Re: Has anyone seen this build error before?

[Tom Tromey <tromey at cygnus dot com>]

Mo> Humm, how many are there? If it is just one or two, could you post
Mo> them to the list or post a pointer to where I could find them?

Alex posted these:

Alex> I think Tom is talking about patches to gcj/297 and gcj/283
Alex> (Tom, please tell me if I'm wrong.) But I don't think it's going
Alex> to help with your case. Egcs isn't in a very good state today. I
Alex> can't even build it past libiberty/cplus-dem.c (I hit this when
Alex> I tried to reproduce your problem yesterday night.)

I think those are the ones.  I'm not positive -- it was some patches
you emailed me last week in response to the java.util.Collections
problem.

Note that today Mark switched on the new ABI so you have to configure
gcc specially to get Java to build.

Mo> I just want to get a working version of gcj so I can get some
Mo> pass/fail numbers from the Jacks test suite before you fix all the
Mo> problems it exposes :)

We don't do all that great right now:

    gcj:	Total	543	Passed	503	Skipped	1	Failed	39
    Number of tests skipped for each constraint:
	    1	pc

This is with the latest gcj, plus a couple lexer patches that haven't
gone in -- one which fixes the \uffff bug and resynchronizes the gcj
notion of identifier character with the latest Unicode tables, and
another one for `0 - 2147483648' which I haven't submitted yet.  If
you run the tests you'll probably get an additional 3-4 failures.

I've been fixing lexer bugs because the lexer is relatively
self-contained.

The biggest problem for us is that the constant expression code in gcj
is very underpowered.  There are a lot of cases it doesn't handle, and
it is only used for declarations (and not, as far as I could tell,
case labels).

We also fail all of the "@" tests since I haven't had time to change
the test suite to add a constraint covering that case.  There are
around 10 of these.

Tom