[Aldor-l] Expected results from regression tests

[pip88nl at gmail.com]

Hi,

I currently have a collection of aldor source files that are compiled and
executed as well as interpreted. This way I have some form of automated
assertions of the compiler still doing the same thing as before I changed
anything. The problem I am now having is twofold: platform dependent
tests (those that give error messages with memory addresses or memory
sizes) and tests that just do two completely different things on
different platforms are useless. I can't automatically compare them
sensibly.

What is the test "aldor/lib/axllib/test/pack3.as" supposed to output,
theoretically? The SVN contains expected outputs from many tests, among
which is pack3. The expected output is:
  2
  array(579376, 579408, 579424)

But on Linux/x86_64, I get:
  2
  array(4.6125867383528622e+18, 4.6168651579988644e+18, 4.619342137793918e+18)

and on Linux/i386, I get:
  2
  array(0, 0, 0)

For the interpreter run, it outputs:
  2
  array(2.3999999999999999, 4.5999999999999996, 6.7999999999999998)

for both platforms. What is correct? None of the outputs matches the
expected output as written in the testout directory of axllib,

bug1224.as from aldor-bugs (on the aldor.org website) shows 6 lines of
trivial output and then:
  list(5524816)         # on the compiler, x86_64 and i386
  list(11327888)        # on the interpreter, x86_64
  list(11499584)        # on the interpreter, i386

bug1022.as on the compiler says:
  Constructing D(5489088)       # on x86_64
  Constructing D(5487824)       # on i386

bug1295.as on the compiler says 1,12 followed by:
  12,5421520    # on x86_64
  12,5425616    # on i386
on the interpreter, for both:
  12,-6149101905609984652


Furthermore, on x86_64, "aldor/lib/axllib/test/mandel_t.as" outputs
nothing. That's obviously wrong, so this test is broken for that platform
and is therefore unusable.

All of this just seems so random. Can some more people post the results
of those tests on their systems, with their aldor compilers and if
possible say what output is theoretically correct? For now, I have to
disable these tests.

Pippijn
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://mail.aldor.org/pipermail/aldor-l_aldor.org/attachments/20080818/70442d5a/attachment-0002.bin>