Thanks for looking into that. Some thoughts. This module will have problems with the gnulib model, since it "feels" like a *.in.h module, and yet it can't be written to support arbitrary compilers like the other *.in.h modules can. So it won't be suitable for applications that need to install headers in arbitrary locations (as stdint.in.h is).
Some compilers other than GCC support __aligned__ and __alignof__. Oracle Solaris Studio is one that I know about. I'd feel more comfortable if the module checked for an existing stdalign.h and did not replace one that already exists. This will be more likely to do the right thing for software distributions that we are generating now but that will be compiled on non-GCC platforms in 2012. Given the above, it sounds like it'd be better for 'configure' to test for the presence of the __aligned__ and __alignof__ keywords. That is what GCC's configure.in does, for __aligned__, and that way I could modify Emacs to use the new stdalign module. That would simplify gnulib's maintenance task too, as we wouldn't need to keep track of which compilers support these keywords. Should stdalign use the existing alignof module rather than copying its code, a bit differently? Any namespace pollution issues should be easily resolvable, I'd think. Would you mind changing the list of maintainers to "all", or at least to include Jim and me?
