Re: [cs] pipe infrastructure support

[Mike Stump <mrs at apple dot com>]

On Wednesday, September 24, 2003, at 04:03 PM, Zack Weinberg wrote:
> My understanding is that collect2's only function, for systems with a
> modern GNU linker, is handling -frepo mode template instantiation.
> Mike had a patch to avoid the use of collect2 entirely when it wasn't
> going to do anything - whatever happened to that?

On the back burner, the 1-2% speed up, isn't better than what I'm 
working on now, and will be a bit of work to finish for all platforms 
and we already have that speed up but having a broken -frepo.  It might 
be possible to start the job and stage the work in, but doing it in 
stages might be worse.  I kinda wanted to just do the whole thing all 
at once.  If done in stages, you can do the simple case, no AIX, no 
Solaris, no HPUX, and be 96% of the way there for all platforms.  The 
few odd balls can use the external collect2, until such time as the 
functionality is folded into collect2 the subroutine.  Or maybe a port 
expert can stick it in.  There is no way it could be worse than 
collect2 today, I'd say, just let them play on it till it works.  The 
issues are, all the great target hooks in tm.h and friends don't apply 
to gcc.c.

And folding gcc.c into cc1plus, well, is actually folding all the 
compilers into gcc.c, a much harder task.