[coreboot] Breakage in building traverse/geos

[Philip Prindeville]

On 7/30/11 7:07 PM, Philip Prindeville wrote:
> I'm seeing breakage when building traverse/geos. Is this a known issue?
> 
> [philipp at builder ~/coreboot]$ make V=1
> ...
>     CC         romstage.inc
> gcc -m32 -Wa,--divide -fno-stack-protector -Wl,--build-id=none  -MMD -Isrc -Isrc/include -Ibuild -Isrc/arch/x86/include -Isrc/devices/oprom/include -include /home/philipp/coreboot/build/config.h -Os -pipe -g -nostdinc -nostdlib -Wall -Wundef -Wstrict-prototypes -Wmissing-prototypes -Wwrite-strings -Wredundant-decls -Wno-trigraphs -Wstrict-aliasing -Wshadow -Werror -fno-common -ffreestanding -fno-builtin -fomit-frame-pointer -D__PRE_RAM__ -Isrc -I. -Ibuild -c -S src/mainboard/traverse/geos/romstage.c -o build/mainboard/traverse/geos/romstage.pre.inc
>     POST       romstage.inc
> sed -e 's/\.rodata/.rom.data/g' -e 's/\.text/.section .rom.text/g' build/mainboard/traverse/geos/romstage.pre.inc > build/mainboard/traverse/geos/romstage.inc.tmp
> mv build/mainboard/traverse/geos/romstage.inc.tmp build/mainboard/traverse/geos/romstage.inc
>     GEN        crt0.S
> printf '#include "config.h"\n #include "src/arch/x86/init/prologue.inc"\n #include "src/cpu/x86/16bit/entry16.inc"\n #include "src/cpu/x86/32bit/entry32.inc"\n #include "src/cpu/x86/16bit/reset16.inc"\n #include "src/arch/x86/lib/id.inc"\n #include "src/cpu/x86/fpu_enable.inc"\n #include "src/cpu/amd/model_lx/cache_as_ram.inc"\n #include "mainboard/traverse/geos/romstage.inc"\n' > build/crt0.S
>     CC         mainboard/traverse/geos/crt0.s
> gcc -m32 -Wa,--divide -fno-stack-protector -Wl,--build-id=none  -MMD -x assembler-with-cpp -E -Isrc/include -Isrc/arch/x86/include -Ibuild -include build/config.h -I. -Isrc build/crt0.S -o build/mainboard/traverse/geos/crt0.s
>     CC         mainboard/traverse/geos/crt0.romstage.o
> gcc -m32 -Wa,--divide -fno-stack-protector -Wl,--build-id=none  -MMD -Ibuild -Wa,-acdlns -c -o build/mainboard/traverse/geos/crt0.romstage.o build/mainboard/traverse/geos/crt0.s  > build/mainboard/traverse/geos//crt0.disasm
> build/mainboard/traverse/geos/crt0.s: Assembler messages:
> build/mainboard/traverse/geos/crt0.s:444: Error: junk at end of line, first unrecognized character is `.'
> build/mainboard/traverse/geos/crt0.s:4287: Error: junk at end of line, first unrecognized character is `.'
> make: *** [build/mainboard/traverse/geos/crt0.romstage.o] Error 1
> [philipp at builder ~/coreboot]$ sed -n -e 444p -e 4287p build/mainboard/traverse/geos/crt0.s
>  .section .section .rom.text.unlikely,"ax", at progbits
>  .section .section .rom.text.startup,"ax", at progbits
> [philipp at builder ~/coreboot]$
> 
> 
> which means that we're seeing:
> 
>  .section .text.unlikely,"ax", at progbits
>  .section .test.startup,"ax", at progbits
> 
> in the input stream before transformation.
> 

Forgot to add... My toolchain is:

[philipp at builder ~/coreboot]$ rpm -q binutils gcc
binutils-2.21.51.0.6-2.fc15.x86_64
gcc-4.6.0-9.fc15.x86_64
[philipp at builder ~/coreboot]$

and the problem goes away with crossgcc.

>From what I can tell, the sed scripting just needs to be a little more selective about what it matches.