Re: [PATCH]: C++ mangling patch that is about to be committed

[Michael Snyder <msnyder at redhat dot com>]

Daniel Berlin wrote:
> 
> Kevin Buettner <kevinb@cygnus.com> writes:
> 
> > On Oct 10, 11:55am, Daniel Berlin wrote:
> >
> > > Elena Zannoni <ezannoni@cygnus.com> writes:
> > >
> > > > BTW, I thought
> > > > we agreed to leave the do--while construct in the
> > > > SYMBOL_INIT_DEMANGLED_NAME macro.
> > > I'd rather not.
> > > It's not used in if statements, and *never* should be.
> > > The argument that someone, someday, might want to, just isn't
> > > convincing, because they shouldn't.
> >
> > Daniel,
> >
> > This is really not the way to handle this kind of change.  Elena is
> > right.  The consensus was to leave the ``do ...  while (0)'' construct
> > in the SYMBOL_INIT_DEMANGLED_NAME macro.
> 
> No, there was no consensus, we only really discussed *why* it existed that
> way. Nobody provided a convincing reason to keep it the old way, except "It
> was that way".

If it was discussed and there was no consensus, then 
the DEFAULT action should be "no change".  Especially
if the maintainer responsible for that file did not
approve the change.

> > I believe that turning it
> > into a proper function was also discussed and is regarded as a viable,
> > perhaps even superior, alternative.
> 
> Feel free to do this.
> I'm not playing with symbol tables anymore, I don't have time to wait
> for my patches to get approved.

I'm sorry you feel that way.  Waiting for patches to be approved
is part of the consensus that must exist, to work on a group 
project.  I hope you can get used to it.

> > I know that you don't like the do ... while (0) construct, but it is
> > not right for you to try to sneak changes eliminating it past the
> > maintainer multiple times, particularly when you agreed to leave it
> > in.
> 
> I haven't tried to sneak it in, the elimination exists in every
> repository i've applied any version of the C++ patches to, so i have to hand
> patch every single diff I generate to put it back.
> 
> It's not like I can just magically make it reappear in all the
> repository copies I have around.
> 
> The reality is it exists in the one repository that matters in this
> case:
> #define SYMBOL_INIT_LANGUAGE_SPECIFIC(symbol,language)                  \
>   do {                                                                  \
>     SYMBOL_LANGUAGE (symbol) = language;                                \
>     if (SYMBOL_LANGUAGE (symbol) == language_cplus                      \
>         || SYMBOL_LANGUAGE (symbol) == language_java                    \
>         )                                                               \
>       {                                                                 \
>         SYMBOL_CPLUS_DEMANGLED_NAME (symbol) = NULL;                    \
>       }                                                                 \
>     else if (SYMBOL_LANGUAGE (symbol) == language_chill)                \
>       {                                                                 \
>         SYMBOL_CHILL_DEMANGLED_NAME (symbol) = NULL;                    \
>       }                                                                 \
>     else                                                                \
>       {                                                                 \
>         memset (&(symbol)->ginfo.language_specific, 0,                  \
>                 sizeof ((symbol)->ginfo.language_specific));            \
>       }                                                                 \
>   } while (0)
> 
> 
> I really resent the implication.
> I split the patches and did modifications based on the existing set of
> patches, which is why the patch has it.
> 
> >
> > Kevin