GENERAL
-------

This file explains and announces testing related issues:

Our goal is to run uade on most computing platforms that support at least
32-bits arithmetics (floating points only required through software
emulation).

At the moment we have had working solutions on following processors:

	x86, alpha, pa-risc, sparc, mips, ppc

One ppc version also ran on a non-unix platform (morph os). In my opinion
UAE team has done surprisingly good job with their infrastructure. I applaud
them.

We welcome any feedback, including test cases. PLEASE tell us if you feel
that our tests suck, or are false or unreliable.

TEST CASES
----------

md5sum test case 1:

40fa2d92145c8bced1803449940cf40f  MDAT.TurricanII_1-Dessert
8acd771885f5aa45382e88010ebd2e01  SMPL.TurricanII_1-Dessert

Kimmo Kinnunen <kimmo.kinnunen@cs.tut.fi> reported us that uade works on
Alpha GNU/Linux platform. But what I (shd) was not sure about was that it
would produce exactly the same raw sound output as x86 GNU/Linux uade with a
very different compiler.

alpha gnu/linux uade 0.80-pre2 
compiler version: ccc-6.4.9.005 (switches -arch ev56 -fast)
uade: module 'MDAT.TurricanII_1-Dessert' (13024 bytes)
511a0b694c3c0ab7df1d229cf7d0d7cd  paska.1.raw

x86 gnu/linux uade 0.80-pre2
compiler version: egcs-2.91.66
uade: module 'MDAT.TurricanII_1-Dessert' (13024 bytes)
511a0b694c3c0ab7df1d229cf7d0d7cd  paska.2.raw

(uade songname -outpipe 1 > paska.raw was used to produce raw dumps of sample
data)

note 1: this case does not apply to the xmms plugin.

note 2: uade 0.80-pre2 and uade-0.80-pre3-cvs produced different md5sums
	on x86 GNU/Linux. We have a pretty good idea why this has happened
	(sound core has changed => initial timing has probably changed).
