CRC and NID errors (reaction shot related?)

[Fomka]

Hello all,

I have a question about CRC errors reproduction. Do you have ideas on
how to reproduce that errors in hotseat?

Serge mentioned in his note to issue 461 (http://ufo2000.net/mantisbt/
view.php?id=461) that "Game synchronization problems can be reproduced
in hotseat " and suggested the following setup.

1. Tweak soldiers initialization code to set starting morale value
equal not to 100, but to some lower value (I used 30)
2. use a small 4x4 map with an open area (so that there is no cover)
3. deploy two opposing teams with a large number of soldiers in such a
way that they can easily reach each other
4. start the game and just move each soldier one step per turn towards
the opposing team
5. due to lots of panic and reaction fire happening, crc errors show
up eventually (usually quite fast)

I followed that setup and did not get CRC errors as was supposed to.
So I've uploaded my test suite (team setups, weapons, games saves, a
replay) to the same issue page. The question to Serge remains the
same: can you give more details on how you reproduced the problem?

--
Best regards,
Fomka

[Siarhei Siamashka]

Actually it turned out to be another bug, reproducible only on big endian
hardware (such as playstation3). I have fixed it here:
http://ufo2000.svn.sourceforge.net/viewvc/ufo2000?view=revision&sortby=rev&revision=1149

Regarding the possible source of the whole class of these synchronization bugs.
They are likely related to the differences in precision of floating point
calculations, discussed in:
http://ufo2000.net/mantisbt/view.php?id=250

This floating point issue was never fully resolved, but only partially
workarounded. A real solution is the use of fixed point math for
everything related to game logic. And even if the primary source of these
synchronization problems is some other bug, we just can't simply ignore
this known floating point related problem.

--
Best regards,
Siarhei Siamashka