Re: bug in readutmp module

[Bruno Haible]

Jim Meyering wrote:
> However, I'm reluctant to remove the AC_REQUIRE, since
> that would make the code+.m4 combination depend silently on
> having a particular implementation of free.

Yes, I understand. By looking at the source code, it's not immediately
clear which free() variant is meant.

> Thinking about it some more, it seems backwards to move the
> dependency information from the .m4 file to the module file.
> I think of the .m4 file as recording dependencies inherent in the
> corresponding source files.

The benefit of the module dependency is that it is immune against changes
in the interface of the other module. For example, when you rename
jm_SSIZE_T to gl_TYPE_SSIZE_T, or add other lines in the "configure.ac"
section of the module, the higher-level module doesn't need to be changed.

The drawback is obviously that if you have several source files gathered
in a module, you lose track which file needs what, if you don't have a
more fine-grained way of tracking the dependency. You could choose to
track them through comments. Or in the .m4 file. Either is fine with me.

> How about if we leave the now-redundant
> AC_REQUIRE in place for now?

Done. I committed only the other part.

Bruno