<- Merge/foundation release/1.10.1 resolved: cross-client tests and tests generators/fillers #353
Conversation
Date: 2021-03-31 06:45:50-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 06:57:24-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 07:08:19-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 07:08:40-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 07:15:05-05:00 Signed-off-by: meows <b5c6@protonmail.com>
…tests Stopping here because I'm realizing that its going to be more complicated than just writing one genesis+chain config for each Fork. Each StateTest defines its own genesis env, so that each test can have its own genesis. We have to write ONLY chain configs. Not genesis. Date: 2021-03-31 07:32:33-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Use only chain config values (not genesis). Date: 2021-03-31 07:39:58-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 07:41:20-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 07:56:14-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 07:57:03-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 08:28:27-05:00 Signed-off-by: meows <b5c6@protonmail.com>
… specs) Date: 2021-03-31 08:49:56-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 08:51:03-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 08:53:41-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 08:55:02-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 08:55:58-05:00 Signed-off-by: meows <b5c6@protonmail.com>
… chainspec'd difficulty tests Date: 2021-03-31 08:58:09-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 08:58:25-05:00 Signed-off-by: meows <b5c6@protonmail.com>
…gain Running difficulty tests yields missing gaslimit for genesiss type, which means the file only contains the chain config and not the genesis values. I assume this is because there were pre-existing files which were not overwritten with genesis information. Date: 2021-03-31 09:11:47-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 09:12:16-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 09:13:56-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 09:14:08-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 09:14:28-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 09:14:53-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 09:22:31-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-03-31 19:15:50-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 03:51:44-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 04:13:50-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 04:17:59-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 04:32:24-05:00 Signed-off-by: meows <b5c6@protonmail.com>
…e transations Notably these tests need to work for the matrix of all configuration schemas (eg. Multigeth) which is not always super pretty. The special case for EIP1283 is handled specific to the troublesome/indicative test for now; it could also be handled at the Configurator method level as a general case, but that's going too deep for me right now. Date: 2021-04-01 05:50:23-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 06:43:25-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 06:44:18-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 06:46:31-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 08:05:14-05:00 Signed-off-by: meows <b5c6@protonmail.com>
| @@ -102,15 +107,17 @@ tests-generate: tests-generate-state tests-generate-difficulty ## Generate all t | |||
| tests-generate-state: ## Generate state tests. | |||
| @echo "Generating state tests." | |||
| env COREGETH_TESTS_GENERATE_STATE_TESTS=on \ | |||
| env COREGETH_TESTS_CHAINCONFIG_FEATURE_EQUIVALENCE_COREGETH=on \ | |||
There was a problem hiding this comment.
COREGETH_TESTS_CHAINCONFIG_FEATURE_EQUIVALENCE_COREGETH forces all Forks values used by the tests to be confp.Converted on init to &coregeth.ChainConfig type values. These config values are dumped JSON encoded to an adjacent file.
No longer do clients' test runner have to somehow magically know what "Berlin" means; they can look it up and see exactly what config should be used.
| @@ -60,6 +60,8 @@ var ( | |||
| EIP145FBlock: big.NewInt(9573000), | |||
| EIP1014FBlock: big.NewInt(9573000), | |||
| EIP1052FBlock: big.NewInt(9573000), | |||
| // EIP1283FBlock: big.NewInt(9573000), | |||
| // PetersburgBlock: big.NewInt(9573000), | |||
There was a problem hiding this comment.
I've just added these as documentation to reflect the related Constantinople config; there were issues with EIP1234 that came up in other places and this was part of my investigation.
| return nil | ||
| } | ||
| } | ||
|
|
||
| c.ensureExistingRewardSchedule() | ||
| c.BlockRewardSchedule[*n] = vars.EIP649FBlockReward | ||
|
|
There was a problem hiding this comment.
These two changes are important. The prevent a configuration from overwriting an already-configured EIP1234 value with a same-block EIP649 configuration; essentially codifying that EIP649 can not be set after EIP1234.
Date: 2021-04-01 08:25:26-05:00 Signed-off-by: meows <b5c6@protonmail.com>
| @@ -313,6 +313,8 @@ func (c *ChainConfig) SetEIP1052Transition(n *uint64) error { | |||
| } | |||
|
|
|||
| func (c *ChainConfig) GetEIP1283Transition() *uint64 { | |||
| // This is special case handling for multigeth configuration. | |||
| // It makes the tests pass. | |||
| if c.ConstantinopleBlock != nil && c.PetersburgBlock != nil { | |||
| if c.ConstantinopleBlock.Cmp(c.PetersburgBlock) == 0 { | |||
| return nil | |||
There was a problem hiding this comment.
This is not an ideal thing, and it differs from the goethereum implementation. This is due, in part, to core-geth's original inheritance of the multi-geth configuration format and need to support some old strange versions of the multi-geth config.
There was a problem hiding this comment.
I looked at generalizing the interaction of EIP1283 and EIP1283Disable as Disable-type methods at the Configurator.(Equivalent|Compatible), but decided that was too involved for now; and this case currently only applies to 1283, being the only EIP which has been explicitly disabled.
| @@ -85,18 +85,17 @@ type stInfo struct { | |||
| LLLCVersion string `json:"lllcversion"` | |||
| Source string `json:"source"` | |||
| SourceHash string `json:"sourceHash"` | |||
| WrittenWith string `json:"written-with"` | |||
| Parent string `json:"parent"` | |||
| ParentSha1Sum string `json:"parentSha1Sum"` | |||
There was a problem hiding this comment.
These three fields were specific to core-geth test generation metadata. I've removed them since their information be inferred with version control, and I use the FilledWith field to define the core-geth version (instead of WrittenWith).
There was a problem hiding this comment.
Also Chainspecs is disused since the tests generation creates the GeneralStateTests_configs.json which defines the test configurations; making references to configs by file unnecessary.
| WrittenWith string `json:"written-with"` | ||
| Parent string `json:"parent"` | ||
| ParentSha1Sum string `json:"parentSha1Sum"` | ||
| Chainspecs chainspecRefsT `json:"chainspecs,omitempty"` | ||
| Labels map[string]interface{} `json:"labels,omitempty"` |
There was a problem hiding this comment.
Labels was apparently added at the tests suite.
Date: 2021-04-01 17:11:08-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 17:11:36-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 17:12:37-05:00 Signed-off-by: meows <b5c6@protonmail.com>
This reverts commit 977f5c3.
This reverts commit d74187e.
This reverts commit f8ee352.
Date: 2021-04-01 17:22:35-05:00 Signed-off-by: meows <b5c6@protonmail.com>
Date: 2021-04-01 17:33:45-05:00 Signed-off-by: meows <b5c6@protonmail.com>
meowsbits
deleted the
merge/foundation-release/1.10.1-resolved-writing-chainconfigs-regenerate-again
branch
This is a refactoring and improving of cross-client test generation.
Issues provoking this
deprecatedParity config-format support for EIPs installed after Istanbul.Does
tests/testdata/GeneralStateTests_configs.jsonas a map keyed on the subtest Fork names used in those tests, egBerlinandETC_Magneto.I'll make more notes inline with the code to point out important stuff.
See generated tests diff at etclabscore/tests#4