On Thu, Feb 26, 2004 at 12:12:13AM -0800, Jordan Hubbard wrote:
>
> On Feb 23, 2004, at 11:34 PM, Paul Guyot wrote:
>
> >? (At) 22:16 -0800 23/02/04, Jordan Hubbard ?crivait (wrote) :
>
> No, and we really should just pick one of the many many tcl debuggers
> out there and hook it in enough that you can debug a port using "port
> debug" or some similarly easy way into it. It doesn't have to be an
> ideal experience, just basic. I think everyone looking into this up to
> now has gone in hoping to find something like gdb for tcl and left
> disappointed when they couldn't find anything like that. We need to
> set our sights a little lower here and be happy to have ANY debugger
> rather than continue to accept "none" as an acceptable state of
> affairs. :-)
This suggestion obviously requires a bit of knowledge about Tcl internals,
but it works well enough for just me: Run Tcl under gdb; break on the Tcl
functions you want to examine more closely.
It's not too painful to step through execution this way, though it's
obviously not an optimal solution.
Also - Does anyone object to NOT calling portexec {} in a catch {} in
port(1) if the -d (debug) flag is used? That would provide a backtrace
on most errors.
-landonf
|
