RE: Win 98 Compiler OK

[Mark Butcher <M_J_BUTCHER at compuserve dot com>]

Hi Wayne

>I get confused because you mentioned compiling with MingW
>and DOS, not Cygwin.  This concerns me because you need
>Cygwin's 2.95.3 version of gcc working to bootstrap the
>version of 3.x you want.

I am using MinGW as compiler. gcc -v gives me "gcc version 2.95.3-6 (mingw
special)" on both Win 98 and Win NT computer under dos and gygwin. The
tried compiling and running my test file under dos (Win 98) and it was
successful. When I repeat the exercise under cygwin the result is identical
and also successful. I am sorry if I confused you with only the results of
the test under dos. Until now I have never found any difference when
executing programs under dos or cygwin.

The reason I chose MinGW is because it has been recommended in at least one
recent mailing as suitable for building gcc (although they say it is a
minimum version it is pretty big !). I did search around for others but
didn't find anything which did not require half a dozen tools to go with it
to make it work. I was expecting to find a binary for a native compiler at
cygwin bun couldn't actually find one. I didn't know that it can be a part
of the cygwin installation - I was of the opinion that I got a complete
installation because it seemed as though all parts were activated. Perhaps
I should repeat the cygwin installation and go through checking every
possible option to be sure.

>By the way, for what reason are you wanting to upgrade
>to gcc-3.0.4?
I assumed it best to use the very latest version and didn't want to start
any cross compiling until I know that everything works well with native
compiling. If you think that this is a bad idee I can start again with
older (established) versions.


>By the way, when I was first on this road, I ran into someone
>who was reporting successful builds of gcc-3.0x based on the
>fact that gcc -v came back with stuff that more or less made
>sense, he never compiled anything.  He was honest enough to
>try when I asked him, and lo, it did not work.

This is interesting. I went into the office today and repeated the compiler
test on the Win NT computer. The gcc -v worked fine but it really can not
compile a simple program. The error on the NT computer is therefore really
a compiler problem (it says installation problem), although the
installation was performed identically with identical roots, etc. (or so I
think)

So I deleted everything and put the mingw.tar in a directory mingw (before
it was MinGW). I extracted using tar -xvf and then it worked (compiling and
running  Hallo World under cygwin) ! Not sure whether the capitals were the
problem or whether there had really been an installation error. But it
works now and that is the important point.


Now the attempted configuration aborts exactly as on the Win 98 computer
(Syntax error while checking whether make sets $(MAKE). This means that at
least the error is stable.


These are my thought:
1. I could go after the error message. Perhaps there is really a syntax
error somewhere which could be corrected.
2. Perhaps the MinGW compiler is the cause and I could try another one.
3. Perhaps cygwin has a problem. I could try an alternative, someone
suggested MSYS, which is part of the MinGW project.
4. Possibly I could reinstall cygwin again, with a built in compiler.

What would you suggest ?


I have managed to find someone who has a little experience with Unix (at a
software company in the vicinity) and he did some quick tests concerning
the error message which I receive. As we understand it, the configuration
has created some make files for the build and wants to check their
validity. There is a make file called conftestmake in objdir/gcc. It has a
simple echo test:
all:
         @echo 'ac_maketemp="$(MAKE)"'

and we are assuming that this is being used to check that the make utility
is reponding correctly and is possibly what is not working properly. We
then tried running the conftestmake with 'make -f conftestmake' and it
answers (from cygwin shell) with 

  ac_maketemp=\c:/MinGW/BIN/make.exe

telling us that the make file is working and that the variable is ok. We
then wonder why, if it works, should the checker abort saying that there is
a syntax problem(?)


One idea we had is that it may be the German version of the Windows
operating systems so I borrowed an NT computer with the English version of
the operating system and repeated the installations (exactly - directory
structures, c-tests etc.) and the resuilt was exactly the same - the syntax
error ! I wonder ??


Cheers

Mark