Currently only OS/2 is supported.
Start Makefile.PL like this:

perl Makefile.PL TOOLKIT=F:/TOOLKIT

(we need to know the path to the toolkit header files and libraries).

Now Makefile.PL has some code to determine the location automatically.

To run t/animal.t and t/wrong_method.t: put animals.dll (from the TOOLKIT samples) in the current directory, and make sure that SOM.IR from the animal directory is on your SOMIR path.

If the voodoo in the current versions of these tests is voodooish enough, this should work automatically now.

Possible bugs: I could not manage to make starting up SOMDD and WPDServer as reliable as I want it. I put workarounds against many bugs/undocumented-limitations, but still things are not that good. Things run smoothly if the deamons are already running.

Thus the test scripts won't shutdown the servers unless $ENV{PERLTEST_SHUTDOWN_SERVERS} is true.

===================== What to do if the servers get stuck?

Sometimes killing WPS and SOMDD (in this order?) helps. There is a tool in utils/dsom_shutdown_servers.pl which performs the shutdown of DSOM in a documented way (with many fallbacks), but it usually blocks. Killing the blocked utility is benign.

To clean this condition:

  1. run utils/dsom_shutdown_servers.pl (probably some errors are going

    to be reported);

  2. kill the second copy of PMSHELL (if it would not die itself

    when SOMDD terminates);

  3. run utils/dsom_start_servers.pl.

[Fact: when WinRestartWPDServer() returns success, WinIsWPDServerReady() does not return true for a long time (may be a second).

Conjecture: even after WinIsWPDServerReady() returns true (we wait for this), the server is not yet actually ready. This is why starting it several seconds before the tests makes everything run smoothly.

No, putting a 5sec delay does not help... The damage happens when the servers are shut down. This is why everything is OK if the servers run before starting the tests: then they are not shut down by the tests. ]