- After release of the 2.0 SDK, Geoworks released a modified version of Swat
which was updated to improve its ability to run under OS/2; these changes should
be included in the OmniGo version of the SDK, and it should also be available on
CIS. Remember that when you update SWAT.EXE on your Host, you must also update
the SWAT.EXE stub on your target machine.
- Modified settings in the (fullscreen) DOS box (not sure which of these really
affect Swat performance): [assuming that Swat is connected via serial line on
COM1]
HW_TIMER=1
IDLE_SENSITIVITY=100
DPMI_DOS_API=ENABLED
SESSION_PRIORITY=32
COM_HOLD=1
COM_SELECT=COM1
XMS_MEMORY_LIMIT=0
EMS_MEMORY_LIMIT=0
EMS_FRAME_LOCATION=NONE
AUDIO_ADAPTER_SHARING=None
- I'm running Swat with the -noTimeout parameter, because I was under the
impression that this might resolve problems occuring sometimes when trying to
connect to the Target for the first time (not sure about this either). I have
set up two DOS icons calling a batch file which sets all the environment
variables required by Swat, one calling Swat with and one without the "-n"
option (forcing connection to an EC or NC target build).
- The most important thing: despite of all this, I have never been able to run
Swat under OS/2 on a 386/40 machine (not even by replacing COM.SYS by Ray
Gwinn's SIO.SYS) - after upgrading to a P75 board, it worked right from the
start, so I suspect that there is a timing problem in the interrupt handling
somewhere, which causes reproducible DOSXNT "abends" on slower machines [btw, I
have seen Swat fail with the same offending address on Win95 as well when trying
to connect it to a target running in the background on the *same* machine, thus
also taking away CPU time].
ciao marcus
ceterum censeo: Rebuilding the Geos tools as native OS/2 applications would be a
Good Thing, because it would avoid the compatibility problems with the Dos
Extender, and it shouldn't be too hard taking into account that the memory model
of OS/2 is much closer to the one assumed under Unix than that offered by Dos.
But, then, we should be happy that there is an SDK for ISVs at all, shouldn't
we? :-)
ps: Has anyone else experienced crashes when trying to run the OmniGo simulation
under OS/2? On my setup, the session just aborts about one in three times it
tries to play a sound.