Re: gcj -bootclasspath doesn't work anymore?

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Per" == Per Bothner <per@bothner.com> writes:

Per> More to the point, it works if you put the -C first:
Per> gcj -C -bootclasspath ''  hello.java -v

I find that a bit surprising.  We need to document this restriction.

Per> Now why it doesn't work if you put the -bootclasspath before the
Per> -C is because process_comand in gcc.c processes -b and -V options
Per> at the start of the command line, before doing anything else.
Per> This is a crock, attempting to handle the historical use of -V -b
Per> even when the gcc driver changes incompatibly.

For some reason I thought -b and -V were deprecated.  I don't have a
reference though.

I have two concerns here.  First, requiring -bootclasspath after -C is
a bit odd; most GNU tools don't work that way.  Second, whether we can
rely on this continuing to work in future gcc releases.

The first can be partially addressed through documentation.  I suppose
the second could be addressed by adding a test case.


Still, why not ship a `javac'-like wrapper script?  I haven't seen an
argument against doing it.  Debian does it, presumably indicating that
they've seen a need for one.

Tom