One more thing: I haven't recompiled gcc, but I don't think I should because stubs are added by linker. Hope that clarifies things, Adam Adam Turowski wrote: > > I recompiled newlib and double checked what binutils version was used > during linking stage. Unfortunately still have no success. > > What I found is that almost __from_arm stubs all broken - there are zeros > in there instead of proper code. Only one stub __from_arm is OK (the call > comes from assembly file) and all __from_thumb calls looks OK in > dissassembly. > > There is another weird thing I found when I've converted from .elf to > binary using "arm-elf-objcopy -O binary" command. The resulting binary > values generated for these stubs code are 4 byte shifted towards front of > the file. For example stub code starting on 0x5EA8 address in dissassembly > file starts on 0x5EA4 > address in resulting binary. The rest binary code (I mean non stub) is in > the right places. > > Have you got any suggestions? > > Thanks > Adam > > > Miguel Angel Ajo Pelayo wrote: >> >> I think I didn't need to recompile the newlib, but try it, and check >> that your new ld/gcc is being effectively used at link time. >> >> Miguel Ángel >> > > -- View this message in context: http://www.nabble.com/gcc-4.4.0-arm-thumb-interwork-_from_arm-bad-stub-generation-tp24016313p24256768.html Sent from the gcc - Help mailing list archive at Nabble.com.
