View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0005604 | Spring engine | General | public | 2017-06-05 07:17 | 2017-06-13 12:26 |
| Reporter | Google_Frog | Assigned To | Kloot | ||
| Priority | normal | Severity | major | Reproducibility | sometimes |
| Status | closed | Resolution | unable to reproduce | ||
| Product Version | 103.0 +git | ||||
| Summary | 0005604: 103.0.1-1019 Desync 2 | ||||
| Description | Logs: * https://github.com/ZeroK-RTS/CrashReports/issues/1314 * https://github.com/ZeroK-RTS/CrashReports/issues/1316 * https://github.com/ZeroK-RTS/CrashReports/issues/1317 * https://github.com/ZeroK-RTS/CrashReports/issues/1318 Battle: http://zero-k.info/Battles/Detail/459195 | ||||
| Tags | No tags attached. | ||||
| Checked infolog.txt for Errors | |||||
| related to | 0005603 | closed | 103.0.1-1019 Desync |
|
|
I'm going back to 103.0.1-910-g9ce3d70 because I don't think people will stand for this level of desync and the ongoing memory management change are relying too much on live testing. |
|
|
the ongoing memory management changes are a forced consequence of players getting into huge games, then reloading N times and hitting limits the engine would have never exposed prior to reload support. live testing is the only practical way to find and fix these, so I think reverting is a mistake. |
|
|
Ok, I wasn't aware the memory changes were bought on by reload usage. |
|
|
nearly every recent major issue is in some way related to repeated reloading, which uniquely affects ZK since all other games still follow the traditional launch-play-exit cycle using external lobbies. the engine's present state does not yet (and may never quite) allow to do so indefinitely, though it is far better than 103 at managing memory over time. one problem is that players tend to dig holes down to the edges of stability faster than we can fill them. |
|
|
infolog of the desyncing player (snoke) is missing. at the same time the number of desyncs has dropped, correlating with forced path-cache regeneration in -1026. that leaves 0005580 and pool corruption as known desync sources, of which the first can be helped by a restart. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2017-06-05 07:17 | Google_Frog | New Issue | |
| 2017-06-05 07:52 | Google_Frog | Note Added: 0017791 | |
| 2017-06-05 11:05 | Kloot | Note Added: 0017795 | |
| 2017-06-05 11:06 | Kloot | Relationship added | related to 0005603 |
| 2017-06-05 11:28 | Google_Frog | Note Added: 0017798 | |
| 2017-06-05 16:16 | Kloot | Note Added: 0017799 | |
| 2017-06-05 16:18 | Kloot | Note Edited: 0017799 | |
| 2017-06-13 12:26 | Kloot | Assigned To | => Kloot |
| 2017-06-13 12:26 | Kloot | Status | new => closed |
| 2017-06-13 12:26 | Kloot | Resolution | open => unable to reproduce |
| 2017-06-13 12:26 | Kloot | Note Added: 0017837 |
