Rainer Orth wrote: > George Vasick <George.Vasick at Sun.COM> writes: > >> Highlights: >> >> - binutils 2.19 installed in /usr/gnu/bin with g-prefixed links placed >> in /usr/bin. a new package, SUNWgnu-binutils, will be created. > > Doesn't this need to go by ARC if you try to change the previous case in > this way?
Definitely. I want to get people's feedback on the proposed changes and then take the revised case back to ARC. > >> - existing binutils 2.15 will be retained in /usr/sfw. binutils 2.19 >> will not build ON without source source changes: >> >> /export/george/onnv/usr/src/tools/proto/opt/onbld/bin/i386/aw [...] -o >> debug64/i86_subr.o ../../intel/ia32/ml/i86_subr.s >> ../../intel/ia32/ml/i86_subr.s: Assembler messages: >> ../../intel/ia32/ml/i86_subr.s:1252: Error: suffix or operands invalid >> for `push' >> ../../intel/ia32/ml/i86_subr.s:1258: Error: suffix or operands invalid >> for `pop' >> >> It appears we need to keep binutils 2.15 around until the ON source can >> be ported to the newer version of the x86 assembler. Note, the ON build > > I'd very much prefer that this be fixed in O/N instead of working around > the issue: from what I've read (I know almost nothing about x86/amd64 > assembly), this is invalid assembler code on amd64, and gas correctly > errors out here. Instead of cheating with continued use of gas 2.15, fix > the issue at the source. I agree, but I think we need to do this in two steps: 1) get the new binutils, and gcc 4.3 for that matter, out there so people can migrate their code. 2) remove the old binutils. Once everything is migrated to the new binutils, the idea is to simply upgrade to future versions as they become available without trying to keep multiple versions of binutils around. Thanks, George > >> succeeds on Sparc using binutils 2.19. The compatibility problem is >> limited to x86. > > Obviously: gcc 3.4.3 on SPARC uses Sun as. > > Rainer >
