On 27/03/2012 6:46 PM, Joseph S. Myers wrote:
Fair enough. I don't remember where I got the original patch from (years ago) -- some mailing list's archives, most likely -- but I just added the windows/cygwin conditionals following the existing patches' use of Darwin/FreeBSD conditionals. I lack the autotools-fu to create the feature test from scratch, though.On Tue, 27 Mar 2012, Ryan Johnson wrote:Hi all, At the request of the folks at crosstool-ng.org, I'm attaching three small patches for eglibc-2.12 that allowed me to bring up an x86_64-unknown-linux-gnu toolchain (gcc-4.5.3 and 4.6.1) on both cygwin-1.5 and 1.7 hosts. I suspect the patches will work for other eglibc versions as well, but I haven't tested that.I don't think conditionals on the build system OS are a good idea; the GNU practice is to use feature tests where necessary instead.
Disabling i18n would probably remove the need for the two *libintl.patch files, which fix linker errors involving -lintl; the third patch (bsd-u_char) resolved type-already-defined errors; I forget which system header contained the original definition, but it gave virtually identical typedefs to the ones in sunrpc/rpc/types.h; the patch just copies verbatim the #ifdefs that guard the original definition.I've been gradually working on merging the cross-rpcgen changes to FSF glibc; my most recent version, <http://sourceware.org/ml/libc-alpha/2012-03/msg00375.html>, disables i18n for the cross-rpcgen, which might help with avoiding libintl dependencies there (or maybe further i18n disabling is needed as well - you didn't say anything about what the actual problems are that your patch is addressing or why you chose a particular approach).
In any case, the patches are mostly intended to identify the problem and show where the solution lies; I'll leave the polishing to someone more adept at autotools hacking than I am.
Regards, Ryan _______________________________________________ Patches mailing list Patches@xxxxxxxxxx http://eglibc.org/cgi-bin/mailman/listinfo/patches