[DotGNU]Portable.NET Status 30 Nov 2002

[Rhys Weatherley]

Hi all,

Since the 0.4.8 release and our "Big Slashdot Day Out", I've
mainly been cleaning up various loose ends in the code base.

In "pnet", there were some fixes for interoperability with
both MS and Mono.  MS'es CLR can be very pedantic about the
format of IL binaries, even about information that it isn't
interested in using!  Very annoying.

In "pnetlib", I've spent most of the week writing test cases
for "System.Security.Cryptography" and fixing the bugs that
were shown up during the process.  A few internalcalls had
to fixed in "pnet" along the way.

Symmetric ciphers, hash algorithms, and random number generation
now seem to be working pretty solidly.  I still need to test
asymmetric ciphers, certificate parsing, and the utility routines.

Array.Copy finally bugged me enough to actually fix its
handling of zero-length copies.  This should get rid of a
lot of bugs, as it is used pretty much everywhere.

There was a very strange bug in the compiler that was causing
DateTime to give very bogus values for Day and Year.  Four hours
and much head scratching later, I eventually tracked it down to a
bug in 64-bit constant folding.  Unfortunately, this margin is
too small to describe the complete circumstances - as I said,
it was very strange.

Gopal has started work on dynamic flow analysis for variables
in the compiler, but will be exam-bound for a few weeks so
try not to bug him too much.

Jonathan Springer has checked in his number formatting code,
which is very cool.  One amusing anecdote: there was a bug
in string handling within the assembler that Jonathan worked
around by doing "A" + "B" in his test suite.  Then Gopal
optimized string concatenation to get rid of the "+", which
broke the tests again!  I've fixed the assembler now and
everything is fine.

Let this be a lesson guys.  Don't work around the bugs!
Fix them! :-)

Michael Moore has begun implementing DateTime formatting,
and James McParlane is inching closer to getting pnet working
on Alpha processors.

Looking at the status page, the remaining areas to tackle in
pnetlib for ECMA-compatibility are "System.IO", "System.Net",
and the ever-annoying "System.Xml".  The first two are the
most critical.  We are very close now - I can almost smell it.

Cheers,

Rhys.