Conventions for test output
% $Id$

Lines beginning with % are comments and will be ignored
in all automatic processing. They'll only be useful
when running test programs manually.

All other lines will be of the form:

<variable>=<value>

Where <var> is a string of alphanumerics and blahs, and
<value> is any printable chars up to the end of line.

The output should begin with identifying lines something
like this:

test_program=
platform=
product=
version=

No line should be more than 80 chars long.

Obviously this isn't the direct output from the test
program. Things like abortion, assertion failure &c &c
will tend to produce other junk in the test output; this
will be processed by a (platform-specific) program to produce
the eventual test output. The processing program will
pass straight through all lines beginning with %, and all
blank lines. It will strip a ! from the beginning of any
line which begins with one (use for !<var>=<value>),
and do platform-speciic stuff on all the remaining lines.
It will complain about (at least) the first line it
doesn't understand.


