|
|
Suspected A6 publishing error
#169586
11/27/07 01:54
11/27/07 01:54
|
Joined: Mar 2003
Posts: 5,377 USofA
fastlane69
OP
Senior Expert
|
OP
Senior Expert
Joined: Mar 2003
Posts: 5,377
USofA
|
We have a level that we work with in the development version. Here is a screenshot of it: However, when we publish it and do NOTHING else to it, we get the following: We did some investigation and used ent_next to cycle through all the entities. We did this in both our development and published versions: As you can see, some level geometry is completely missing from our published version. Notice also that the two golems on the left are missing in the list and replaced by starbases (the blocky objects). Similarly, the tree to the right is replaced by our ball entity and also missing from our published list. Another symptom (not seen in these shots) is that non-level geometry that is created by clients will come in with the wrong orientation and the wrong scale. We have all the nosends on but even when we turn them off and manually set the scale to 1, the problems persist. Any clue as to what could be causing this?
|
|
|
Re: Suspected A6 publishing error
[Re: jcl]
#169588
11/27/07 14:05
11/27/07 14:05
|
Joined: Mar 2003
Posts: 5,377 USofA
fastlane69
OP
Senior Expert
|
OP
Senior Expert
Joined: Mar 2003
Posts: 5,377
USofA
|
Well we don't save or load games so that's not an issue.
As to the multiplayer, we have the exact same code on developement as on publishing. We have the same levels in all our application, we bind all the needed models in every client and the server, and we are very careful to separate our network indicies from our local indicies. Furthermore, if it was a bad index issue, it should affect development and publishing equally and as you can see, in development things run fine.
If the difference that you are refering to is the list we have above, note the following: in development everything is running fine and when we output all the entities in the level (using ent_next() calls), the "development version" list is what we get; when we publish and follow the exact same procedure, using the ent_next() results in the "published version" column and thus the missing models.
In other words, the missing models are not of our doing; that is one of the symptoms that we are noticing and everything is exactly the same in publishing than what we have in development.
|
|
|
Re: Suspected A6 publishing error
[Re: jcl]
#169593
11/28/07 19:19
11/28/07 19:19
|
Joined: Mar 2003
Posts: 5,377 USofA
fastlane69
OP
Senior Expert
|
OP
Senior Expert
Joined: Mar 2003
Posts: 5,377
USofA
|
Quote:
the speed is different
We knew this one. We have several timing issues in our game where we use wait(1) and this causes irregular behaiviour. However, we've ruled out timing in this issue since we have no control over entity creation when it's part of the level.
Quote:
internal lists, responsible for the entity indices, can also be different.
Now this is interesting! We had noticed that when we do dev vs. pub, the handles seemed to be different from run after run. However, we were not sure WHY they would be different.
Could you expand a bit on this point? Why would the internal lists in publishing be different than in development? I would like to know for as you stated before, we need to make sure that our network indicies do not interfere with the other indicies. The solution we've used is to put anything that will go over the network (strings, texts in strings, variables, function/actions) as a header in the begining of the file and replicate this header among all our applications. Hence, in principle, every application gets the same index for every network object. But by what you say, going from dev to pub may actaully interfere with this process.
Curious to know more about how the internl list is managed in dev and pub! 
|
|
|
|