Re: Small example of livelock regression in garbage collector forGCJ3.3 under W

[Ranjit Mathew <rmathew at hotmail dot com>]

>  > I meant not having to build libgcj while building the cross compiler - for
>  > the crossed native compiler, I realise that I need to rebuild libgcj.
>
> I'm not sure I understand what the cross compiler is for.  If it's
> only to build a native compiler surely you don't need C++ or Java,
> just a C compiler.  You might as well disable all other languages.

I build a crossed-native (build=linux, host=target=win32) GCJ
for Win32 using a cross (build=host=linux, target=win32) GCJ.

This two phase process is *much faster* and far more reliable
than a purely native Win32 build - more details on this process
are at:

http://ranjitmathew.tripod.com/phartz/gcj/bldgcj.html

The cross GCJ is used to build the libgcj for the
crossed-native GCJ (as it can't build the libgcj itself).

So "java" is needed and therefore "c++" and thus "c" - I
configure only with "--enable-languages=c,c++,java".


>  > If I need to affect the crossed-native compiler's libgcj, is there any way I
>  > can do this *without* having to build the cross compiler's libgcj as well?
>
> Well, I don't know what the crossed-native compiler is for, so I can't
> say.

The crossed-native GCJ is for all practical purposes as good
as a native GCJ for Win32!

In a bootstrapping native build, I can understand that the
configure.host changes can possibly affect the libgcj
build, but since I am using a cross GCJ (that doesn't have its
own libgcj or libgcj.spec) to build the final libgj, I am
not so sure the changes will be effective.

I'll investigate this issue a bit more.

Ranjit.

--
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/