Re: Gtk2 1.210 (unstable) available

[Thierry Vignaud <tvignaud mandriva com>]

Torsten Schoenfeld <kaffeetisch gmx de> writes:

> > This still doesn't include muppet's work on shrinking big number of
> > relocations done in perl-Gtk2 on startup... :-(
>
> Yes, because the discussion wasn't really conclusive.  The patch is
> huge, contains some evil, and the Gtk2.xs part is rather fragile.
> When timing this last year, I found:
>
> > [...] a 9% size reduction of the shared object (11MiB -> 10MiB, with
> > debugging symbols) and a 7% start speedup (0.067s -> 0.062s,
> > measured via timing many runs of `perl -MGtk2 -e'exit'Â).
>
> These numbers don't justify such an invasive change, I think.

well, if a compiler optimization brings a 7% startup speedup, it would
be considering amazing.
So if we can got it for free by preventing the linker to do stupid and
useless things, I think that it justifies it.
 
> Rather, I'm with Kevin: why don't we change xsubpp (or
> ExtUtils::ParseXS actually) to have an option that makes it declare
> every xsub (except for the boot xsubs) as 'static'?
> <https://rt.cpan.org/Ticket/Display.html?id=42574>.

Why not...
I'm not against going that way.