Re: problems w/ -traditional-cpp

[Neil Booth <neil at daikokuya dot demon dot co dot uk>]

Jason R Thorpe wrote:-

> I'm finally getting around to testing the new -traditional-cpp on
> the NetBSD kernel .S files, and I've run into this:

Great!  Could you confirm this is GCC as of the last 12 hours?  I
fixed some header file issues about 12 hours ago.

> #0  0x808cd38 in paste_tokens (pfile=0x8412000, plhs=0xbfbfcf60, rhs=0x8415528)
>     at ../../../gcc/gcc/cppmacro.c:392

Uhh, there's no way you should be pasting tokens with the traditional
preprocessor.  This code should never be run; it's not surprising it
segfaults because things aren't set up for it to run 8-)

So this should be easy to track down.

> /usr/local/gnu/bin/x86_64-unknown-netbsd-gcc  -x assembler-with-cpp -traditional-cpp -D_LOCORE  -Dx86_64 -I.  -I../../../../arch -I../../../.. -nostdinc  -DMAXUSERS=2 -D_KERNEL -D_KERNEL_OPT -c /u1/netbsd/src/sys/arch/x86_64/compile/SIMICS/../../../../arch/x86_64/x86_64/microtime.S

I wonder if the problem is -assembler-with-cpp.

> This is because the output is incorrect.  Chunks of a header file are in
> there which should not be, and the ENTRY() macro did not get expanded
> properly.
> 
> If you want to tell me exactly which bits you want to try and debug this,
> let me know, and I'd be happy to pack them up for you.

Please give me the smallest set of files that shows a bug.  If you can
trim down the files that's great, but not necessary.  If it's not
obvious to someone who doesn't know the code, please also briefly
explain what part of the header file is being incorrectly included
or whatever.

Thanks,

Neil.