RE: autoconf/configure problem on text mounts

["Dave Korn" <dave dot korn at artimi dot com>]

lemkemch wrote on 10 August 2008 18:46:

> I am not familar enough with autoconf to be able to tell how that line
> gets into configure.  I can't find it in configure.in.  All I can see is
> stuff like AH_VERBATIM([HAVE_UINT64_T] and AC_CHECK_TYPES([uint8_t,
> int8_t, uint16_t, int16_t, uint32_t, int32_t, uint64_t, int64_t]); I'd
> guess it is releated to the latter.

  Yep.  Those are macros that get expanded using m4 scripts based in
/usr/share/autoconf to generate the configure script.

> The contents of conftest.val are used later like
> 
> ac_cv_sizeof_unsigned_int=`cat conftest.val`

  'cat' always opens files in binary mode, as required by POSIX.
 
> which adds a \r to the assignment causing later tests like
> 
>     if test "${ac_cv_sizeof_unsigned_int}" = "4"; then
>     ...
> 
> to fail.
> 
> Running autoconf on cygwin results in the same broken configure script.
> 
> Any ideas what should be done other than not using a text mount?  To me
> opening the conftest.val file as text is wrong.

  Well... the content is plain text, after all; the problem arises on
re-opening it implicitly in binary mode by using cat.  Then again you could
argue that since it's going to be opened by cat, it should be written in
binary.

   This is an autoconf bug upstream really.  They'd know best what how to
patch it long term.  If you wanted a quick local fix, you could edit
/usr/share/autoconf/autoconf/c.m4 and change that fopen command at source,
then any time you regenerate a configure script it'll come out ok for your
text mount.


    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/