Message-ID: <B0000064264@stargate.astr.lu.lv>
From: "Andris Pavenis" <pavenis AT lanet DOT lv>
To: Frank Donahoe <fdonahoe AT wilkes1 DOT wilkes DOT edu>, djgpp AT delorie DOT com
Date: Mon, 18 Jan 1999 21:48:46 +0200
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: Will gcc 2.8.1 compile itself under DJGPP 2.02?
In-reply-to: <Pine.A32.3.96.990117192828.48701A-100000@wilkes1.wilkes.edu>
X-mailer: Pegasus Mail for Win32 (v3.01d)
Reply-To: djgpp AT delorie DOT com

On 18 Jan 99, at 12:22, Frank Donahoe wrote:

> With a new pentium II machine and nothing much except Windows 98
> loaded it seemed a good time to try something for which I would
> never find enough memory or time on my old machine.
> 
> So gcc281s.zip was loaded, (Oct 22, 1998) along with the binaries
> and associated GNU binaries, (even acnf212b.zip when a first run
> reported it couldn't find autoconfig), and djdev202.zip.

autoconf should not be needed unless You edit configuration related
files (e.g. configure.in)
> 
> As per instructions, from %DJDIR%/gnu/gcc-2.81/build.djg the
> command "sh djmake.sh" started the build process.  This ran into
> trouble about six minutes into the build where it was working
> on `libgcc2.a'.  The error was different on each run.  From a
> full DOS screen, windows snapped it shut so fast that nothing
> could be seen.  Once the error reported was "Load error:  no DOS
> memory".  Windows accepted 600K as the largest block of DOS
> memory it would reserve for the DOS prompt.  Working from a
> window rather than full screen, the message:

If You see similar things close DOS session and start a new one.

> "This program has performed an illegal operation and will be
> terminated.  Exit all programs and then restart your computer."
> 
> additional information:
> 
> invalid page exception
> Fault location:  0028:C00025E6
> Faulting component:  VMM[01] +000015E6
> Interrupts in service: None
> 
> Only once did I notice any error messages in the DOS window,
> but that was lost when, in trying to move the windows error
> box out of the way, it was accidentally closed.
> 
> Finally the build got past that point.  It happened while
> I was no longer watching the screen.  Nothing else had been
> changed from the previous run which failed.
> 
> But the script now went down a blind alley.  Following the
> conditional at the target "f77-runtime:" (Makefile: line 1626)
> it produced:
> 
> entering directory "../gcc-2.81/build.djg/f/runtime"
> 
>      "..No rule to make target all"
> 
> which is quite true since that directory contains nothing but
> three subdirectories.  To find any action, it would have
> to enter the directory:
> 
>     "../gcc-2.81/f/runtime"
> 
> where a Makefile awaits.
> 
> Has anyone successfully built gcc 2.8.1 with itself under
> DJGPP V2?  I'd appreciate any help.

Of course I had build it even more than once (last time in October).
But not under Win98. But I don't more use gcc-2.8.1 anymore (only egcs-1.1.1)
> 
> Incidentally, removing `g77.exe' as a target in the top
> level of the Makefile (all.internal:) does not stop fortran
> functions from being compiled.

Correct way how to remove Fortran support is:
	- remove directory  f  (from sources, not in build.djg)
	- run ./djconfig.sh 
> 
> Multiple warnings are issued about "size_t" (and wchar_t)
> being defined in <stdio.h> and redefined in <atddef.h>
> and "va_list" defined in <stdio.h> and redefined in <stdarg.h>.

It's know thing.   The problem is with  fixinc.djgpp.  However You can safely
ignore these warnings.

> All three files are included in "djdev202.zip".  Is the
> compiler being too pickey or should these files be protected
> from eachother?

Ignore these messages.
> 
Andris