Re: Error message revelation

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Keith" == Keith Seitz <keiths@cygnus.com> writes:

Keith> The problem is that we have TWO ways of returning information
Keith> to the tcl code: the interpreter's result and result_ptr. I
Keith> really think that both of these methods are needed in our
Keith> case. The confusion comes (as you've noted) that there doesn't
Keith> appear to be any consistent style by the rest of the code, so
Keith> no one (including me) really knows what is supposed to be "the
Keith> right way". (Something more for my styleguide/internals doc).

Keith> I like the explicit method: if YOU (a command writer in gdbtk)
Keith> say GDBTK_IN_TCL_RESULT, then the result from tcl is used,
Keith> otherwise, it's the one you've built in result_ptr->obj_ptr.

I have a question: is it more common that you'd want to use the Tcl
result, or is it more common that you'd want to use the gdb result?
That is, maybe it makes sense to invert the flag to
GDBTK_IN_GDB_RESULT.

You would know this better than me.  In fact, I think really you
should do whatever you think is best, since you'll be working on this
code much, much more than I will.  I think as long as the convention
is documented and clearly used in all the gdbtk code, then it will be
fine in the long run.  That's particularly true if you use a helper
macro and have some automatic way to test for new bugs being added
(this is a technique we use in Automake, and it really does help -- we
have a "maintainer-check" target which checks for the checkable parts
of our coding standard.  We just run this occasionally, eg before
releases, and fix whatever it points out).

Checking in a jumbo patch to fix all the gdbtk*.c problems would be
really cool.  Unfortunately it is a daunting amount of code.  Finding
a day or two to do cleanup is hard :-(

Tom