tweak auto-unparking message
ClosedPublic
Actions

Authored by markblundeberg on Jan 14 2020, 08:53.

Details

Reviewers

deadalnix

Group Reviewers

Restricted Project

Commits

rSTAGING638e4857a2c4: tweak auto-unparking message
rABC638e4857a2c4: tweak auto-unparking message

Summary

UnparkBlock will only get called once per chain, i.e., we only see one

log message regardless of how many blocks got unparked on that chain,
since UnparkBlock will unpark all ancestors of pindexTest and it
results in everything from genesis to pindexNew being unparked. So
"unpark chain" rather than "unpark block" makes more sense as a message,
as I'd expect to see the latter one for every single block.

Ideally when this message happens we should be in a state where

pindexNew == pindexTest, however it might somehow happen that this
isn't the case and we should print the tip of the chain that is being
unparked.

Test Plan

ninja check

Diff Detail

Repository

rABC Bitcoin ABC

Branch

preciseunpark

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 8920
Build 15814: Default Diff Build & Tests
Build 15813: arc lint + arc unit

Event Timeline

markblundeberg created this revision.Jan 14 2020, 08:53

Herald added a reviewer: Restricted Project. · View Herald TranscriptJan 14 2020, 08:53

Herald added a reviewer: deadalnix. · View Herald Transcript

Build Bitcoin-ABC / Diffs / Diff Testing started.

Build Bitcoin-ABC / Diffs / Diff Testing passed.

Harbormaster completed remote builds in B8920: Diff 15434.Jan 14 2020, 08:58

Fabien added a subscriber: Fabien.Jan 14 2020, 13:11

Fabien added inline comments.

src/validation.cpp
2620	Is there any non-bogus case where `pindexNew` can be different from `pindexTest` ?

markblundeberg added inline comments.Jan 14 2020, 13:36

src/validation.cpp
2620	It would be pretty rare it seems to me. Example 1: A -- B -- C* You start out with tip on C. A -- B -- C* \ -- D Block D comes in and gets parked because it's a 'deep reorg'. A -- B* -- C \ -- D You invalidate C for some reason, so now your tip is on B A -- B* -- C \ -- D -- E Block E comes in and is not parked during AcceptBlock because it wouldn't count as a 'deep reorg' anymore. Then FindMostWorkChain looks at E and sees it's the longest chain, so it unparks D. Example 2: Something similar to the above happens because you shut down your node and restarted with -parkdeepreorg turned off -- the new incoming block on alt chain doesn't get parked. Example 3: Probably can make something weird happen using `parkblock` RPC call. (and I think this could also happen with multiple blocks on each chain)