Re: New Erlang-related macro

[Romain Lenglet]

Paul Eggert wrote:
> Romain Lenglet <address@hidden> writes:
> > Here is a small patch that adds a new Erlang-related macro:
> > AC_ERLANG_SUBST_ERTS_VERSION.
>
> This goes against the usual Autoconf philosophy, which is to
> test for the features you need, rather than to try to guess
> features from version numbers.

I agree.
I also agree that I should add macros to test for the presence of 
Erlang modules, and of specific Erlang functions in modules. I 
am working on this.

But the AC_ERLANG_SUBST_ERTS_VERSION macro will not be used to 
test for features.

> Why is this macro useful in Autoconf scripts?  Perhaps there's
> something more Autoconfish that could have the same utility.

In Erlang application "release files", the ERTS version number 
must be given explicitly. In the general case, applications 
really don't depend on a particular version, but a precise 
version number is still required in every release file.
I added that macro to allow substituting that version number in 
release files automatically.

That was the motivation also for the substitution of 
ERLANG_LIB_VER_<lib> variables in macro 
AC_ERLANG_CHECK_LIB(lib): in a release files, the precise 
versions of libraries to be distributed with the Erlang 
application must be given. With those variables, one can 
substitute them automatically.

Does that convince you?

When I will submit a patch for the testing of features (modules 
and functions), I propose to also add a word about the purpose 
of those variables, in the doc, to discourage their use for 
testing features.

Regards,

-- 
Romain LENGLET