getblocktemplate on testnet giving wrong target and bits?

fireduck

sr. member

Activity: 392

Merit: 251

Quote from: gmaxwell on February 19, 2015, 06:02:59 PM

Quote from: fireduck on February 19, 2015, 05:26:21 PM

It is also possible that everything is fine, and jerks are just generating testnet blocks two hours in the future to mess with me.

Well, not to mess with you but because they can. Of course, if you generate a single block at the full difficulty you'll reorg out a wad of those blocks that came after you... so they shouldn't be blocking you from mining, only potentially from mining at difficulty 1.

Sure. I've managed to do the testing I intended by accidentally having my node misbehave enough to isolate itself from the network and then build some blocks on my own side chain.

gmaxwell

staff

Activity: 4284

Merit: 8808

Quote from: fireduck on February 19, 2015, 05:26:21 PM

It is also possible that everything is fine, and jerks are just generating testnet blocks two hours in the future to mess with me.

Well, not to mess with you but because they can. Of course, if you generate a single block at the full difficulty you'll reorg out a wad of those blocks that came after you... so they shouldn't be blocking you from mining, only potentially from mining at difficulty 1.

fireduck

sr. member

Activity: 392

Merit: 251

It is also possible that everything is fine, and jerks are just generating testnet blocks two hours in the future to mess with me.

fireduck

sr. member

Activity: 392

Merit: 251

I'm pretty sure there is some bug somewhere in the block template caching relative to the testnet special difficulty after 20 minutes.

I just generated a block by commenting out a line.

pow.cpp
GetNextWorkRequired

I commented out this line:
if (pblock->GetBlockTime() > pindexLast->GetBlockTime() + Params().TargetSpacing()*2)

In my testing, the difference between
pindexLast->GetBlockTime() and pblock->GetBlockTime() is always 6004.

My guess is that something is creating a template block and then not updating it because there are no new transactions and because the next target hasn't changed. And the next target isn't changing because the block time is never changing.

But some of that is guesswork and I'm not sure.

fireduck

sr. member

Activity: 392

Merit: 251

If it is indeed working correctly, why do all the recent found blocks have a much higher hash than the target I am seeing (both before and after the block is found)?

"previousblockhash" : "00000000680260e45f57f0cdbe264366aa5095b4b60cbdf47b3a89b416b5d228",
"previousblockhash" : "00000000b76a0528040b33a15c474261c799ac626fcc889754d01e37f0351a36",

"target" : "00000000000003a2d80000000000000000000000000000000000000000000000",

Looking at the recent blocks, it appears they are coming in almost exactly every 20 minutes. I imagine the code that drops the difficulty (in testnet) if there are no blocks takes effect then. Maybe the node finding them has a clock set a few seconds ahead such that it sees the 20 minutes elapsed and makes a new block before I see the difficulty adjustment. I'm checking every second and am not seeing the window.

gmaxwell

staff

Activity: 4284

Merit: 8808

Getblocktemplate is working correctly there.

Getdifficulty is telling you the difficulty of the prior block. "Difficulty" is not an interface that is suitable for use in mining, its fundamentally imprecise. It's a human friendly number for display.

fireduck

sr. member

Activity: 392

Merit: 251

I've tried this with 0.8 release branch, 0.9 and head. I get the same results for each.

The getdifficulty returns 1.0 and the found blocks confirm that this seems to be the case.

However, getblocktemplate is showing a much harder difficulty in the target:
00000000000003a2d80000000000000000000000000000000000000000000000

Broken up for readability:
00000000000003a2
d800000000000000
0000000000000000
0000000000000000

For that difficulty it should be something like:
0000000100000000
0000000000000000
0000000000000000
0000000000000000

The most recent found block agrees with this:
0000000076f8abad2

./bitcoind -testnet getblocktemplate

...transactions omitted...

   "coinbasevalue" : 2500081729,
   "target" : "00000000000003a2d80000000000000000000000000000000000000000000000",
   "mintime" : 1424334619,
   "mutable" : [
   "time",
   "transactions",
   "prevblock"
   ],
   "noncerange" : "00000000ffffffff",
   "sigoplimit" : 20000,
   "sizelimit" : 1000000,
   "curtime" : 1424334619,
   "bits" : "1a03a2d8",
   "height" : 323482
}

./bitcoind -testnet getdifficulty
1.00000000

I'm trying to run some blocks on testnet to make sure my mining pool software still works correctly.

Huh

Topic: getblocktemplate on testnet giving wrong target and bits? (Read 1993 times)