Python 2.0 vs OS/2+EMX

[Andrew MacIntyre]

In attempting to build 2.0 on OS/2 (v4 + FP12) with the EMX gcc compiler
(0.9d fix03 w/gcc 2.8.1) I've run into a couple of oddities.

I've patched the source based on patches done by Andrew Zabolotny for his
EMX port of 1.5.2, and have succeeded in building both single and
multithreaded versions.

I have been able to run the regression test with the single threaded
interpreter and a fair proportion of the tests pass, and most of the rest
fail graciously although a number need to be followed up as to why they
fail.  However I'd appreciate thoughts about the issues causing the
following problems:-
1) when running the regression test, regrtest.py reports that it is unable
to create the output files for each test (EACCES is the C lib error).
However I can manually fire up the python binary and create the exact same
file no problems.  ????

2) the longexp test causes python to consume all available memory until
SWAPPER.DAT exhausts available disk space.  Reducing the number of REPS to
30000 (from 65580) doesn't help, but REPS = 10000 works fine.  On the face
of it a memory management problem of some sort, but where to start?

The multithreaded interpreter fails on startup with a message something
like "don't have interpreter lock".  Unfortunately I haven't kept the
multithreaded build, so I can't check this at the moment, and in any case
its a lower priority than getting the single threaded interpreter running.

Any thoughts or suggestions gratefully received.

--
Andrew I MacIntyre                     "These thoughts are mine alone..."
E-mail: andrew.macintyre at aba.gov.au    (work) | Snail: PO Box 370
        andymac at bullseye.apana.org.au  (play) |        Belconnen  ACT  2616
        andymac at pcug.org.au           (play2) |        Australia