On Fri, Jan 02, 2004 at 03:47:54PM -0500, Nathanael Nerode wrote: > From the 4.3.0 TODO Thanks a lot for doing this, Nathanael. > 001_linux_cf.diff > ----------------------- > Most of this is present upstream in 4.3.0 branch. > Essentially all of the rest is in 003_linux.cf_and_xfree86.cf.diff. > > There are some definite oddities in 003_linux.cf_and_xfree86.cf.diff, however. > I don't pretend to understand these files, but it's worth checking that the > folowing are deliberate: > > * linux.cf: GccAliasingArgs is in OptimizedCDebugFlags upstream, but dropped > in the Debian patch, for HPArchitecture, ia64Architecture, > Mc68020Architecture, MipsArchitecture, s390Architecture, SparcArchitecture, > and SuperHArchitecture. Daniel, can you explain this? > * linux.cf: -fomit-frame-pointer is in OptimizedCDebugFlags upstream, but > dropped in the Debian patch, for s390Architecture, s390xArchitecture. Daniel, can you explain this? > * In xfree86.cf, DefaultGcc2i386Opt, DefaultGcc2AxpOpt, and DefaultGcc2PpcOpt > are defined to contain GccAliasingArgs. But in linux.cf, they're defined to > *not* contain it. > > Note that GccAliasingArgs was introduced upstream sometime after 4.2. How > precisely should it be handled for Debian? Good question. I'd like to solicit suggestions from the mailing list. :) If the XFree86 codebase is strict-aliasing safe, we can do without -fno-strict-aliasing. > * The following silly little fix for unspecified architectures was in > 001_linux_cf.diff, but isn't in the new patch: > @@ -686,6 +693,7 @@ > # endif /* x86_64Architecture */ > # else > # define CcCmd gcc > +# define CplusplusCmd g++ > # define AsCmd CcCmd -c -x assembler > # define LdCmd CcCmd -nostdlib > # define AsmDefines -DUSE_GAS This should probably be added back. > * linux.cf: HPArchitecture doesn't wrap its definition of OptimizedCDebugFlags > in the usual #ifndef upstream, and the new patch doesn't either. But the old > patch did. It should do so. > I hesitate to submit patches due to my lack of understanding, but if any of > the issues are verified and solutions outlined, I could do so. > > Once these issues are fixed (or verified to be non-issues), 001_linux_cf.diff > can be marked as merged to the new 003_linux.cf_and_xfree86.cf.diff. I think it was a bad idea to merge these, but oh well. When Daniel started this work he didn't fully understand the principle of least surprise, but I think he and I have gone around enough now that he does. :) > 004_xfree86_cf.diff > ---------------------------- > Entirely present upstream. Great. -- G. Branden Robinson | Ambition: an overmastering desire Debian GNU/Linux | to be vilified by enemies while branden@debian.org | living and ridiculed by friends http://people.debian.org/~branden/ | when dead. -- Ambrose Bierce
Attachment:
signature.asc
Description: Digital signature