-
Notifications
You must be signed in to change notification settings - Fork 138
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
test: Add trace input via JSON #1265
Conversation
* Add other steps to read/write test * Improve error logging in trace handler test * Push ActionType handling into Unmarshal/Marshal and out of own class * Rename generic trace handler files to be clear they are for json * Rename to marshalling to be more generic * Add marshal/unmarshal for proposals * Export unexported field * Add test for marshal/unmarshalling chain state * Fix pointer issues * Fix typo: action -> proposal * Log proposal string in test * Use json.RawMessage instead of map[string]any * For uniformity, also use RawMessage for step unmarshalling * Add tests for extra proposal types * Add more proposal types to test and unify names * Add handling for ParamsProposal * Regenerate traces * Chore: Export forgotten field * Use string, not int, to help marshal/unmarshal
* Add other steps to read/write test * Improve error logging in trace handler test * Push ActionType handling into Unmarshal/Marshal and out of own class * Rename generic trace handler files to be clear they are for json * Rename to marshalling to be more generic * Add marshal/unmarshal for proposals * Export unexported field * Add test for marshal/unmarshalling chain state * Fix pointer issues * Fix typo: action -> proposal * Log proposal string in test * Use json.RawMessage instead of map[string]any * For uniformity, also use RawMessage for step unmarshalling * Add tests for extra proposal types * Add more proposal types to test and unify names * Add handling for ParamsProposal * Regenerate traces * Chore: Export forgotten field * Use string, not int, to help marshal/unmarshal * Add rapid to go.mod and .sum * Add rapidpbt for chainState marshalling * Rename file to make clear it only relates to chainState * Add error return to TraceWriter and TraceParser * gst * Add generators for actions and steps, utilize in test driver * Restrict range for time * Add test for time marshal/unmarshal * Make time have a lower bound, since negative numbers are not supported by RFC3339 * Improve label string for argument to time.Unix * Correct lower bound for time: 1900 years negative instead of 2000 * Convert timestamp to utc * Format file * Ignore testdata folder * Add go comment * Add docstring to GetTraceGen
) | ||
|
||
var selectedTestfiles TestSet | ||
|
||
var stepChoices = map[string]StepChoice{ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I separated step choices from test runs.
Why: When we input files, we need to give the test runs separately anyways. This makes that behaviour universal. Also seems more future proof, e.g. for running the same steps with different test runs/contexts
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
To a semi outsider like myself, the abstractions here are not super intuitive.
"steps" seems like a set of steps to make in running a test.
"test runs" seems synonymous to "steps"
@@ -66,15 +66,15 @@ var slashThrottleSteps = concatSteps( | |||
stepsStopChain("consu", 2), | |||
) | |||
|
|||
var democracySteps = concatSteps( | |||
var democracyRewardsSteps = concatSteps( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The naming was confusing before: democracySteps were the input for democracy-reward, and rewardDenomConsumerSteps were the input for democracy. This makes the mapping of test cases to step names more consistent.
@@ -56,3 +56,17 @@ jobs: | |||
go-version: "1.20" | |||
- name: E2E tests | |||
run: make test-e2e-short-cometmock | |||
Trace-Tests: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Add a test that reads traces to the CI to check that trace-reading does not break. Could be expanded in the future
@@ -0,0 +1 @@ | |||
testdata |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ignore temporary folder that rapid puts test failures in
I'm about to review, first impression of reading the description: this work all seems awesome. There's a lot of changes added in one PR tho. Is it possible to split out these individual components?
Totally understand if it's not worth the effort to split everything out at this point, just throwing out the idea |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Really cool stuff! My comments are mostly in regard to e2e test abstraction rather than functionality directly related to the PR title. Might be worth splitting things out if not too challenging
@@ -36,7 +36,7 @@ jobs: | |||
with: | |||
go-version: "1.20" # The Go version to download (if necessary) and use. | |||
- name: E2E changeover test | |||
run: go run ./tests/e2e/... --tc changeover | |||
run: go run ./tests/e2e/... --tc changeover::changeover |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
From description
for test-file, specifying the test runner separately is necessary
It seems to me like specifying testrunner is mostly relevant to the --test-file
flag. Is it neccessary to also require that callers specify the testrunner when using --tc
?
