colinux-devel

[coLinux-devel] Mingw patch for CoLinux console/deamon

[Steven E. <ste...@ya...>]

Hello All,
Well here it is in all of its glory. All 5 lines of it. It works but we
(ReactOS team) have not tested it beyond booting the kernel.

There are a few changes that will need to be made before this can be
merged in and we can drop Cygwin I guess. But I beg and plead that we
do this so we can drop Cygwin totaly if we are not going to need it for
the Xserver.

1. The FLTK console wont build on the Mingw cross-compiler as we lack
support for it. I guess we could ship our own copy.

2. The user has to go download/install the mxml sources for the
mingw-crosscompiler. Once again its small we could ship this with the
coLinux sources.

3. You need to change i*86-cygwin-blah to -*86-mingw32-blah for all of
the commands in your makefile to support a mingw cross

4. We need to add some configure like checking for these headers I have
added the ifdefs for in this patch.

5. It links to ws2_32 but I have not tested the networking stuff so we
will have to adapt the rest to use use Winsock but it will still be
faster than using Cygwin->Winsock.

Thanks
Steven

__________________________________
Do you Yahoo!?
Yahoo! Small Business $15K Web Design Giveaway 
http://promotions.yahoo.com/design_giveaway/

[coLinux-devel] Re: Mingw patch for CoLinux console/deamon

[Dan A. <da...@co...>]

On Fri, Apr 09, 2004 at 02:46:20PM -0700, Steven Edwards wrote:
> Hello All,
> Well here it is in all of its glory. All 5 lines of it. It works but we
> (ReactOS team) have not tested it beyond booting the kernel.

So the kernel really booted. Can you post a screenshot?

> There are a few changes that will need to be made before this can be
> merged in and we can drop Cygwin I guess. But I beg and plead that we
> do this so we can drop Cygwin totaly if we are not going to need it for
> the Xserver.

I agree, Cygwin should be dropped.

> 1. The FLTK console wont build on the Mingw cross-compiler as we lack
> support for it. I guess we could ship our own copy.

I got it built using the Mingw cross compiler, but I couldn't test the 
library because of some linkage problems.

I've fiddled a few hours with creating a i686-pc-mingw32 cross compiler, 
and found a bunch of annoying problems with gcc's build - actual build 
system bugs. Do you have scripts that automatically create a mingw32 
cross compilation toolkit on Linux using recent versions of gcc 
and binutils? What versions do you use? I'd really benefit from it.

I also tried to use the -mno-cygwin gcc flag. It worked for the daemon, 
but then I found out that libstdc++ depends on a symbol inside cygwin1.dll! 
So I had to specially compile a new libstdc++ that doesn't refer that 
symbol (___getreent. I had to use --disable-threads and pull more defines 
from the hat to get it working).

Anyway, as from the 20040411 snapshot, the daemon and console-nt don't 
depend on Cygwin, so you can test console-nt to see if it works on
ReactOS.

> 2. The user has to go download/install the mxml sources for the
> mingw-crosscompiler. Once again its small we could ship this with the
> coLinux sources.

Or we could ship a patch for it.

> 3. You need to change i*86-cygwin-blah to -*86-mingw32-blah for all of
> the commands in your makefile to support a mingw cross

It only needs to be changed in one location.

> 4. We need to add some configure like checking for these headers I have
> added the ifdefs for in this patch.
> 
> 5. It links to ws2_32 but I have not tested the networking stuff so we
> will have to adapt the rest to use use Winsock but it will still be
> faster than using Cygwin->Winsock.

That networking code in the daemon is not that important (used for 
debugging).

-- 
Dan Aloni
da...@co...

