Re: [RFC] Alternate approach to keeping convenience variables

[Daniel Jacobowitz <drow@false.org>]

On Wed, Jan 04, 2006 at 10:37:06AM -0800, Jim Blandy wrote:
> On 1/4/06, Andrew STUBBS <andrew.stubbs@st.com> wrote:
> > Perhaps you could modify copy_type_recursive such that it creates a
> > cleanup chain specific to the internal variable. Then call do cleanups
> > in set_internal_var.
> >
> > Might be tricky with shared type copies though. Just a thought.
> 
> Exactly.
> 
> You can keep throwing complexity into the code to deal with this stuff
> until it buries you.  And it'll never be a full solution.  Or it'll be
> complete until someone wants to add some perfectly reasonable feature
> (just as is happening with "let's preserve the values of convenience
> variables across symfile loads" right now).  At some point, we have to
> take a step back and think about a complete solution.

Right.  For now, I'm OK with saying "this is a small memory leak when
symbol files are unloaded".  If someone wants to investigate adding
boehm-gc to GDB and using that, instead, then I'm OK with that too :-)

Every once in a while I do get the urge to go through GDB with a
leak checker.  But not often enough to jump through huge, rarely tested
hoops for that case.

-- 
Daniel Jacobowitz
CodeSourcery