Re: rhug status

[Bart Locanthi <bart at sabl dot com>]

ahhh.. this is my bad.

with the dependency issue that seemed to be there (the build immediately 
failed on gnu.readline, perhaps because of objdir==srcdir), i wrote my 
own makefiles and didn't set that flag.

btw the art of efficiently making gcj targets seems to elude both make 
and ant. is there a best standard practice for this?

i understand the intent of using traditional .java->.class via make to 
bootstrap, but the ant stage of doing it en masse (yay!) still leaves 
the issue of generating .o's, which is almost as slow as going straight 
to .o from .java.

i've taken to having rules like below and loading the few resulting .o's.

   gcj -c -o dir.o -classpath dir `find dir -name '*.java'`

you need a slightly more complex thing when other classpath entries are 
necessary.

Anthony Green wrote:

> On Mon, 2003-09-08 at 10:08, Bart Locanthi wrote:
>  
>
> > gcj has been barfing on stuff in comments in BCEL and xalan. in the 
> > former, the literature reference symbol (as in $3.5.2 where $ is 
> > replaced by something that looks like a line integral), in the latter, 
> > stuff like u-umlaut in Fahrfergnugen. i elided these rather clumsily so 
> > i have no line numbers for you.
> >    
>
> This shouldn't be happening.  Do you see --encoding=8859_1 on the
> compile line?  And you say this is with GCC 3.3.1 on a RHL9 machine? 
> How did you configure the compiler when you built it?
>
> BTW, I just checked in a fix for the jakarta-regexp classpath problem in
> Xalan.
>
> Thanks,
>
> AG
>
>