chore: conform to the documented conventions - #486
Open
retr0h wants to merge 31 commits into
Open
Conversation
CONTRIBUTING requires the MIT header on every .go file. Four had none: three in the Docker container provider and one in telemetry's process conditions. Eight more carried the header but separated the copyright line from it with a bare // rather than a blank line, which is how the other twelve hundred files in the repository write it. They are the darwin files under the ping and netplan DNS providers. The logging rules ask for typed attributes; one call in the audit export command passed a positional pair, which compiles but bypasses the type safety the rule is there for. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The conventions ask for the multi-line form: one parameter per line, with the closing parenthesis and return types on a line of their own, so adding a parameter shows as one added line rather than a rewritten signature. Four hundred and eighty-eight declarations were written on one line. This rewrites them; no signature changes, only how it is spelled. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The public suites send raw HTTP through the Echo stack and then checked the recorder in the loop: a status code from one column and a list of substrings from another, iterated over. The row says what it expects of the response now, and the loop hands the recorder over. Two hundred and two tables across a hundred and two files. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The agent processors and the job client branched on an expectError column, compared the message against another column, and then called an optional validate the row might not supply. A row supplying none checked only that the result was non-nil. Each row states what it expects of the result and the error. Where the inlined body declared its own err for a later decode, that one is named decodeErr so it no longer shadows the callback's parameter. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The gate is stated as a percentage and the file declaring the target is named, but not the file narrowing what the percentage covers. .coverignore excludes cmd, gen, mocks, ui and main.go, so a reading taken without it does not match what the gate measures. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Thirty-nine files held tables whose loop finished by comparing a want column against the call. The comparison is the assertion, so it moves into the row. A nil written in one of those columns is emitted as the column's typed zero. testify compares types as well as values, and a bare nil against a *bool reads as unequal; two rows in the agent handler table showed that before the types were carried through. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The platform stubs assert that every Darwin method returns ErrUnsupported, which the rules ask for by name. The row said only its own name and the loop held the whole expectation; the row states it now. The CLI tables assert through the package-level assert with suite.T() rather than the suite's own methods, which the earlier pass did not read. Same conversion, same shape. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Twenty-nine more files of the same shape. Some assert only that the result is nil and the error is ErrUnsupported, without the separate Error call the earlier pass required, which is why they were left behind. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
validation's Struct and Var tables read an ok column and then, only when the call failed, iterated a contains column. Each row states both. The job subject label table branched on wantErr for the same two-line choice. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Twenty declarations of boolPtr, strPtr, intPtr and intervalPtr sat across sixteen files, all the same two lines, feeding three hundred and fifteen call sites. Generics made that one function years ago and the ecosystem settled on ptr.To, so the copies go and the calls read ptr.To(v). Seven support_public_test.go files held nothing else and are deleted rather than renamed. The remaining three are named for what they hold: sha256_public_test.go for the digest helper, test_agent_public_test.go for the agent builder its eight callers share, and the job client's mock setup folded into the one file that used it. File naming said a test file is named for the production file it tests, which says nothing about a test file holding no tests. That absence is what let support, fixture and mock_support all appear; the rule now covers it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Missed in the previous pass. It holds newTestAuditHandler, which three of the audit test files share. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Fifty more files. The loops branched on wantErr, expectError or expectErr and held both arms; each row now carries the arm it takes, with the row's own literals substituted in. The stub tables that assert Empty rather than Nil are converted too; the earlier pass only read the Nil form. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Some stubs assert Equal(0, got) rather than Nil or Empty; the matcher now keeps whatever the loop asserted about the result. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The general case: everything after the call is what the row expects, and the columns only that part reads are expectations rather than inputs. Those move into the row with their literals substituted; columns the setup also reads stay where they are. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Twenty-eight files carried the row's assertion in a column called validate. It is the same thing under a different name, so it takes the one the conventions use. Where the loop guarded the call, the guard is gone with it: a row that supplied no callback asserted nothing and passed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Thirty-seven tables declared one column, name, and held one row. The loop ran once and the row varied nothing, so the table said only what the sub-test name already said. The conventions ask for table-driven cases because rows are where success, errors and edge cases live. A table of one is ceremony around a single test, and bolting a callback onto it would only add more. The sub-test keeps its name and the loop goes. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
This reverts commit 662ca7e.
