Re: gobject-introspection and MinGW

[Earnie <earnie users sourceforge net>]

Earnie wrote:
> Earnie wrote:
>> Earnie wrote:
>>> I'm building gobject-introspection-1.30.0 with MSYS and MinGW.
>>> I'm to the point of the GISCAN GLib-2.0.gir and it reports "No
>>> mscvr71.dll loaded." I'm using the python.org provided Windows
>>> distribution version 2.7. I had to modify the hack in
>>> giscanner/utils.py at line 101 to replace('.libs\\.libs',
>>> '.libs') to get this far, as well as set environment variables.
>>> Any hints on how to get past the reported error?
>>
>> Ok, so with the help of http://tml.pp.fi/g-i.windows.diff I got
>> past the "No mscvr71.dll loaded" issue by loading mscvr90.dll
>> instead. But now we're aborting on the syntax of stdio.h.
>>
>
> So thinking more about how to get the proper MSVC runtime I was
> thinking we should be able to have the configure script define it for
> us. We can ``strings $PYTHON | grep -i msvcr'' to obtain the runtime
> module which can be used to define a macro in config.h.
>
> I'm still looking for a resolution to the scanner syntax issue.

Trying to debug this I'm finding several issues with the call to the
compiler and linker.  First, for some reason utils.get_libtool_command
isn't returning None.  So at the beginning of the command I add a
condition for os.name == 'nt' and return None.  This led me to the fact
that the command argument ordering is out-of-sync with any OS but those
using ELF libraries.  For MinGW the list of objects and libraries must
be specified in order for resolution of references.  This means the
GLib-2.0.o object file should be listed before the required libraries. 
Also there is a mention for -llibglib-2.0-0 and libgobject-2.0-0 which
do not exist; why are these added?  I'll try to figure out how to move
the object file and remove the unneeded libraries.

Any speculation here can help.

Earnie