Re: portability checks, errors and warnings

[Bruno Haible]

Paul Eggert wrote on 2007-02-05:
> >      he will get a link error about function 'rpl_strstr', and will likely
> >      report it as a bug in gnulib.
> 
> I don't think this will be much of a problem in practice.  We can
> assume maintainers will read the gnulib documentation so that they
> know what rpl_ is for.  Briefly, defining strstr to rpl_strstr means
> "Do not use any library implementation of strstr, regardless of
> whether the library implements strstr."

A linker warning with a one or two line text is better than a link error
because

  - We can enable a link warning even for cases where 90% of the uses of the
    functions are ok, such as
      dlopen()
        If the file name argument is not absolute, the file is searched for.
        The search algorithm is system specific.
      glob()
        Some systems may store additional flags in the @code{gl_flags} field.

  - A link errors aborts the "make" process, a link warning doesn't.

  - The documentation (functions.texi) is not yet in the autoconf manual.

  - The documentation (functions.texi) does not say which gnulib module
    solves the portability warning; the rpl_* symbol doesn't say either.
    But the link warning says it.

Bruno