[Spread-users] BUFFER_TOO_SHORT && endian_mismatch >= 0

[Jeremy Hylton]

We have a Spread application that is periodically failing with an
unexpected BUFFER_TOO_SHORT error from SP_receive().  We're using the
Python wrapper, which normally hides those errors and retries
SP_receive() with a larger buffer.  So it should not actually be
possible to see the error in Python code.

The one exception, however, is if endian_mismatch would be >= 0.  n
that case. the Python wrapper raises an exception.  It looks like this
is happening, but there's nothin in the docs that suggests
endian_mismatch could be non-negative.  (Unfortunately, I don't know
what the actual value is.)

I glanced at the code in sp.c and couldn't rule out the possibility
that endian_mismatch was 0, although it doesn't seem like the code is
intended to set it to 0.

Any ideas?  Is it possible to get a non-negative endian_mismatch?  If
yes, what would it mean?

Jeremy