Asking because the cli for e2e tests is becoming quite verbose to run a simple test from the manual (golang defined) test cases
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Also we should document the required format for each flag in a readme if that's not already done
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Also we should document the required format for each flag in a readme if that's not already done
You can see it by running --help, do you think that is good enough?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
From description
for test-file, specifying the test runner separately is necessary
It seems to me like specifying testrunner is mostly relevant to the
--test-file
flag. Is it neccessary to also require that callers specify the testrunner when using--tc
?Asking because the cli for e2e tests is becoming quite verbose to run a simple test from the manual (golang defined) test cases
Okay, I can see that. I'm between making it optional and removing it for the test cases, and I think for now I will remove it for --tc. Can be added back later if we need it.
@@ -0,0 +1,485 @@ | |||
package main | |||
|
|||
import ( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Am I understanding correctly that this file serves to fuzz test the ser/deser of json specified e2e actions?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Exactly
|
||
// This needs to be adjusted manually when new actions are added and should | ||
// include generators for all actions that are mentioned in main.go/runStep. | ||
func GetActionGen() *rapid.Generator[any] { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Will prob need more context to properly review this file
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Currently splitting this out, then we can e.g. go over that part, or I'll write more explanation
"github.com/davecgh/go-spew/spew" | ||
) | ||
|
||
func TestProposalUnmarshal(t *testing.T) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is there a reason these tests don't use rapid like action_rapid_test.go
does?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes. I wanted to expliticly give a json string as an input, since this is hard to do with Rapid - we can get a json input by marshalling the rapid-generated objects, but this implies trusting that the marshal function does things correctly. this is able to test unmarshal and marshal in a slightly more isolated manner
case "main.submitConsumerAdditionProposalAction": | ||
var a submitConsumerAdditionProposalAction | ||
err := json.Unmarshal(rawAction, &a) | ||
if err != nil { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is it possible to have
if err != nil {
return nil, err
}
live outside the switch statement so as to not repeat code. Maybe just create a var err error
above the switch that can be set
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Good call, f5b0355
case "main.ConsumerAdditionProposal": | ||
prop := ConsumerAdditionProposal{} | ||
err := json.Unmarshal(inputMap, &prop) | ||
if err != nil { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Same as above with local err variable
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
) | ||
|
||
var selectedTestfiles TestSet | ||
|
||
var stepChoices = map[string]StepChoice{ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
To a semi outsider like myself, the abstractions here are not super intuitive.
"steps" seems like a set of steps to make in running a test.
"test runs" seems synonymous to "steps"
func getTestCases(selection TestSet) (tests []testRunWithSteps) { | ||
type testRunWithSteps struct { | ||
testRun TestRun | ||
steps []Step |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Related to above, it seems like "steps" should be a field of TestRun
, as in a test run is a high level abstraction with config + actual steps to be executed
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have renamed TestRun to TestConfig, since that is much more fitting to reality.
Good call, easy to get blind to these weird, historical naming choices. I've refactored this.
The abstractions are now:
A TestRun is made of
steps []Step
testConfig TestConfig
lmk if you still find it confusing
testRuns []string | ||
} | ||
|
||
type TestRunChoice struct { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
TestRun
already has a name
field right? Is this new type neccessary?
Co-authored-by: Shawn <[email protected]>
Thanks for the review, very useful stuff! I'll give splitting this as much as possible a try and apply your comments. |
Closing in favor of a feature branch https://github.com/cosmos/interchain-security/tree/feat/trace-format |
Description
Closes: #1240
Changes are as follows:
Main e2e test entry file: For the test main.go, change the behavior flags are processed:
Trace format:
interchain-security/tests/e2e/trace_handlers_test.go
Line 84 in e2175c2
interchain-security/tests/e2e/json_utils.go
Line 50 in e2175c2
interchain-security/tests/e2e/json_utils.go
Line 403 in e2175c2
Author Checklist
All items are required. Please add a note to the item if the item is not applicable and
please add links to any relevant follow up issues.
I have...
Reviewers Checklist
All items are required. Please add a note if the item is not applicable and please add
your handle next to the items reviewed if you only reviewed selected items.
I have...