Fix for Payment Request reorganizations

[aguycalled]

This PR prevents payment requests with invalid hashes (not set yet or out of the main chain) to count for the already requested balance of a proposal.

[mxaddict]

@aguycalled I see that your travis-ci run for this PR on x86_64 linux is failing with similar error to my test, any ideas on what the cause is?

[mxaddict]

I'm not really familiar with the logic in the wallet CFUND code, so I don't think I'm qualified to review that.

But the changes i the unit test seem good.

I just need you to make 1 edit

[mxaddict]

Maybe we should add .editorconfig to the project?

[mxaddict]

So we can standardize the white space and some other basic configs

[aguycalled]

+1

[aguycalled]

The issue this PR fixes can be reproduced as follows:

NODE1> staking false
NODE2> staking false
NODE2> addnode 127.0.0.1:node1port onetry
NODE1> generate 300
NODE1> donatefund 100
NODE1> getnewaddress
Save output in %1
NODE2> getnewaddress
Save output in %2
NODE1> sendtoaddress %2 5000000
NODE1> createproposal %1 100 3600 test
Save hash in %3
NODE1> generate 1
NODE2> listproposals
Assert output.size > 0
NODE1> proposalvote %3 yes
NODE1> generate 250
NODE1> createpaymentrequest %3 100 test true
Save output in %4
NODE2> disconnectnode 127.0.0.1:node1port
NODE1> forcetransactions [“%4”]
NODE2> forcetransactions [“%4”]
NODE1> staking true
Wait for 3+ blocks.
Store best block hash in %5
NODE1> staking false
NODE2> staking true
Wait for 1 block. Node 2 only has one output so it will only stake once.
NODE2> staking false
NODE2> addnode 127.0.0.1:node1port onetry
Check that Node2.bestblockhash == %5

Tested succesfully with 2 nodes locally in devnet.
RPC test unit needs to be added.

[mxaddict]

Test looks good.

[proletesseract]

compiling and starting to review this now.

[mxaddict]

compiling and starting to review this now.

Looks like the latest build has an issue with cfund-paymentrequest-state-reorg.py

[mxaddict]

Roger that.

…

On Thu, May 2, 2019, 14:26 proletesseract ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In qa/rpc-tests/cfund-paymentrequest-state-reorg.py <#456 (comment)>: > + while self.nodes[1].getblockcount() == blockcount_1: + time.sleep(1) + + print("nodes staked!") + + self.nodes[0].staking(False) + self.nodes[1].staking(False) + + node_0_best_hash = self.nodes[0].getblockhash(self.nodes[0].getblockcount()) + node_1_best_hash = self.nodes[1].getblockhash(self.nodes[1].getblockcount()) + node_0_nbits = int(self.nodes[0].getblock(node_0_best_hash)["bits"], 16) + node_1_nbits = int(self.nodes[1].getblock(node_1_best_hash)["bits"], 16) + + # Node 0 and Node 1 have forked. Difficulty in Node 0 must be higher (hence lower nbits) + assert(node_0_best_hash != node_1_best_hash) + assert(node_0_nbits < node_1_nbits) Yeah, the hashes are different but the nbits are the same. im reverting the target time to 30 seconds and seeing if it makes a difference. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#456 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAIDAKJXIJSP6G5BDKQTAY3PTKCRTANCNFSM4HJPJQKA> .

[proletesseract]

Roger that.

You mean ACK :P

[mxaddict]

Roger that.

You mean ACK :P

I see ACK on alot of opensource projects, what does it actually stand for? :)

[proletesseract]

Roger that.

You mean ACK :P

I see ACK on alot of opensource projects, what does it actually stand for? :)

It's just short for "Acknowledged" AFAIK.

see: bitcoin/bitcoin#6100

[aguycalled]

I've changed the check so it looks at chainwork instead of nbits.
chainwork = prevblock.chainwork + nbits

[mxaddict]

Looks like final commit fixed the issue with the test.

@craigmacgregor ready to merge? :)

[proletesseract]

Travis failing again. Investigating.

[aguycalled]

it's failing again uniquely because of a random fee in the coldstaking_* tests

[proletesseract]

my local build and tests pass on Ubuntu 18.04, waiting for Travis.