Re: [cs] pipe infrastructure support

[Mike Stump <mrs at apple dot com>]

On Wednesday, September 24, 2003, at 01:44 PM, Benjamin Kosnik wrote:
> > I guess you don't want a c++ compiler that is 2.2-2.5x faster do you
> > then?  Some people actually care about compile speed.  Hopefully the
> > people that review this won't actually stand in the way of progress.
> > If you have a better way, put it in now, otherwise...
>
> I know! How about making the last piece of Apple-contributed work that
> "makes the c++ compiler 2-2.5x faster" actually work again?

That's not enough!  My increases are incremental on top of PCH.  If you 
remove PCH, then you have to come up with 5x-8x or more someplace else! 
 :-)

I know, I already have one of those patches in the compile server, we 
can just merge it in!

> If these compile-server bits are anything like the PCH support, Neil is
> right to be concerned about long-term maintainability: there is
> effectively no maintenance of PCH. Why is that?

It just is how things are.  An enterprising person that doesn't mind 
being a mind reader that likes abuse can try their hand at cleaning up 
one of the posted patches.  I don't think it would be that hard.  I 
might do it, if the problem remains unfixed by the time we want to 
merge in the compile server.

As for each, there are complexities in each, that we have to weigh the 
cost of verses the benefit.  The compile server should remove most all 
the benefit of pch, so, maybe we just remove the PCH code, if people 
want.  What it doesn't remove, is the once per server pch build time.  
Presently, cs uses pch to avoid that.

> Has Apple decided that the compile server is a better way to go? Or, 
> does PCH work on darwin,
> and is just broken everywhere else? Do tell.

It works on Darwin on our 3.3 branch.  It fails in 3.4 on Darwin.  
Geoff will have to fix this before it is a product.  He's doing 3.3.x 
now, and after that, he'll merge in 3.4 top of tree, then, 3.4 PCH 
functionality, or lack of it, will become even more critical, but, 
he'll merge in from the cs branch, so, we'll just pick up one of the 
already posted patches.  Hum...