's not related to flags.
I was actually thinking that it was nice.  Debug::breakpoint() was
something that I wanted to enable/disable with flags.  For example,
it'd be nice to have intelligently placed breakpoints for things like
startup(), module initialization (if using m5 as a python lib),
instantiation, etc.  The one that I want the most is startup() so that
the simulator stops and you can do things like call functions.

  Nate
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev

Reply via email to