Client Prediciton Debugging Command/Option
ultranewb
Pro Bug Hunter Join Date: 2004-07-21 Member: 30026Members
Pro Bug Hunter Join Date: 2004-07-21 Member: 30026Members
The Situation: My usual investigation starts with looking at my own ping, the ping of the target, and finally the tick rate of the server. The usual case is one player's ping is high (aka 'euro') and shots don't land on just them. The rarer case is that the entire server is bogged down. The rest of the time, there's no logical reason for the error and this is the case I'm interested in.
Currently, the only way to provide "proof" for inspection is to record a demo before joining any server, or using a third party screen capture with net_stats enabled. The first provides a detailed log, but is impossible to do once you're already joined a server. The second provides scant information and requires a tremendous amount of horsepower to do so.
The Suggestion: The "real" fix would be to fix demos to allow them to record at any time, but I'm assuming the amount of work required to support this is high. Demos, however, really only help the dev team and profile logs are arduous to traverse.
A second best fix, but more appropriate, would be to add a console command to enable hit detection debugging. This would only report anomalies in hit detection and much more information for anomalies that seemingly have no probable cause. Once turned off, some final statistics could be logged.
Currently, the only way to provide "proof" for inspection is to record a demo before joining any server, or using a third party screen capture with net_stats enabled. The first provides a detailed log, but is impossible to do once you're already joined a server. The second provides scant information and requires a tremendous amount of horsepower to do so.
The Suggestion: The "real" fix would be to fix demos to allow them to record at any time, but I'm assuming the amount of work required to support this is high. Demos, however, really only help the dev team and profile logs are arduous to traverse.
A second best fix, but more appropriate, would be to add a console command to enable hit detection debugging. This would only report anomalies in hit detection and much more information for anomalies that seemingly have no probable cause. Once turned off, some final statistics could be logged.