repairing teleport tubes w/ a hammer doesn't restore them to functionality.

[fluxionary]

a player reports:

When a sender teleport tube (tp) is broken, use hammer to repair. Now the tp tube sender is not working anymore. Expected: tp tube sender should tp the items as it did before breaking, doesn't work even after setting the channel name again. Workaround: destroy the receiver tube and then place it again and configure to get the sender tp tube working again.

[S-S-X]

Looks like #36, if this is happens with up to date version then I guess that functionality is now broken.

[fluxionary]

Looks like #36, if this is happens with up to date version then I guess that functionality is now broken.

server is using b6c02ac, which is after #37, might be a regression.

[fluxionary]

i recently discovered there's a separate issue that causes receive tubes to get delisted from the teleport db, if mod_storage fails to serialize w/ a file-based backend, due to the db getting too large for LuaJIT. it's possible it affected this issue as well, i haven't done any thorough testing.

[TurkeyMcMac]

The move to mod storage was supposed to avoid the LuaJIT size limit. Each entry is serialized individually by Lua. Could you provide more details of this error?

[fluxionary]

The move to mod storage was supposed to avoid the LuaJIT size limit. Each entry is serialized individually by Lua. Could you provide more details of this error?

possibly i'm wrong, it looks like the file-based metadata is being stored in JSON and not via minetest.serialize.

[fluxionary]

in any event, for some reason, the "receive" tubes are disappearing from the tube DB "randomly" (which might be after a server restart). i still suspect this is because things aren't being written cleanly when using the file-based backend when the server crashes.

[TurkeyMcMac]

The mod storage should be written unless there is an abnormal crash without a backtrace etc.

[fluxionary]

The mod storage should be written unless there is an abnormal crash without a backtrace etc.

yup, that does happen from time to time. moving to sqlite backend should (mostly) make that a non-issue.