RE: Odd behaviour when building on Win32 using MinGW

["Espen Harlinn" <espen harlinn seamos no>]

Hi,

> Hmm, if you can use paths like /f/MinGW then you are using
> the Cygwin gcc (and linker). I am not saying that it doesn't 
> work as well as the mingw one (it's supposed to, as you use 
> -mno-cygwin), it's just that I haven't tried...

No I'm using MinGW, MSYS and related tools. I've just edited my
%MSYS_ROOT%\1.0\etc\fstab file so /f is really f:/

I started out using the pre-built binaries and dev packages from your
site, and they seems to work fairly OK except that building other
packages on top of them is rather painful - mostly because of old
libtool versions - no __imp_ prefixed exports in your packages. Since
I'm also rather fond of Borland C++ builder, I would like to add a set
of _ prefixed exports too. The resulting binaries should be usable with
M$ vc, gcc(mingw) and Borland C++ builder.

I would also like to be able to debug the libraries, and I would really
like to be able replace (read fix) some of the code - most versions of
relocatable.c does not include windows.h - and that means that it can't
do what it's supposed to do. (not really a glib problem - but since glib
depends on libintl & libiconv ...)

IMHO: In an ideal world glib for windows should include a set of
functions returning the paths to the various directories usually
specified in the makefiles on unix. Theese functions should return
directories based on the location of the glib dll. On unix they can
return whatever was specified in the makefile. This would be portable,
and it would ease building and installing the libraries correctly on
win32. I would be happy to contribute the win32 part of this code,
porting it to unix should be a set of one liners.

Initially I tried to build things using the procedure from
glib-2.2.2\README.win32 and I did remember to put the path to
/f/MinGW/bin in front of everything else inside /etc/profile. As I
mentioned in my posting this didn't work very well - configure found a
lot of cygwin stuff and decided to create makefiles that depended on
cygwin.

So now I'm trying to use MinGW, MSYS and related tools.

I'm currently not using the code from the cvs, since I thought that
things would be better tested if I used the release archives.




> FAIL: module-test

This seems to be a bug in the test script, not gmodule.

This is from module-test:
if test -f "$progdir/$program"; then
    # Add our own library path to PATH
 
PATH="$thisdir/../gmodule/.libs:/f/Src/MinGW/Source/glib-2.2.2/glib/.lib
s:$PATH"

    # Some systems cannot cope with colon-terminated PATH
    # The second colon is a workaround for a bug in BeOS R4 sed
    PATH=`$echo "X$PATH" | $Xsed -e 's/::*$//'`

    export PATH  <--- THIS IS WHERE THE PATH GETS EXPORTED

    # Add the dll search path components to the executable PATH
 
PATH=:/f/MinGW/lib:/f/Src/MinGW/Source/glib-2.2.2/gmodule/.libs:/f/Src/M
inGW/Source/glib-2.2.2/glib/.libs:$PATH  <--- THIS IS WHERE IT GETS
MODIFIED

    if test "$libtool_execute_magic" != "%%%MAGIC variable%%%"; then
      # Run the actual program with our arguments.

      exec $progdir/$program ${1+"$@"}

      $echo "$0: cannot exec $program ${1+"$@"}"
      exit 1
    fi

My guess is that the path should be exported after the last
modification.

Best regards
Espen Harlinn