[fixed] r4757: CRC mismatches

[tomreyn]

Titi (Windows7-64, using a 32-bit build produced on cudgellady) and I (atibox, using self-compiled Linux builds, hosting) just played several games using the CRC check monitor. Sadly, on several builds Titi got disconnected with CRC mismatches quite fast. We always played with the same game setup. Whenever there was a CRC mismatch, it occurred during the first large inbound attack.

Our games:

r4757 resulted in a mismatch:
http://tomreyn.megaglest.org/debugcrcworld_tomreyn_4757_20131110_1.tar.xz
http://titi.megaglest.org/logs/debugCRCWorld_4757.log_client.7z

r4747 resulted in a mismatch:
http://tomreyn.megaglest.org/debugcrcworld_tomreyn_4747_20131110_1.tar.xz
http://titi.megaglest.org/logs/debugCRCWorld_4747.log_client.7z

r4732 resulted in a mismatch:
http://tomreyn.megaglest.org/debugcrcworld_tomreyn_4732_20131110_1.tar.xz
http://titi.megaglest.org/logs/debugCRCWorld_4732.log_client.7z

r4716 resulted in a mismatch:
http://tomreyn.megaglest.org/debugcrcworld_tomreyn_4716_20131110_1.tar.xz
http://titi.megaglest.org/logs/debugCRCWorld_4716.log_client.7z

r4707 resulted in a mismatch:
http://tomreyn.megaglest.org/debugcrcworld_tomreyn_4707_20131110_1.tar.xz
http://www.sendspace.com/file/97ho66 (terminal output: http://pastebin.com/raw.php?i=KqXSuraC )

r4697 resulted in a mismatch:
http://tomreyn.megaglest.org/debugcrcworld_tomreyn_4697_20131110_1.tar.xz
http://www.sendspace.com/file/n5xvia (terminal output: http://pastebin.com/a2tGV3Qh )

r4694 worked, no CRC mismatch, we did not play the entire game this time, but clearly after the point where it crashed before. (Titi had the feeling that this version was slower on windows.)

r4682 worked, no CRC mismatch, we played an entire game (roughly 45 minutes). ( big game with lots of units and large battles )

I had no noticeable performance differences, but Titi said it was faster for him on r4682.

The W32 snapshot builds we used are temporarily available at http://snapshots.megaglest.org/windows/DONOTUSE/


Note:All updates have been added above

Update #1: added r4694

Update #2: added r4716

Update #3: added r4707

Update #4: added r4697

[titi]

And this testing took very long   
but we found something 

[tomreyn]

This seems to be fixed. Coldfusionstorm (W32 client) and I (atibox host) played the same game today on SVN head, using EnableNetworkGameSynchMonitor=true and there was no CRC mismatch detected. We will have to ensure to always test with the network game synch monitor (CRC / checksum checks) enabled before releases and sporadically during release cycles in the future.