Re: Strange LT_INIT behavior

[Ralf Wildenhues]

Hi Peter,

* Peter O'Gorman wrote on Mon, May 25, 2009 at 07:05:42PM CEST:
> 
> This is the "Ok to commit?" message. Jeff is already in THANKS. Ralf, if
> you want time to review, that's fine.

> >From 012c3c35cfd7c15542f70b360269d9f27c5aafb5 Mon Sep 17 00:00:00 2001
> From: Peter O'Gorman <address@hidden>
> Date: Mon, 25 May 2009 10:00:42 -0500
> Subject: [PATCH] Don't run compiler checks twice.
> 
> libltdl/m4/libtool.m4 (_LT_PROG_FC, _LT_PROG_F77, _LT_PROG_CXX): Remove these 
> macros.
> libltdl/m4/lt~obsolete.m4: Add removed macros here.
> libltdl/m4/libtool.m4 (_LT_LANG_FC_CONFIG, _LT_LANG_F77_CONFIG, 
> _LT_LANG_CXX_CONFIG): Remove references to removed macros, and move 
> functionality here.
> NEWS: Update.

Please wrap lines in the commit log entry.  (I just take the ChangeLog
entry, remove the initial lines and the leading tabs, and add an empty
line after the first one.)

This patch is ok; it could be better even, if we had a test case as
well: one that has AC_PROG_{CXX,F77,FC} before LT_INIT, one with
reversed order and one without them (to ensure that no C++ or Fortran
compilers are tried out).  Not sure if it would be necessary to try
each of the compilers separately; also, some of these are probably
already covered by the old or the new testsuite, I haven't checked.

> --- a/NEWS
> +++ b/NEWS
> @@ -35,6 +35,8 @@ New in 2.2.8 2009-??-??: git version 2.2.7a, Libtool team:
>    - Argument mangling of execute mode has been improved (i.e., lessened).
>    - Fix 2.1b regression that caused nm to not be the default name lister.
>      The regression affected mainly (arguably broken) cross compiles.
> +  - Fix long standing bug that caused compiler checks for fortran and
> +    c++ compilers to run twice.

Fortran, and C++, please.

Thank you!
Ralf