Re: [Texmacs-dev] [Announce] nogencc-0.6

[Joris van der Hoeven]

On Mon, 6 May 2002, David Allouche wrote:

> On Monday 06 May 2002 11:34, Joris van der Hoeven wrote:
> > Why is C++ without automatic templates a pain? It would be nice
> > to compare code size and efficiency before taking a decision.
> 
> When you have automatic templates disabled, you must somehow implement a 
> way to specify explicit instanciations.
> 
> There are essentially two ways to do that:
> 
> Either by hand, specifying instanciation as the linker complains for 
> undefined symbols. That lead to equivalent code generation as with 
> automatic templates. In that approach one would allow automatic templates 
> in the compilation units that defines automatic instanciations. It is a 
> pain to define the first time, and then it will only accumulate cruft, 
> because no one will ever want to clean it up.
> 
> Or semi-automatically. Here the trick would be to define macros which do 
> explicit instanciations for a given set of extern (that is, not inline) 
> class and method templates. That macro would be used in a way similar to 
> the parameterized code_* modules. That approach is the closest possible 
> to what did gencc, but it has has no advantage over a automatic template 
> instanciation which is smart enoungh to avoid duplicate code generation. 
> Actually, it will generate more code, thus slowing down the compilation 
> and possibly leading to bigger binaries.
> 
> As you see, no solution is simple to implement, and even if I think that 
> the automatic template instanciation system of g++ is a pain, it is good 
> enough not to use explicit instanciation.

Hmm, it might be a good idea to check whether we win something
when disabling the automatic templates in the official distribution.
that will give us a good hint on whether we should torture ourselves
with this issue or not...

> > I am not responsable for your initiative to remove gencc.
> [snip]
> > I decided to do you a pleasure with removing gencc,
> > which takes me a lot of time and which was not a priority.
> > In your turn you should do me a pleasure and allow me
> > to adopt the changes in a smooth and rigourous way.
> 
> Ok, ok... I'll stop complaining.
> 
> That is getting on our nerves and wastes our time.

Hundred thanks.

> > OK, I will install 2.95.3 and see whether the problems persist.
> > But maybe 2.95.4 is better?
> 
> I did not found 2.95.4 on the GNU website either. It looks like a Debian 
> specific prerelease.

Yes, I also could not find it. So I will install 2.95.4 and 3.0.4.

> > Keep on the good work, keep calm, don't get angry and
> > smoke a sigaret at each time that you want to kill me
> > (when I am not there in preference).
> 
> Well basically, I think that I should get more in a "you boss, I worker" 
> state of mind. I could probably have a boss who pays more, but he may 
> also be more obnoxious.

Well, it is not really a question of boss and worker.
I appreciate your comments and ideas and I am willing to discuss
most of them. But I am also responsable for the global coherence
of the project, so we need to streamline the cooperation and
reduce to a minimum the risk of bugs. This can only be done
if we share some working conventions. Rigour is one of them.
Making the changes in a clean order is another one.
Also, at the moment, I see you more as a student than
as a worker: you have many good ideas, but you still need
a bit of direction to put them in practice in an efficient way.
Why make similar errors as I sometimes did in the past if
you precisely know someone who can avoid that?

Keep up the good work,

<Joris>