[Bug other/15194] [fastjar] Check for "long long" before using it

[Kelley Cook]

This trivial patch should fix PR/15194, which was a regression from 3.2 
present on 3.[345].

It was bootstrapped (with no object code changes) on i686-pc-cygwin.
Unfortunately, I do not have ready access to a compiler without long 
long support, so I cannot absolutely verify does fix the reporter's bug; 
though, clearly, it cannot make things worse.

OK to install?

As an aside, I've come to believe that it would be appropriate to 
arrange for the toplevel Makefile so that, unless requested with a 
--without-system-jar, when bootstrapping only compile fastjar for the 
*build* system if there is not already a suitable "jar" present.  I 
think that it currently gets built and goes unused.

Clearly in a cross compile you would need to build fastjar for the host 
system.

I'll think through the logic of that over the weekend, but any comments 
from the build experts would be welcome.

Kelley Cook
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: pr15194.patch
URL: <http://gcc.gnu.org/pipermail/java-patches/attachments/20040625/4e91eb8a/attachment.ksh>