Paul Eggert wrote:
> > https://www.gnu.org/software/gnulib/manual/html_node/Modules-that-modify-the-way-other-modules-work.html
> >
> > How would you answer the question
> > “If my module occurs among the modules of gltests, should it
> > have an effect on the modules in gllib?”
> > in this case?
>
> I assume the new module should be like malloc-gnu, so the answer should
> be "no". Is there a subtle point I'm missing?
Hmm. Let's look at the two main scenarios.
* A gnulib testdir. Suppose the 'alloc-0-nonnull' module is not present
in gllib, but is present among the dependencies of a module T among the
gltests. This module T will call *alloc(), expecting 0-nonnull semantics.
If it's done as carefully as with the two different AC_LIBOBJs in
modules/malloc-posix and modules/malloc-gnu, there will be different
realloc functions in gltests/libtests.a and in gllib/libgnu.a, and
linking any of the tests that uses module T will use the one from
gltests/libtests.a. So we're fine.
* A package with two uses of gnulib-tool, for example, libpoke and poke.
The same mechanism will work there.
It seems that the documentation section that I pointed to is out-of-date
for these scenarios. The correct pattern to use is the one introduced by
https://lists.gnu.org/archive/html/bug-gnulib/2022-01/msg00020.html .
Bruno
