Sorry, somehow I thought those patches were for glibc review, not gnulib
review.
The gnulib part is patch 5/5. I see a few problems there:
These days _GL_ATTRIBUTE_PURE is defined by m4/gnulib-common.m4 and so
there should be no need to include stdwhatever.h to define it if one is
using gnulib.
Similarly, to break the dependency on modules/stdlib, modules/obstack
should depend on and use snippet/_Noreturn, as modules/stdlib does.
Third, you need to modify modules/obstack to remove the dependencies
you've removed.
Fourth, suppose a project uses both obstack and exitfail? Then its
exit-failure isn't propagated. This suggests that obstack should
conditionally use exitfail if present.
I haven't looked at the other 4 patches carefully but assume they're OK
for glibc (or your glibc reviewers will tell you :-).
