Re: emacs gud does not work with libtool

[Stefan Monnier]

> Huh? AFAIK, I'm using the *same* heuristic that M-x gdb uses when invoked
> normally.

You keep the same heuristic as far as deciding which directory to use,
based on the executable's filename, but you change the heuristic used
to get the executable's filename from the command line.

I think the way gud tries to solve the problem of finding the source file,
is showing its limits and we need to add some way for the user to give more
info to gud so we can hopefully address all special cases, rather than
adding to gud some extra heuristics to deal with a handful more cases.

The two are not mutually exclusive, but if we just keep adding special
handling for special cases, it'll never end.  Furthermore, when a new
problem shows up, we can only hope to say "we'll try to fix it in the next
release" rather than just say "you simply need to take this extra step
because your situation is special".

I.e. I'm not necessarily opposed to your patch, but given the amount
of complaints we've received around this part of gud, I'd like to try
and think of a more long term solution.

As for your patch, I don't understand exactly what it tries to do.
It seems that it tries to skip the first argument if it doesn't start with
a dash.  This works for `libtool gdb -fullname file' but not for
`libtool -foo gdb -fullname file'.  Is that indeed what your patch does?
Was that the intention?  Also let's not forget that this part of the
code is used for all gud backends, so I don't know what kind of impact it
might have on perldb, for instance.


        Stefan