Re: Small example of livelock regression in garbage collector forGCJ 3.3 under W

[Andrew Haley <aph at redhat dot com>]

Ranjit Mathew writes:
 > >  > > So, the only libgcj that needs to be built is by the linux->mingw
 > >  > > compiler, and that libgcj must be built with the correct libgcj.spec.
 > >  > > Everything should work fine, modulo bugs in the configury.  I guess
 > >  > > it's just a matter of trying it.
 > >  > 
 > >  > I tried this last night and as expected
 > >  > "libjava/configure.host" is not touched if I give
 > >  > "--disable-libgcj" while building the cross GCJ.
 > >  > 
 > >  > So there's no libgcj.spec created that contains
 > >  > the necessary flags and hence the crossed-native
 > >  > libgcj is built incorrectly.
 > > 
 > > So what libgcj.spec does the compiler use while building libgcj?  This
 > > sounds totally screwy.
 > 
 > None. :-/
 > 
 > Is it *that* terrible?

There must be something badly wrong, because the configure process
that creates libgcj.spec does a whole bunch of other things, without
which you can't build libgcj at all.  So, we need to know why some
part of the configure process isn't happening.

Andrew.