Ian Piumarta <ian.piumarta at inria.fr> wrote:
> Folks,
> 
> On the offchance that there's anybody out there who still cares about
> my VM, I've put an updated 3.1beta in the usual place (source and
> binaries):
> 
>   http://www-sor.inria.fr/~piumarta/squeak
> 
> Since 3.1beta-4478pre1:


Awesome!  Downloading now....


"Stephen Pair" <spair at advantive.com> wrote:
> I am also insterested.  But, could someone who knows about these VMs
> post something describing the differences between Ian's VM and the
> Sourceforge VM (or any other VMs floating about)?  Also, if you could
> just tell me which one is better, that would be helpful.  If neither one
> has a clear advantage over the other, maybe we should start another
> SourceForge site to make a third VM that combines the best aspects of
> the other two VMs.
> 

Let's see, at a quick glance,  SF has:

 - VMMaker layout
 - current WWW browser plugin
 - current files for building RedHat and Debian packages
 - improved image printing
 - better handling of backspace and delete
 - freeing shared memory even on an abnoral exit
 - Ctrl-Break triggers an interrupt
 - restarting readdir() on EINTR (important over NFS!)
 - makefiles that are friendly to BSD-make 
 - cleaned up window-resize handling, which avoids some crashes
 - tweaked window-creation code that is necessary on some Solaris boxes
 - the cut-and-paste tinkering that was discussed a little while ago
 - the use of -1 instead of 0 for some invalid descriptors
 - some fixes for dealing with the case that the X server is not
connected
 - newer NAS support (recording is in there now)
 - less complaining when there is a directory with the name of an
external plugin


Ian's has:

 - a from-scratch aio module.  I don't know the differences from a first
glance.
 - -swapmod and -swapbtn
 - -memory-mmap as a a run-time configuration option, so you no longer
have to choose at compile time
 - SIGHUP triggers an interrupt
 - more consistent indentation style  :)
 - probably various things I am overlooking


This list is smaller than it's been in a while, and most of the stuff
should be easy to integrate (like the different versions of aio).  Some
parts are hard, unfortunately (like anything to do with sqXWindow.c),
and probably some fixes will be lost in the integration.

Ian's is probably the one to grab right now, unless one of the various
things above affects you directly.



Bert Freudenberg <bert at isg.cs.uni-magdeburg.de> wrote:
> >  - modifier mappings are saner (on Apple keyboards they should now be
> >    identical to MacOS; for non-Apple keyboards there's a switch
> >    `-swapmod' to make life easier in some cases)
> 
> Alt-. does not work. Alt-anything does not work. Oh, it works with
> -swapmod.  I think the default should be different for non-Apple
> platforms.  ("Help..." -> "Command Key Help" states "use with Cmd key on
> Mac and Alt key on other platforms")

"command" now seems to be mod-2.   Why did the BSD people have to pick
option as mod-1 intead of the apple key?  What a mess!    

Maybe the default should be configurable?  I hate to add the complexity,
but this will lhave a big out-of-the-box impact.  People should not have
to learn an obscure option or play with xmodmap in order to get a basic
key working.  On  my machine it's especially funny: num-lock turns on
mod-2, and since I had numlock turned on, every mouse click brought up
halos!  Weird!  Anyway, if we do make the default configurable, then
-swapmod should probably be renamed, like, -applemod vs. -noapplemod, or
-commandIs1 versus -commandIs2.  That way the affect of the flag doesn't
depend on the compilation options.

Alternatively, maybe we could just dump the option key?!  The option key
is not portable, anyway, and it should surely be deprecated for Squeak
code.  Maybe both mod-1 and mod-2 could generate "command" on Unix?


Lex