On 08/03/10 11:44, Somebody in the thread at some point said:
Unfortunately there's no simple way of working around that by glueing that option into the defaults I could find.
That seems like a bad idea in any case -- then people would start inadvertently linking against x86 object files and wondering what went wrong.
Yeah, I'm just thinking about getting through this but you're right at the distro level it'd be a pretty bad workaround.
I studied your 4.1 libgcc.a contents as well, there isn't any thumb related routine in that old version but nothing in that guy reports "no machine".
Right, I didn't run into this problem myself at any point with the 4.1 stuff.
Hum if I go down into the rpmbuild/BUILD where it was made, I can see the original .o and it's fine
[agreen@otae libgcc]$ armv5tel-redhat-linux-gnueabi-objdump -f _thumb1_case_sqi.o
_thumb1_case_sqi.o: file format elf32-littlearm architecture: arm, flags 0x00000011: HAS_RELOC, HAS_SYMS start address 0x00000000
The one that is extracted from the .a file -- surely it is from the above file though -- is broken
[agreen@otae tmp]$ armv5tel-redhat-linux-gnueabi-objdump -f _thumb1_case_sqi.o
_thumb1_case_sqi.o: file format elf32-little architecture: UNKNOWN!, flags 0x00000010: HAS_SYMS start address 0x00000000
-Andy