The loop was asserting on what the call returned, so the expectation sat in the loop rather than in the row it belonged to. The rows now carry a validateFunc and the loop only makes the call. Three files spelled the column validateFn or validateResult; the convention names it validateFunc. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Where the loop called and then asserted, the assertion belonged to the row: it is the row's expectation, not the loop's. The rows now carry a validateFunc holding what the loop used to assert, with the row's own values written in, so the guards that selected between them are gone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The loop served the request and then asserted the status the row had named. Serving is the call; the status is the row's expectation, so the recorder now goes to the row's validateFunc. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The same move as the previous commits, in the tables that reach for the assert and require packages rather than the suite's own methods. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
These rows named a whole struct or a built error as the thing they expected, so the expectation moves into the row's validateFunc with it. An error built only to have Error() called on it is now the string. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The validation loops chose which call to make from the row and then asserted for it; choosing stays in the loop, the assertion moves to the row. The job tables follow the same move. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
These loops assert and then tear down, so only the assertions move; the teardown stays. The config tables spelled the column wantCheck, which is the same idea under another name. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The converters treated a range clause as a call and a redeclared err as a borrowed variable, so these tables were passed over. They follow the same move as the rest. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Moving wantContains into the row flattened it: a row that named
[]string{`"error"`, `drain initiated`} came out asserting only that the
body contained error, without the quotes and without the second string.
Both are back, taken from the rows as they stood before the move.
Every raw string literal on main is again present in the file that held
it, and the only interpreted literals gone are three fmt imports the
simplification made unnecessary.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The middleware loops built a flag and a recorder and then asserted on them; both now go to the row. The processor tables carried an optional validate callback, which is the same thing under another name, so it is no longer optional and no longer needs the branch that guarded it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The remaining tables the converters could take: the agent constructor, the heartbeat time, the compact table's headers, and the readiness body. What is left asserts nothing about a result — it starts a server or a goroutine and lets the mock expectations do the checking. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Where the loop cannot give up the call — it needs the server again to stop it, or what it checks is a value the call left behind — the call stays and the variables go to the row instead. The signing table is written out by hand: its error row returned early, so the rows that do not error carry the envelope checks the return used to skip. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
The seed and template rows named what they expected and the loop read it back; the rows now hold the assertion and are handed the error, the result and the filesystem the loop built. HandleError does the same with the log it captured and the exit code it recorded. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
Two tables asserted that a call does not panic. The call is what the loop makes; not panicking is what the row expects, so the row takes the call and says so. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP
retr0h
force-pushed
the
chore/conform-to-contributing
branch
from
September 8, 2026 04:45
b91993f to
d15d50e
Compare
Codecov Report✅ All modified and coverable lines are covered by tests. @@ Coverage Diff @@
## main #486 +/- ##
==========================================
- Coverage 99.94% 99.94% -0.01%
==========================================
Files 484 484
Lines 22877 22876 -1
==========================================
- Hits 22865 22864 -1
Misses 12 12
... and 4 files with indirect coverage changes Continue to review full report in Codecov by Harness.
🚀 New features to boost your workflow:
|
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Brings the repository in line with
CONTRIBUTING.md, in one pass.What the audit found
build constraint from the header with a
//line instead of a blank one.parameter per line with the closing paren and return types on their own.
logger.Errorpassed a positional pair where theconventions ask for a typed
slog.Any.ptrTo-style helpers, with 315 callsites, duplicated
k8s.io/utils/ptr.To.support_public_test.gofiles held only thosehelpers and went away with them; the rest are named for what they hold.
row. They now carry
validateFunc: the loop makes the call, the row sayswhat it expects.
Tables
1,286 of 1,298 tables carry
validateFunc. The twelve that do not start aserver or a goroutine and return nothing — there is no result to hand over,
and the mock expectations do the checking. gohai and nats-server have the
same handful for the same reason.
One commit here reverts another:
test: drop the tables that had nothing to tableremoved 37 one-row tables, which no other repository in theorganisation does, so it is reverted rather than dropped from the history.
test: restore the assertions the handler tables had lostfixes a defect theconversion itself introduced: moving
wantContainsinto the row flattened[]string{"error",drain initiated}down to a single unquoteds.Contains(rec.Body.String(), "error"). Both the quotes and the secondstring are back, taken from the rows as they stood on
main. Every rawstring literal on
mainis present again in the file that held it, and theonly interpreted literals gone are three
fmtimports that a simplificationmade unnecessary.
Two notes
just readyfails onmainas well:go generateregenerates againstoapi-codegen v2.7.1, which the committed
*.gen.gofiles predate. Nottouched here.
go.modmoves more thank8s.io/utils. Every fmt, lint and coveragerecipe in the shared
gomodule runsgo get -tool, so the tool pinsfollow whatever was current when the recipes last ran.
Verification
just go-fmt,just go-vet(0 issues) andjust test(coverage 100.0%,target 99.9%) after every commit.
🤖 Generated with Claude Code
https://claude.ai/code/session_01CM5qrNtYQvdKcemkS5YpiP