Implement mocknet and CSV output of HF21 transition #1800
+3,450
−394
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Doy-lee
force-pushed
the
doyle-mocknet-rebased
branch
2 times, most recently
from
860087b to
da5f3e8
Compare
tewinget
reviewed
Mar 5, 2025
|
|
||
| for (auto& [addr, amt] : rewards) { | ||
| auto address_str = "0x{:x}"_format(addr); | ||
| auto amount = static_cast<int64_t>(amt * BATCH_REWARD_FACTOR); |
There was a problem hiding this comment.
Should be covered in #1802 but github won't let me comment here in that PR. I assume that this amt * BATCH_REWARD_FACTOR should be converted to the "db vs actual" coin functions?
There was a problem hiding this comment.
Yep. Just separated them because that change could be completely isolated out and that this PR focuses on just getting mocknet in.
tewinget
reviewed
Mar 5, 2025
tewinget
approved these changes
Mar 5, 2025
Boost 1.87 removed many things which were previously only deprecated, some of which we were still using.
changing from including boost asio io_service to io_context is roughly equivalent, but the former has a typedef for compatibility. I replicated the typedef, but gated it on the boost version that removed the io_service include.
the bulk of the code for transitioning to HF21 is present, and much testing code is present. This code does not yet work, and the tests are not yet complete. add hf21 transition mechanism to sqlite db
After HF21, any nodes which did not transition to SENT stakes need to be removed. Doing them all at once would be chaotic with swarms, so they will be removed by the existing decommission/deregister system. That system now has a config-determined limit of deactivations per block so that this transition (but also huge network outages) won't cause swarms to disappear (or other possible chaos resulting from mass removal).
Staked oxen addresses which are not registered for the SENT transition will have their pending service node rewards paid out in full on the final block before HF21. Modified local devnet testing to leave a staked static-startup address out of the registration for testing this (and adjusted expected numbers accordingly).
nodes which did not transition to hf21 will be removed by the decommission->deregister mechanism rather than by the bls-key-not-in-the-contract mechanism
script needs parameters set as appropriate: - oxen -> eth address mapping - transition bonus mapping - staking requirement and scaling ratio TODO: some mainnet nodes have a larger stake, scaling needs to be normalized for those
They cannot ever come back, so there's really no point of leaving them for a week. (We *don't* want to immediately dereg them because going into decomm still allows SN comms between storage servers, allowing the leaving nodes to still transfer their files out to the remaining nodes upon swarm dissolution).
nodes which do not transition at hardfork 21 should not be in the reward payment rotation
calling this method returns a before and after as if hf21 would happen the very next block. Because outstanding oxen rewards will differ from the exact fork height (and from block to block), this can only be approximate, but it should be fairly close to what the actual result will be.
- Windows.h is only needed if we don't have readline - Use WIN32_LEAN_AND_MEAN to #define out a bunch of headers you don't need
- Address some issues with the network not correctly transitioning to mocknet when a height is given that has already passed. - Fix block leader not being set in the cache in certain scenarios causing bad blocks to be generated - Ensure that the Pulse quorum is hijacked at the leaf-most function for generating a quorum to ensure all code paths requesting a Pulse quorum correctly get hijacked with mock nodes - Simplify the mocknet interface by dropping the post CN block add hook and instead kill P2P comms when a mock pulse block is being constructed
This lets the caller control the final values that the transition will run with. For ordinary usecases this is not important, the transition parameters are pulled from the compile time constants. For testing and mocking it can be useful to modify these values at runtime to simulate different transition conditions.
For mocknet and testing we now allow changing the staking requirement at runtime (i.e. for testing different staking requirements and conversion ratios) the staking ratio needs to be updated with the appropriate values. Allow a higher precision conversion ratio to allow for more fine grained conversions. For the $OXEN to $SESH conversion we now defer to a 128 bit multiply to avoid overflow for the conversion.
I was counting the number SESH calculated for each contributor as the amount they contributed to the node. This was incorrect because that number is not the amount they were able to stake, just the amount the were required to have to be able to keep the node alive. This mean that `tokens_allocated >= staking_requirement` was incorrectly attributing a node as having made it past the transition when it actually didn't. Instead of trying to be clever and calculating the partial collaterised nodes that will be zombified, only count tokens from a fully collateralised node.
```diff
std::bitset<sizeof(uint16_t) * 8> bitset_view16(uint16_t val) {
- std::bitset<sizeof(uint16_t) * 8> result = val;
+ std::bitset<sizeof(uint16_t)* 8> result = val;
return result;
}
```
clang-format constantly tries to re-format this line, moving the '*'
back and forth all the time. Seems to be a bug that we can avoid
completely by just pulling out the expression into a constexpr value.
Doy-lee
force-pushed
the
doyle-mocknet-rebased
branch
from
cfe33df to
118ae8f
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Goes ontop of https://github.com/oxen-io/oxen-core/pull/1790/files
I did have to modify the transition code a little bit to extract out information for producing the reports. The notable changes:
-D WITH_MOCKNET=ON. The intention is that in our release and packagable builds everything mocknet should be completely compiled out and there's no way of getting it to activated at all in public builds.Most of the implementation is compartmentalised into
src/network_config/mocknet.{h,cpp}and compiles to no-op macros when disabled.Some cursory header cleanup, removing unused headers, moving things to implementation file if it can be, removing dead/unused macros e.t.c.
Updating the conversion_ratio to use 64 bits, the numbers we were playing were fractional and we have a
mul128_divprimitive in the codebase that seemed fine to use for this purpose.Stack trace support is not a build flag enabled with
-D WITH_STACKTRACE=ONFix 0 OXEN payout generating an output in the miner TX on transition of the amount of rewards owed is less than 1000 DB OXEN units.
Enabling mocknet can be done by specifying the flag
--fork-to-mocknet <height>at which when the chain height reaches<height>(e.g.top_block.get_height() + 1 == <height>) then mock nodes will be injected into the SNL and all the quorums from that point will be hijacked by those nodes and it will periodically generate blocks to the chain.When this happens, pulse timestamps are disabled and blocks are appended to the chain every 5s.