Re: Has anyone...

[Matthew Delco <delco@bmrc.berkeley.edu>]

I haven't tried compiling on .NET; I don't own a copy.  Although it would
be nice to have a compiler that supports basic Pentium Pro (and later)  
assembly instructions like "cmov" (conditional move), I've been hesitant
to buy a copy and let MSFT think that anyone is actually interested in
using .NET.  Has anyone tested if mash gets noticably faster with the .NET
compiler (perhaps by using /arch:SSE, /G6 and/or /G7)?

Anyway, the May 10 tarball compiles for me now, thanks.  Attached is a
copy of what I've been working on this weekend (originally in 5.2 as
net-ip.cc, and now ported [correctly?] to udp.cc in the current code).  
The proposed change is for Windows, and is an attempt to make a smarter
substitute for INADDR_ANY when joining multicast groups.  This issue was
discussed briefly on this list on Sept 5/6, 2002.  Ivan might like the
change, although I don't know if anyone else will appreciate it.

MD

> Matt, Ivan -
> 
> i think that your problem was a combination of the changes for Visual
> Studio .NET and the small error wei-tsang made.  i know i had to do a
> lot of work on the makefile to compile the h.263 codec.  some of it was
> cosmetic and some was strange (e.g., it appeared that sometimes they
> needed "\" file separators and sometimes "/" was ok - so converted it
> all to "\").
> 
> my question about 64-bit processors got answered -- andrew and peter
> said that the code should compile normally w/o the 64-bit define set. 
> it did and worked fine.
> 
> in answer to ivan's questions about the makefiles for otcl and tclcl,
> you're right they are probably not done correctly.  otcl was written by
> david weatherall when he was a graduate student at MIT.  steve mccanne
> included it into the mash world without many changes.  as far as i know,
> no one has touched the code much.
> 
> tclcl is slightly different.  i'm not sure who wrote it, but the ISI
> folks have taken responsbility for support so we just include what they
> have.  i have no idea if they are doing ports to .NET compilers.
> 	larry
> 
> Matthew Delco wrote:
> > 
> > > Has anyone compiled Open Mash on a 64-bit processor using the Windows
> > > build?  If so, did you have to make any changes to the makefile? Source
> > > code?
> > >       Larry
> > 
> > Does Open Mash even compile on Windows using a 32-bit processor?  The 5.2
> > tarball compiles fine, but the May 9th tarball has several problems.
> > Running build (using Visual Studio 6 SP5) I get:
> > 
> > otcl.c(424) : warning C4101: 'obj' : unreferenced local variable
> > 
> > followed eventually by the fatal error:
> > 
> > makefile.vc(170) : fatal error U1052: file 'C:\Program Files\Microsoft
> > Visual Studio\VC98\platformsdk\include\win32.mak' not found
> > 
> > Anyway, in makefile.vc I changed the non-standard (from .NET?)
> > 
> > !INCLUDE <$(TOOLS32)\platformsdk\include\win32.mak>
> > 
> > back to
> > 
> > !INCLUDE <$(TOOLS32)\include\win32.mak>
> > 
> > I restarted the build but that too resulted in an error
> > (..\tclcl\tcl2C++.exe missing).  I restarted the build from scratch, and
> > eventually blocked on
> > 
> > mid\vidreps.h(50) : fatal error C1083: Cannot open include file: 'tclDvmBasic.h'
> > : No such file or directory
> > 
> > This include file isn't anywhere in either tar, so I give up.  Regarding
> > 64-bit, there are some crosscompilers (e.g., the XP SP1 DDK has a IA64
> > compiler that runs on 32-bit systems) but I haven't tried to compile
> > application code with them.
> > 
> > MD
> 
>

udp.zip