Re: [bug-gnu-libiconv] Including libiconv in another package

[Bruno Haible]

Hi Reuben,

> >   Also, it is very unusual/strange that the *configuration* of recode
> >   triggers a *build* of libiconv.
> 
> 
> I already fixed that: libiconv is now built by "make" as you advised, and
> is only configured by the top-level configure.

Ah, good. I was confused by the lines

  === configuring in libcharset (/d/a/recode/recode/libiconv/libcharset)
  configure: WARNING: no configuration information is in libcharset

immediately followed by

  Making check in libiconv

> I do still have a call to AM_ICONV (I presume, via AM_GNU_GETTEXT). Why
> would that cause a problem only on Windows?

We cannot do efficient communication this way, if all you give me is a
certain git branch in the 'recode' repository and a CI's log file.

What I'm missing is the tarball that was being configured in your CI.
- If there is a mismatch between the contents of the git branch and
  that tarball, we can look at that.
- If the CI's processing is surprising, we can look at what's in the
  tarball at the particular place.
- Additionally, I could run the tarball on one of the machine's I have
  access to, to reproduce and investigate.

Bruno