GT01 is currently still affected by one major problem: in some circumstances it becomes unable to save the game to file.
I ran some hundreds of automatic tests of GT01 under different game configurations, and I measured the failure rates.
Each line of Table 1 is a set of 50 tests under the same game configurations. A test consists exclusively in generating a new game and /save it to file.
The two leftmost columns are the test conditions, the right columns are the related results.
For tests 6th and 7th the map size parameter is directly specified and mapsize is set to fullsize.
Table 2 lists only the settings which differ between GT00 and GT01. The complete sets of settings are on greatturn.org.
All tests are conducted with 11 AI players.
Table 1 legend:
Ruleset: full ruleset used Settings: full set of settings used /save failures: number of tests in the series which presented at least one /save failure segfaults: amount of tests in the series which presented at least one segmentation fault of freeciv-server def: vanilla Freeciv's default configuration, not related with Greatturn GT01: fully respecting GT01 customs [numbers in settings column]: setting numbers from Table 2 which are toggled to the value used by GT00.
In this case all other settings reflect GT01.
|Tests [table 1]|
|ruleset ||settings ||-->||/save failures ||segfaults |
|def ||def ||0 ||2 |
|gt01 ||def ||0 ||0 |
|def ||gt01 ||5 ||6 |
|def ||4679 ||0 ||0 |
|def ||6 ||0 ||1 |
|def ||gt01 size=50 ||0 ||6 |
|def ||gt01 size=14 ||1 ||0 |
|Per-match settings [table 2]|
|n°||gt00 ||gt01 |
|specials ||1||900||400 |
|startunits||2 ||cccwwxx ||cccwwxxx |
|dispersion||3||2 ||3 |
|generator||4||island ||fractal |
|landmass||5 ||35 ||29 |
|tilesperplayer||6||100 ||400 |
|alltemperate ||7||enabled ||disabled |
The ruleset used has no effect on the stability of the server. The settings group GT01 crashes the server.
Toggling tilesperplayer from 400 to 100 (5th test of Table 1) sensibly improves stability; in particular, the 5th test generates maps of the same size of the one used by GT00.
The 6th test uses a map size of 50k tiles: it generates some apparently non-harmful segfaults and all the savefiles are correctly generated.
The 7th test reduces the map size from the previous test, generating no segfaults but 1 harmful failure to save.
The amount of segfaults produced for each series of tests seems to be proportional to the map size, and seems to be non-harmful to the consistency of the game and the savefiles.
The /save failures are observed with map sizes of about 400 tiles per player.
It seems that to gain server stability the map size needs either to be increased or reduced around the value of 400 tiles per player.
I have no other clues.