PATCH: Support stabs in libjava test

["H . J . Lu" <hjl at lucon dot org>]

When stabs is used, I get

/tmp/ccgCe9gi.s: Assembler messages:^M
/tmp/ccgCe9gi.s:29: Warning: .stabs: description field 'a0004' too big, try a
different debug format^M
/tmp/ccgCe9gi.s:263: Warning: .stabs: description field '10001' too big, try a
different debug format^M

and the test is marked "FAIL". I am enclosing a patch to filter it out.
The interesting results on Linux/x86 are:

1. DWARF 2:

                === libjava Summary ===

# of expected passes            2073
# of unexpected failures        1
# of expected failures          18
# of untested testcases         15

output is:
/tmp/ccMCVWbf.o(.data+0x64): undefined reference to `support::Waldo::class$'^M

2. stabs


                === libjava Summary ===

# of expected passes            2075
# of unexpected successes       4
# of expected failures          14
# of untested testcases         6
XPASS: Where compilation from source
XPASS: Where -O compilation from source
XPASS: inner_1 compilation from source
XPASS: inner_1 -O compilation from source

One example:
/tmp/ccO3Lvkg.s: Assembler messages:^M
/tmp/ccO3Lvkg.s:26: Warning: .stabs: description field '90006' too big, try a
different debug format^M
/tmp/ccO3Lvkg.s:53: Warning: .stabs: description field '10001' too big, try a
different debug format^M
/tmp/ccR3ol5j.o: In function `Where::Where[in-charge]()':^M
/home/hjl/work/gnu/src/gcc-3.1/gcc/libjava/testsuite/libjava.compile/Where.java:1:
undefined reference to `support::Waldo::class$'^M
collect2: ld returned 1 exit status^M

It looks like some area of dwarf support in gcc/ld is not as good as
stabs. Shouldn't we fix gcc/ld for that?


H.J.

Attachment: libjava-test.patch
Description: Text document