Re: [coLinux-devel] Re: Mingw patch for CoLinux console/deamon

[Dan A. <da...@co...>]

On Sun, Apr 11, 2004 at 01:06:04AM +0300, Dan Aloni wrote:
> On Fri, Apr 09, 2004 at 02:46:20PM -0700, Steven Edwards wrote:
> > Hello All,
> > Well here it is in all of its glory. All 5 lines of it. It works but we
> > (ReactOS team) have not tested it beyond booting the kernel.
> 
> So the kernel really booted. Can you post a screenshot?
> 
> > There are a few changes that will need to be made before this can be
> > merged in and we can drop Cygwin I guess. But I beg and plead that we
> > do this so we can drop Cygwin totaly if we are not going to need it for
> > the Xserver.
> 
> I agree, Cygwin should be dropped.
> 
> > 1. The FLTK console wont build on the Mingw cross-compiler as we lack
> > support for it. I guess we could ship our own copy.
> 
> I got it built using the Mingw cross compiler, but I couldn't test the 
> library because of some linkage problems.
> 
> I've fiddled a few hours with creating a i686-pc-mingw32 cross compiler, 
> and found a bunch of annoying problems with gcc's build - actual build 
> system bugs. Do you have scripts that automatically create a mingw32 
> cross compilation toolkit on Linux using recent versions of gcc 
> and binutils? What versions do you use? I'd really benefit from it.

Okay, I managed to port the build-cross.sh script to the latest stable 
versions of the tools. The only difference is that it requires the files 
to be downloaded.

-rw-------    1 karrde   karrde   14182109 Apr 10 20:55 binutils-2.15.90-20040222-1-src.tar.gz
-rw-------    1 karrde   karrde   13268861 Apr 11 09:17 gcc-core-3.3.1-20030804-1-src.tar.gz
-rw-------    1 karrde   karrde    2723676 Apr 11 09:11 gcc-g++-3.3.1-20030804-1-src.tar.gz
-rw-------    1 karrde   karrde     308420 Apr 10 21:09 mingw-runtime-3.2.tar.gz
-rw-------    1 karrde   karrde    1336713 Apr 10 21:12 w32api-2.5.tar.gz

[10:49 pts/424,callisto source]% i686-pc-mingw32-gcc -v

Reading specs from /usr/local/mingw32/lib/gcc-lib/i686-pc-mingw32/3.3.1/specs
Configured with: /home/karrde/colinux/res/mingw32/cross/source/gcc-3.3.1-20030804-1/configure -v --prefix=/usr/local/mingw32 --target=i686-pc-mingw32 --with-headers=/usr/local/mingw32/i686-pc-mingw32/include --with-gnu-as --with-gnu-ld --without-newlib --disable-multilib
Thread model: single
gcc version 3.3.1 (mingw special 20030804-1)

I didn't try to build coLinux with it yet. There will be more to follow.

-- 
Dan Aloni
da...@co...

Re: [coLinux-devel] Re: Mingw patch for CoLinux console/deamon

[Steven E. <ste...@ya...>]

Hi Dan,

--- Dan Aloni <da...@co...> wrote:
> Okay, I managed to port the build-cross.sh script to the latest
> stable 
> versions of the tools. The only difference is that it requires the
> files 
> to be downloaded.

One thing I didnt think about was that on cygwin you can force linking
to msvcrt by passing the -mno-cygwin flag to gcc. Could you try
building a Colinux install with it? If this is the case then we might
still be able to use FLTK and keep the cygwin setup. 

Thanks
Steven


__________________________________
Do you Yahoo!?
Yahoo! Tax Center - File online by April 15th
http://taxes.yahoo.com/filing.html

Re: [coLinux-devel] Re: Mingw patch for CoLinux console/deamon

[Dan A. <da...@co...>]

On Sun, Apr 11, 2004 at 01:41:48PM -0700, Steven Edwards wrote:

> --- Dan Aloni <da...@co...> wrote:
> > Okay, I managed to port the build-cross.sh script to the latest
> > stable 
> > versions of the tools. The only difference is that it requires the
> > files 
> > to be downloaded.
> 
> One thing I didnt think about was that on cygwin you can force linking
> to msvcrt by passing the -mno-cygwin flag to gcc. Could you try
> building a Colinux install with it? If this is the case then we might
> still be able to use FLTK and keep the cygwin setup. 

No need, FLTK works with the MingW32 setup.

-- 
Dan Aloni
da...@co...

