PATCH: Speed up libgcj build

[Bryce McKinlay]

Alexandre Oliva wrote:

> > I think Alex said the -I flag was for inlining methods across
> > classes.
>
> Not only that.  I'm pretty sure there was something that broke without
> it, but I'd have to build from scratch to remember what it was :-(

When I removed the -I flag, libtool gave an "unable to infer tag configuration" error. Adding the "--tag=GCJ" argument to the
libtool command fixed that. I havn't come across any other problems.

> >> 2. Can someone remind me why the .java->.class stage was changed to run
> >> gcj separately for every file?
>
> To avoid limits on command-line size.
>
> >> This is now the slowest part of the compile by far.
>
> Wasn't it already?

Well, with the -I flag, the .java->native part was even slower. And previously I could (theoretically, since libgcj still has
enough gcj-isms that it doesn't build on a "regular" java compiler) just substitute jikes for "gcj -C" and get the .class
files built extremely fast.

I agree with Jeff that reordering the classes in the makefile (java.lang first, then java.util, etc) should help.

regards

  [ bryce ]