Re: SIGSEGV in jc1

[Per Bothner <per at bothner dot com>]

"Boehm, Hans" <hans_boehm@hp.com> writes:

> I still see this failure both in the branch and in (my hacked version of)
> the trunk.  I believe it's reproducible by
> 
> 1) Downloading and unpacking
> http://www.hpl.hp.com/personal/Hans_Boehm/crcalc/download/CRCalc.tar.gz

Did that (again).

> 2) Making sure that JAVAC in the Makefile is a usable Java to class file
> compiler, and that the gcj on the path is the right one.
> 3) Typing "make rpn_calc" in the root directory.

kelso<1148>make rpn_calc JAVAC=/opt/jdk1.3/bin/javac
/opt/jdk1.3/bin/javac -d . UnaryCRFunction.java CR.java PrecisionOverflowError.java AbortedError.java TestCR.java
gcj -O -o rpn_calc --main=rpn_calc rpn_calc.java com/sgi/math/*.class
kelso<1149>./rpn_calc
34
34.00
kelso<1151>type gcj
gcj is /home/bothner/GNU/install-gcc3/bin/gcj

This is my cvs version of the gcc3 branch, checked out, built, and
installed yesterday evening, on my RH7 box.

> Based on the stack trace, I would guess that it also occurs in other cases
> in which you're compiling from .class files which include inner classes?

No.

> Is there an easy way to diable the -ffilelist-file mechanism to work around
> this?

You could try commenting out line 378 in jvspec.c:
      || (saw_C && java_files_count > 1)

We need to figure out what is going on here.  Could some other people try
Hans's testcase using a branch or trunk tree, and report the result?
Doing so shold be trivial, once gcc is built.
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/per/