Re: [patch 2/2] Implement multi-component --with-auto-load-dir

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Wed, 09 May 2012 22:51:19 +0200, Joel Brobecker wrote:
> > With Fedora scl-utils ("alternative packaging") you have both regular
> > 	/usr/share/gdb/auto-load/
> > and also the "alternative"
> > 	/opt/rh/upgradeset/root/usr/share/gdb/auto-load/
> 
> I see.  I wish we could manage both with one setting.  Since we need
> a "list" of directories, I'd keep "auto-load script-dir" (should we make
> that script-dir plural?)...

We already have multi-directory "set debug-file-directory".

(It may be called so as it was a single directory in the past; OTOH if you
insist on dir/directories we should rename it and make the original setting
obsolete-compatible.)

We also have command "directory" which properly parses multiple directories:

(gdb) directory /tmp/a:/tmp/b
Warning: /tmp/a: No such file or directory.
Warning: /tmp/b: No such file or directory.
Source directories searched: /tmp/a:/tmp/b:$cdir:$cwd

We can argue about singular vs. plural form like there are many reasons for
both cases in which form to name SQL tables.


> I can see how things can be awkward for turning the safety guard off
> (the equivalent of what we used to do with "auto-load safe-path").
> I think I'd just make that a boolean setting instead.

Do you mean to make 'auto-load safe-path [on|off]'?  This would be
coincidentally fine for me as I both face untrusted inferiors and I dislike
any auto-loads.  But it would be too dangerous for anyone who likes
auto-loaded scripts and who also faces untrusted inferiors.


> > > debugging /work/build/gcc-feature-a/gcc/cc1, you don't need to have
> > > cc1-gdb.gdb defined in your auto-load directory as
> > > auto-load/work/build/gcc-feature-a/gcc/cc1-gdb.gdb. It'd just be
> > > sufficient to have cc1-gdb.gdb defined in your scripts-directory.
[...]
> No, they are not related, and you shouldn't have to worry about that
> (unless you want to!). It's a feature I think some users would enjoy
> having, but if I want it, I should be the one taking care of the
> implementation.

I understand it would be useful; I just do not like any auto-loads myself to
try figure out the best way to do it.  Also /usr/share/gdb/auto-load/ has
never been an idea of mine.


Thanks,
Jan