[coLinux-devel] Re: Mingw patch for CoLinux console/deamon

[Steven E. <ste...@ya...>]

Hi Dan,

--- Dan Aloni <da...@co...> wrote:
> Anyway, as from the 20040411 snapshot, the daemon and console-nt
> don't 
> depend on Cygwin, so you can test console-nt to see if it works on
> ReactOS.
> 
> > 2. The user has to go download/install the mxml sources for the
> > mingw-crosscompiler. Once again its small we could ship this with
> the
> > coLinux sources.
> 

Cool. I sent test binarys to most of the ReactOS developers that are
interested in this when I did my port to mingw. Currently we are
debugging some issues in our named pipes and service control manager
implemenation. Once this is done CoLinux should just "work". I tested
loading linux.sys standalone as a boot driver and ReactOS is able to do
this with no issue so once we have the SCM/Pipe issues worked out I
thnk we should be able to load vmlinux. I have sent a copy of this
message to ros-kernel to try and get a few more developers interested
in kicking out the SCM/Pipe issues.
 
> > 5. It links to ws2_32 but I have not tested the networking stuff so
> we
> > will have to adapt the rest to use use Winsock but it will still be
> > faster than using Cygwin->Winsock.
> 
> That networking code in the daemon is not that important (used for 
> debugging).

who needs a debugger anyway...It may still be a while for ReactOS to
really show use to the CoLinux team as our networking still has large
spots that are not connected. Its my hope that in the next few months
we can plug those holes so that any Linux user that need a Windows
driver can deploy ReactOS with CoLinux running on top.

Thanks
Steven


__________________________________
Do you Yahoo!?
Yahoo! Tax Center - File online by April 15th
http://taxes.yahoo.com/filing.html

[coLinux-devel] Re: Mingw patch for CoLinux console/deamon

[Dan A. <da...@co...>]

On Mon, Apr 12, 2004 at 07:43:24AM -0700, Steven Edwards wrote:
> Hi Dan,
> 
> --- Dan Aloni <da...@co...> wrote:
> > Anyway, as from the 20040411 snapshot, the daemon and console-nt
> > don't 
> > depend on Cygwin, so you can test console-nt to see if it works on
> > ReactOS.
> > 
> > > 2. The user has to go download/install the mxml sources for the
> > > mingw-crosscompiler. Once again its small we could ship this with
> > the
> > > coLinux sources.
> > 
> 
> Cool. I sent test binarys to most of the ReactOS developers that are
> interested in this when I did my port to mingw. Currently we are
> debugging some issues in our named pipes and service control manager
> implemenation. Once this is done CoLinux should just "work". I tested
> loading linux.sys standalone as a boot driver and ReactOS is able to do
> this with no issue so once we have the SCM/Pipe issues worked out I
> thnk we should be able to load vmlinux. I have sent a copy of this
> message to ros-kernel to try and get a few more developers interested
> in kicking out the SCM/Pipe issues.

It's nice to see that coLinux helps to bring ReactOS to completeness.

It will be good if some of the ReactOS developers, especially the kernel 
hackers, could look at the kernel driver's sources and pin-point about 
specific problems. As you may know, we have problems in the SMP area.

    http://www.colinux.org/snapshots

The winnt-dependent kernel code is under src/colinux/os/winnt/kernel,
whereas the os-independent kernel mode is under src/colinux/kernel and
src/colinux/arch.

> > > 5. It links to ws2_32 but I have not tested the networking stuff so
> > we
> > > will have to adapt the rest to use use Winsock but it will still be
> > > faster than using Cygwin->Winsock.
> > 
> > That networking code in the daemon is not that important (used for 
> > debugging).
> 
> who needs a debugger anyway...It may still be a while for ReactOS to
> really show use to the CoLinux team as our networking still has large
> spots that are not connected. Its my hope that in the next few months
> we can plug those holes so that any Linux user that need a Windows
> driver can deploy ReactOS with CoLinux running on top.

Cool. Have you managed to build coLinux using the latest snapshot and
the bunch of scripts I posted?

-- 
Dan Aloni
da...@co...