[qa] allow assert_fee to have a bit more wiggleroom
ClosedPublic
Actions

Authored by schancel on Oct 1 2018, 22:14.

Details

Reviewers

deadalnix
jasonbcox

Group Reviewers

Restricted Project

Commits

rSTAGINGd68bb71616a3: [qa] allow assert_fee to have a bit more wiggleroom
rABCd68bb71616a3: [qa] allow assert_fee to have a bit more wiggleroom

Summary

As per title

Test Plan

make check && ./test/functional/test_runner.py

Diff Detail

Repository

rABC Bitcoin ABC

Branch

util-fiddle

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 3472
Build 5020: Bitcoin ABC Buildbot (legacy)
Build 5019: arc lint + arc unit

Event Timeline

schancel created this revision.Oct 1 2018, 22:14

Herald added a reviewer: Restricted Project. · View Herald TranscriptOct 1 2018, 22:14

Build Bitcoin-ABC / Diffs / Bitcoin-ABC Diff Testing started.

schancel added a child revision: D1859: [qa] use billable_size is testnode fee estimates.Oct 1 2018, 22:20

Harbormaster completed remote builds in B3472: Diff 5216.Oct 1 2018, 22:31

Build Bitcoin-ABC / Diffs / Bitcoin-ABC Diff Testing passed.

Please update the description to explain why this needs to happen.

This revision now requires changes to proceed.Oct 2 2018, 11:24

I actually ended up fixing a different cause of the failure.

schancel reclaimed this revision.Oct 5 2018, 01:11

This revision now requires changes to proceed.Oct 5 2018, 01:11

Okay, so for fundrawtransaction, when fees are dropped, the first transaction sent in a series of fee checks is with a feerate of 100sat/kb. This results in a fee of 0.00000202. This is ceiled so as to not be under the minimum relay fee. It then proceeds to calculate expected fees based on multiples of this due to the dummy signatures that get added during fundraw transaction.

However, the full fee was 0.000002016. So when it tries another transaction with a 10x feerate it expects 0.00002020 due to the previous truncation -- it then fails.

Something like this patch is needed to deal with that kind of problem. This particular method may not be best.

jasonbcox requested changes to this revision.Oct 5 2018, 18:13

jasonbcox added a subscriber: jasonbcox.

jasonbcox added inline comments.

test/functional/test_framework/util.py
28	Depending on how rigid the +/- 10 range is, you could include it as a parameter to this function and allow the individual tests to determine the tolerance.
34	Update this comment and make it reflect how +/- 10 was chosen