ProductionQueue WARNING: Invalid command: can't train 0 units

[elexis]

In this match with r22951, nani somehow managed to send a command from the UI where 0 units were ordered to be trained at a kennel:

turn 6181 500
cmd 8 {"type":"returnresource","entities":[9978,9998,12158,12159,12160,12851,12887,12902,12903,13500,13501,13502,13503,13579,13609,13619],"target":10731,"queued":false}
cmd 8 {"type":"gather","entities":[9978,9998,12158,12159,12160,12851,12887,12902,12903,13500,13501,13502,13503,13579,13609,13619],"target":5766,"queued":true}
cmd 2 {"type":"train","entities":[10077,10092,10105],"template":"units/brit_war_dog_b","count":2}
cmd 2 {"type":"train","entities":[10118],"template":"units/brit_war_dog_b","count":0}
end

In the replay / for all players:

Turn 6181 (500)...
WARNING: Invalid command: can't train 0 units

So there must be some kind of way to trigger this command from the UI, but that should not be possible to send it, i.e. should be fixed in the UI as well, not only simulation.

[elexis]

faction02 reported the way to reproduce the bug:

1. build 3 kennels
2. train 28 dogs (limit is 30)
3. reduce batchcount size to 1
4. Shift+click on the button to train a dog

Then 2 dogs + 1 JS GUI warning is pushed.

The relevant code would be in addTrainingToQueue, flushTrainingBatch of input.js.

[elexis]

• Reproduced with n in {3, 4, 5} kennels, where limit - (n-1) dogs had been trained, batchSize = 1 and then flushing one batch

• Reproduced with batchsize=3 too:

  WARNING: batchedSize 3
  WARNING: appropriateBuildings.length 5
  WARNING: g_BatchTrainingEntityAllowedCount 12
  WARNING: buildingsCountToTrainFullBatch 4
  WARNING: Invalid command: can't train 0 units
  

• It's the remainer part of flushTrainingBatch.

• g_BatchTrainingEntityAllowedCount % batchedSize is only negative if g_BatchTrainingEntityAllowedCount is negative, see specs

• g_BatchTrainingEntityAllowedCount is at least 0, see line ret.canBeAddedCount = Math.max(ret.entLimit - ret.entCount, 0);

• So capturing 0 is sufficient to cover all these reproduced cases.

• batchedSize is at least 1*

• So this non-0 check is sufficient for all "train" commands to capture all ways to trigger this error*

(* (unless there is a bug where g_NumberOfBatches could be 0 and this code called))

• blame r18421, r19826, r12832 and everything

[elexis]

In 23192:

Fix warning about attempted empty batch train occurring if g_BatchTrainingEntityAllowedCount % batchedSize == 0, refs #1432, #4059, rP12832, rP18421, rP19826, fixes #5602, reported by nani, reproduced by faction02.