[fixed] [Bug] in 3.4.0dev+r1116: Workers fail to harvest

[tomreyn]

Slightly rearranged IRC log:

<tomreyn> softcoder: these changes to how the workers harvest are problematic
<tomreyn> workers stop harvesting, or can not even start to (cancel it as soon as they arrive at the harvesting location, but the order to harvest is still present (yellow arrow))
<titi_linux> the workers are corrupt now
<titi_linux> they often simply stopped
<titi_linux> cpu replace was fine
<titi_linux> and cpu has same problems
<titi_linux> much more stuck than before
<titi_linux> and completly nonsens kind of stuck
<tomreyn> got a screen shot?
<titi_linux> no
<tomreyn> well should be easy to reproduce
<titi_linux> I think its very easy to reproduce
<titi_linux>
<titi_linux> but cpu takeover was really fine
<ultifd> can we make it optionable for cpus to replace, sometimes you don't want them to

[ultifd]

Well, we shouldn't worry too much as it is easy to reproduce (I guess). Then trying to find the cause...
Too bad I suddenly had connection problems when we were going to test.
I wonder if titi had a log...

[softcoder]

Try rev 1118, I tried a bugfix for this.

[tomreyn]

This issue is fixed in or before r1121 for me. I played a single player game on this revision and the workers worked fine.

[tomreyn]

Sometime between r1121 and r1125, a related bug is reintroduced. While workers do not cancel their attempts trying to harvest anymore, they will (maybe just in some cases) walk around indefinitely, trying to do so.

I've had two workers involved in this issue which would indefinitely do this while trying to harvest wood:
* after worker A leaves this position, worker B goes stand in some corner which was previously cut into the wood; he stands there for a couple seconds, then walks away to a different position (apparently still trying to reach cuttable wood)
* after worker B leaves this position, worker A now goes stand in said corner which was previously cut into the wood; he stands there for a couple seconds, then walks away to a different position (apparently still trying to reach cuttable wood)
* after worker A leaves this position, worker B now goes stand in said corner which was previously cut into the wood; he stands there for a couple seconds, then walks away to a different position (apparently still trying to reach cuttable wood)
...

I should have taken a video of this but unfortunately didn't come to think of it, so here's just some images hopefully showing it, too...

The map this happened on is "Conflict". My base was a little to the south of this current screen position and the location where the bug occurred is encircled here:


The following series of screens should demonstrate the failure to harvest described above:

[titi]

I think this is no bug. You maybe gave them the command to harvest wood in an unreachable place?
If this is the jungle tileset I see a not harvestable big tree right and left to the "waiting" worker, so he maybe cannot reach the tree this way.
The path to the wood seems to be blocked by big trees.

[softcoder]

With the current code in svn Rev1126, the unit harvesting which cannot get to a location (for whatever reason) will 'eventually' check in a faction wide cache for another location to harvest the same resource type. This is in cases where they get blocked too many times and cannot get to that resource. OF course in a tight space even with this logic they can get continuously blocked by other moving units and objects and walk around 'seemingly aimlessly' due to congestion.

Thanks