Re: [PATCH]: build problems with command line length (ksh93)

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Jeff" == Jeff Sturm <jsturm@one-point.com> writes:

Jeff> I've done some work on Makefile.am to avoid command line length
Jeff> problems once and for all.

This is excellent.

Jeff> -## JAVAC is set to `$(GCJ) -C'.  However, JAVAC is run from the srcdir
Jeff> -## (due to problems running it in builddir).  In this case the obvious
Jeff> -## definition of GCJ won't work; instead we resort to this ugly
Jeff> -## definition that gives us absolute paths.

Jeff> I can't tell what this refers to exactly, nor can I figure out why
Jeff> gcj -C should have any problem running in builddir.

This is a dead comment.  In the past we did this because (I think)
path-searching was a bit broken in gcj.  However, running gcj in
srcdir meant that M-x next-error in Emacs didn't work correctly.  So
at some point I changed it back.


Jeff> -## FIXME: this ought to depend on built_java_source_files, but right
Jeff> -## now it can't.  Ugly.

Jeff> No real explanation, again, so I tried a straightforward dependency.
Jeff> It worked, and so I removed the comment.

There might have been a problem here if you added a new .java file.


Jeff>  ## This little nastiness is here so that the backquoted stuff in the
Jeff>  ## GCJ definition can be correctly expanded, if required.

Jeff> I assume this refers to the definition of $(JAVAC).  I'm not aware of
Jeff> any such use of backquotes, but I left this just in case.

You can delete this one too.


Jeff> Tested on trunk with alphapca56-unknown-linux-gnu.  OK to commit?

Yes, thanks.

Tom