jamespetts / simutrans-extended

Sources for the Experimental branch of the popular game Simutrans
http://forum.simutrans.com/index.php?board=53.0
Other
82 stars 53 forks source link

Std2406 01 #663

Closed Ranran-the-JuicyPork closed 5 months ago

Ranran-the-JuicyPork commented 5 months ago

Thank you for testing. I think it's a step forward that we found out that the cause is in there.

This result suggests that the bug is not related to file changes in the network folder.

We know where the send of NWC_GAMEINFO failed occurs, and related changes may include, for example, gameinfo_t not being read/written correctly, so we will split the patch further and submit it so that it doesn't include that commit.

The new split is mostly codebase changes, including a change to translate "%" to "%%". If you could test this, that would be appreciated.

For your reference, the bug that caused the time to freeze for clients who connected immediately after starting the game may have been fixed in the previous r10996 and std2406-1 branches, but the bug still exists, at least in the std2406-01 branch. I don't think this issue was seen in the r10996 branch.

Ranran-the-JuicyPork commented 5 months ago

The std2406-01 branch does not respond well to text input. This is a bug of the standard. The next branch, std2406-02, is ready. This branch will fix the bug. std2406-02 version adds scroll threshold options and increases save revisions.

jamespetts commented 5 months ago

Thank you very much for this. I realise that I have been making an error in my test method which showed the error that I had been reporting incorrectly. When I saw that this version also produced the error, I wanted to verify my test methodology, so I repeated the test with the master branch and found that it also produced the same error. What was happening was that I was clicking the IP address in the server list using the internet IP, which was being blocked by the local firewall. However, if I manually entered the loopback address (127.0.0.1) in the query server dialogue, the information would appear correctly.

I have now re-tested the r10996 branch with this method, and found that it works. I have also successfully tested the ex-15 version of this branch (merged with the latest master). As a result of that, I am about to commit this version and merge it with master.

I can only apologise most sincerely for the error, which will have cost you some unnecessary time. I think that this error arose because an earlier version that I had tested failed both on selecting in the list and manually typing in the address, and I had not remembered this detail when I came to test again. If it is any consolation, you having produced this even earlier version did assist in the process of realising that the earlier test method was incorrect.

In any event, thank you very much for your work on this integration from Standard - it is much appreciated. It will be interesting to see whether the losses of synchronisation are reduced when this goes live.