Re: [Chicken-users] 2.207 tarball doesn't build on MinGW or VC++

[felix winkelmann]

On 11/3/05, Brandon J. Van Every <address@hidden> wrote:
>
> I notice that the ChangeLog only goes through Sept. 24th.  Perhaps it is
> not a priority for a tarball.  However, if ChangeLogs are generated by a
> script, then perhaps it is an error, so I mention it.

Absolutely right. Noted.

>
> I notice that README.darcs is not present in the tarball.  Is that
> because it's being revised?  If it's to keep people from getting
> confused, then I think it would be better to include it.  For instance,
> the main reason I need a newer tarball is to bootstrap from darcs.

The darcs README is intended for darcs users, after a first checkout.
I don't think it is necessary to include it in the distrubution.

>
> In my VC++ full retail environment, vcbuild.bat has many type errors in
> pip.h.  This is related somehow to WS2tcpip.h.
>

Oh boy. What header is now the required one? winsock2.h doesn't
seem to be enough, judging from earlier reports by several people.

Do I need to include both headers? Can someone try to build the
system with #include <winsock2.h> added before the include of
ws2tcpip.h, please?

> In my MinGW environment, ./configure  still does the following odd path:
> checking for
> d:/lang/MinGW/bin/../lib/gcc/mingw32/3.4.4/../../../../mingw32/bin/ld.exe/libws2_32.a...
> no
> As a directory, ld.exe is weird.  Previously I've reported this as
> likely a sed problem, but I didn't provide a patch.
> It probably causes problems later in the build.
> The build dies in posix.c with many errors; looks like it's not building
> posixwin.c as it should.

Yes, that's the wrong one. I'm unable to fix this, since I don't have access
to a Windows machine and am fed up with autoconf enough already.

>
> I notice that CMakeLists.txt is not present in the tarball.  I've tried
> adding it from the latest darcs and running cmake as per below.

Sorry, my mistake.

>
> In my MinGW environment, when I do 'cmake -G "Unix Makefiles"',
> makefiles are generated without error. However, 'make' dies looking for
> -llibchicken.  It seems the dependency to build libchicken is broken.  I
> can issue 'make chicken_lib', and this looks like it may almost be
> working, but then it dies.  'nologo' could be some kind of a sed error;
> maybe it's cmake's fault.

Looks like CMake's mingw support isn't so good after all...

> In my VC++ retail environment, If I do 'cmake -G "NMake Makefiles"',,
> makefiles are generated without error.  However, 'nmake' dies looking
> for libchicken.lib.  As before, the dependency to build libchicken is
> broken.  I can issue 'make chicken_lib', but this dies with many
> undefined types in pip.h.  It seems to be related to WS2tcpip.h.

See above for the header file. Can someone look into that library
dependency issue?

>
> I have no solutions at this time.  Must go gather signatures; shoulda
> done so already.  Consider all of this a field report and food for thought.
>

Thanks, Brandon. Trying all this out and messing with build-environments
and options is no fun. Your help is much appreciated.


cheers,
felix