Re: Call for testers for pch-branch

[Geoff Keating <geoffk at geoffk dot org>]

Andreas Schwab <schwab@suse.de> writes:

> Geoff Keating <geoffk@geoffk.org> writes:
> 
> |> I do know that all the PCH tests using -g fail on dwarf2 targets, I
> |> will look into it later.  I'm not sure if the generated debugging
> |> information is wrong or just different (but it doesn't matter, I don't
> |> want to hand-verify whether the two dwarf2 output sequences mean the
> |> same thing).
> 
> Is this an instance of this bug, or something else?
> 
> /tmp/cvs/gcc-20021201/Build/gcc/gcj -B/tmp/cvs/gcc-20021201/Build/ia64-suse-linux/libjava/ -B/tmp/cvs/gcc-20021201/Build/gcc/ --encoding=UTF-8 -C -g -classpath '' -bootclasspath /tmp/cvs/gcc-20021201/Build/ia64-suse-linux/libjava:../../../libjava \
>              -d /tmp/cvs/gcc-20021201/Build/ia64-suse-linux/libjava ../../../libjava/gnu/awt/xlib/XToolkit.java
> ../../gcc/gcjh -classpath '' -bootclasspath . java/lang/AbstractMethodError
> make[2]: *** [java/lang/AbstractMethodError.h] Segmentation fault

No, that's definitely something else.  The dwarf2 problem just affects
the dwarf2 output, making it different (and maybe not even incorrect)
when a PCH file is used.  It doesn't affect non-PCH compilation.

However, it is not unlikely that there will be problems with java.
There are significant changes to java on the branch that I tested when
I made, but can't test any more because java isn't enabled for Darwin
or any simulator I can build on Darwin.

If you can come up with a testcase, I can try building a cross-cc1java
and try debugging this.  Chances are it's just some simple merge bug.
Even a backtrace would help.

-- 
- Geoffrey Keating <geoffk@geoffk.org>