Re: [gcj-eclipse-merge-branch] MinGW ecj: Cross-built ecjx?! (was Re: ecj branch)

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Mohan" == Mohan Embar <gnustuff@thisiscool.com> writes:

Mohan> After spending some time hacking and slashing my way through the
Mohan> (linux,mingw) cross of gcj-eclipse-merge-branch, I'm calling it
Mohan> a night. The patch to regenerate the top-level configure 2.13
Mohan> is definitely needed.

Please check this in on the branch...

Mohan> After commenting out all the park/unpark
Mohan> related stuff and punting on the fact that ecjx doesn't seem
Mohan> to be getting installed to the right place, I'm scratching my
Mohan> head about the fact that ecjx(.exe) is being built by the cross
Mohan> compiler. Is this right? If I understand directly, ecjx is supposed
Mohan> to be equivalent to jc1, right? If so, shouldn't this be built
Mohan> by the build compiler?

ecj1 (which is what ecjx is called when it is installed) is sort of
like cpp, back in the days when we had a separate preprocessor.  gcj
will invoke ecj1 first, then pass its (class file) output to jc1 for
further compilation.

And yeah, I think I just made a mistake here.  ecj1 has to run on the
host, not the target.  So we have to use the build-x-host compiler,
not the build-x-target compiler, to build it.

Tom