Date: Thu, 3 Aug 2000 08:55:03 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: Matt Lewandowsky <matt DOT l AT techie DOT com>
cc: djgpp AT delorie DOT com
Subject: Re: Uh oh. Another newbie. (Sorta...)
In-Reply-To: <4.3.2.7.0.20000802130606.00aa5270@mail.subdimension.com>
Message-ID: <Pine.SUN.3.91.1000803085443.10519C-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: djgpp AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk


On Wed, 2 Aug 2000, Matt Lewandowsky wrote:

> I can't get any configure scripts to recognise PDCurses (although 
> I installed 2.2, built 2.4, and then installed the prebuilt 2.4), nor can I 
> figure out how to get configure to recognise wattcp-32.

I don't see any sign of these two failures in the trace of the
configure run you posted.  Where do you see the test for curses and
network libraries in what the configure script does?

In any case, the configure script leaves a file config.log which
describes each failed test, including the source of the test programs
it tries to compile and run.  Look there for the clues why doesn't the
script work as you think it should.

> For example, this is the output from trying to compile pdksh under 9x. I 
> can't figure out quite why it's failing

It fails because the sources seem to be broken for DJGPP:

> jobs.c: In function `j_init':
> jobs.c:237: `SIGCLD' undeclared (first use in this function)
> jobs.c:237: (Each undeclared identifier is reported only once
> jobs.c:237: for each function it appears in.)

My guess is that the program uses SIGCLD without testing whether it is
defined.  DJGPP doesn't support the SIGCLD signal, so signal.h doesn't
define it.

The usual way to do this portably is like this:

    #ifdef SIGCLD
    ... code that uses SIGCLD ...
    #endif

However, since this program seems to be a shell, it is possible that
without SIGCLD some major features (like job control, as the name of
the source file suggests) will not work.  It's even possible that
other, unrelated features will break.  I suggest to browse through the
sources and see what else depends on SIGCLD.

> but under "pure" DOS (my usual environ), it won't even get through
> ./configure.

Probably due to long file-name problems with files like config.h.in.

> (BTW, for the configure option specified, mmap() and flock() are
> required. I have no clue if they are implemented in DJGPP...)

The configure script detected, correctly, that they are not present in
the library.  I would expect a package that tries to find them at
configure time to be able to provide a fallback.  If you want to see
what the package will do without mmap and flock, grep the sources for
HAVE_MMAP and HAVE_FLOCK.

> sh-2.03$ autoconf
> configure.in:257: warning: AC_TRY_RUN called without default to allow cross 
> compiling

I'm not even sure why did you need to run Autoconf.  I would simply
run the configure script supplied with the package.

Note that I know nothing at all about this particular package; all I
said above is from general knowledge of the configure machinery.