Limited 0.45 Cosmos SDK release

[robert-zaremba]

Summary

There is a need in the community to release v0.45 sooner. The idea is to not add any significant effort nor any changes which could create any risk.

Also, we would need to think if with v0.45 release will stop support for 0.42 (which will sunshine anyway at the end of January 2022) and how long to support 0.44.

Here is an "in progress" list of what we an merge into potential 0.45.

Included:

• IterNextCostFlat not charged #10127
• No gas is charged for key length in gas meter #10243
• feat: Add HasSupply method to bank keeper #10393
• feat: add configurable iavl cache size #10561

Maybe:

• fix: Panic if SetOrder* functions forgot modules #10711
• fix: should revert tx when block gas limit exceeded #10770
• Support migration ordering for in place migration #10604
• Add rollback support in the event of an incorrect hash #10281

---

For Admin Use

• Not duplicate issue
• Appropriate labels applied
• Appropriate contributors tagged
• Contributor assigned/self-assigned

[ethanfrey]

Those first two gas related ones are critical for a wasmd 1.o release we can safely stand behind reliability wise.

We are targeting a early January release date and starting the 0.42 to 0.44 migration shortly (the last step left)

Many chains have been asking for this and going ahead on untested versions, and we would be very happy to have those first two gas related prs in an official Cosmos sdk release we can depend on.

I am happy for any other minor changes as well, but please no larger breaking ones

[amaury1093]

Let's remove #10348 :

• it needs a joint effort with the Ledger team to update the Ledger cosmos app, we talked with Juan, and we're planning to do that along with the next major release (so 0.46) with the tip feature (which also needs Ledger update)
• that PR 10348 adds support for tips + feepayer in Amino. There's some non-trivial cherry-picking if we only want to backport feepayer, plus some non-trivial QA.

[amaury1093]

Current timeline is to release on Tues 14th.

[ethanfrey]

Thank you very much!

[amaury1093]

Due to some critical bugs discovered recently, we decided to release 0.45 early January.

[fedekunze]

We'd like to include the following in the next release for Evmos mainnet launch:

• fix: enable setting sequence (nonce) for module account #10536
• fix: nil pointer panic on NewIntFromBigInt #9627
• fix: populate ctx.ConsensusParams for begin blockers #10725
• feat: Add HasSupply method to bank keeper #10393

Edit: New additions

• perf: Improve the speed of coins.String() #10076
• fix: adding checks to ensure coin denoms are sorted in Balance during Validate #9829
• perf: store/cachekv: avoid a map lookup if unnecessary, clear maps fast, avoid keys sort #10486
• perf!: Add HasAccount to the AuthKeeper to save protobuf decoding time #10022

We are currently using a fork of the SDK that includes these fixes, but we'd like to use a stable release if possible. Also not sure if @odeke-em's perf improvements have been considered for this release

[tac0turtle]

• fix: enable setting sequence (nonce) for module account #10536 -> fix: enable setting sequence (nonce) for module account (backport #10536) #10846
• fix: nil pointer panic on NewIntFromBigInt #9627 -> fix: nil pointer panic on NewIntFromBigInt (backport #9627) #10845
• fix: populate ctx.ConsensusParams for begin blockers #10725
• feat: Add HasSupply method to bank keeper #10393

[ethanfrey]

So the only real blocker here is merging #10770 right?

I see #10281 which is important but don't see a pr for it on master

[robert-zaremba]

#10281 can come in a patch release.

[fedekunze]

are there any blockers now? I think we are ready with all the issues based on this list and the milestone?

@robert-zaremba @AmauryM @aaronc @marbar3778

[robert-zaremba]

@fedekunze we are testing the upgrade. Yesterday, @AmauryM made a test and it failed. If you have available cycles, could you also make an upgrade test on your side?

[robert-zaremba]

our qa-team did another test this morning and it worked, so we are trying to replicate.

[amaury1093]

My localnet repro is:

• on a v044 chain, create a proposal "abc" at height 50
• make some votes for it to pass
• wait for the node to panic at height 50
• ./build/simd start using the v045 node

At startup of the new node, I get:

3:15PM INF applying upgrade "abc" at height: 50
3:15PM ERR WARNING: Attempt to allocate proposer rewards to unknown proposer cosmosvalcons1ha3pfjev0q0e3j7k7yum0mu7catx28rh4hutfd. This should happen only if the proposer unbonded completely within a single block, which generally should not happen except in exceptional circumstances (or fuzz testing). We recommend you investigate immediately. module=x/distribution
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x28 pc=0x4ee02b1]

goroutine 1 [running]:
github.com/cosmos/cosmos-sdk/x/distribution/keeper.Keeper.AllocateTokensToValidator({{0x5857970, 0xc0010966a0}, {0x950a6d0, 0xc001097450}, {{0x950a6d0, 0xc001097450}, 0xc000c0cd18, {0x5857970, 0xc0010966d0}, {0x58579c0, ...}, ...}, ...}, ...)
	github.com/cosmos/cosmos-sdk/x/distribution/keeper/allocation.go:107 +0x51
github.com/cosmos/cosmos-sdk/x/distribution/keeper.Keeper.AllocateTokens({{0x5857970, 0xc0010966a0}, {0x950a6d0, 0xc001097450}, {{0x950a6d0, 0xc001097450}, 0xc000c0cd18, {0x5857970, 0xc0010966d0}, {0x58579c0, ...}, ...}, ...}, ...)
	github.com/cosmos/cosmos-sdk/x/distribution/keeper/allocation.go:95 +0x1111

Maybe I messed up something though

[amaury1093]

Since the qa team managed a successful v044->v045 upgrade, I just tagged a v0.45.0-rc1.

If you have available cycles, could you also make an upgrade test on your side?

echoing this, cc @ethanfrey and @ValarDragon too

[ethanfrey]

awesome work. thanks!

[tac0turtle]

closing as 0.45 is released

[ethanfrey]

Just released wasmd 0.22 using Cosmos SDK 0.45.0
Launching testnet tomorrow