revive.toml

# config for https://github.com/mgechev/revive
ignoreGeneratedHeader = false
severity = "error"
confidence = 0.8
errorCode = 1
warningCode = 0

[directive.specify-disable-reason]
    severity = "error"

#### roughly what golint does.  probably only disable noisy ones.

[rule.blank-imports]
[rule.context-as-argument]
[rule.context-keys-type]
[rule.dot-imports]
[rule.error-naming]
[rule.error-return]
[rule.error-strings]
[rule.errorf]
# [rule.exported] # disabled due to lack of value / encouraging bad habits
[rule.if-return]
[rule.increment-decrement]
[rule.indent-error-flow]
# Disabled because we have 158 packages that need package comments; we could instead add ignore
# directives for existing packages and require it for new packages.
#[rule.package-comments]
[rule.range]
[rule.receiver-naming]
[rule.time-naming]
[rule.unexported-return]
[rule.var-declaration]
[rule.var-naming]

#### higher value stuff

# this is basically errcheck, warns on errs that are not checked.
# strongly desired, but disabled due to 300 failures (to be tackled incrementally).
# [rule.unhandled-error]

# general defer gotchas.
#
# in particular: "recover" warns about unsafe use of recover().
# this has caught bugs that can allow crashes while seemingly safe, and are *extremely* hard to catch in review.
#
# the arguments are excluding only "call-chain", which would disallow `defer someFn(...)()` which is both useful and in use.
[rule.defer]
arguments=[["loop","method-call","recover","return", "immediate-recover"]]

# string(int) is almost always a bug.
# go vet considers this a fatal error, but only in 1.15 or newer, and go.mod currently targets 1.13
[rule.string-of-int]

#### added because we currently have zero violations, and they seem decent enough to retain

[rule.atomic] # correct use of sync code, important
[rule.call-to-gc] # beneficial
[rule.constant-logical-expr] # minor code simplifier
[rule.identical-branches] # code simplifier / failures are pretty dubious
[rule.modifies-parameter] # beneficial
[rule.modifies-value-receiver] # probably beneficial, prevents subtle bugs
[rule.range-val-address] # beneficial
[rule.range-val-in-closure] # beneficial
[rule.unconditional-recursion] # probably a good idea
[rule.unreachable-code] # code simplifier
[rule.waitgroup-by-value] # correct use of sync code, important

#### unused utilities

# [rule.file-header] # could possibly replace `copyright -verifyOnly`?
# [rule.imports-blacklist] # simple way to ban imports / enforce wrappers, likely useful

#### disabled but maybe desirable

# [rule.bare-return] # probably beneficial as it's slightly error-prone, but 2,000 failures
# [rule.bool-literal-in-expr] # minor code simplifier, few failures
# [rule.confusing-results] # maybe beneficial, only a few failures
# [rule.deep-exit] # probably a good idea in most code, some failures, but not trivial to adopt
# [rule.duplicated-imports] # minor, but may be worthwhile.  failures are weird but harmless
# [rule.early-return] # minor code simplifier, a handful of failures
# [rule.get-return] # existing failures are intentional + desirable, but in principle it's a fine idea
# [rule.import-shadowing] # probably beneficial, but 750 failures
# [rule.redefines-builtin-id] # probably beneficial, few failures
# [rule.struct-tag] # probably beneficial, a few failures
# [rule.superfluous-else] # minor code simplifier, a few failures
# [rule.unexported-naming] # probably beneficial, but 300 failures
# [rule.unused-parameter] # minor code simplifier / clarifier, but 250 failures
# [rule.unused-receiver] # minor code simplifier / clarifier, but 500 failures

#### probably undesirable

# [rule.add-constant] # extremely noisy.  18,000 failures, overwhelmingly for tests or 0/1 which seem totally fine
# [rule.argument-limit] # too arbitrary
# [rule.cognitive-complexity] # dubious value, but possibly interesting
# [rule.confusing-naming] # dubious value, ~50 failures
# [rule.cyclomatic] # dubious value, but possibly interesting
# [rule.empty-block] # easily noticed in code review, but also warns on documented no-op branches, which seem fine
# [rule.empty-lines] # low value, many failures
# [rule.flag-parameter] # interesting, but very noisy
# [rule.function-result-limit] # too arbitrary, easily noticed in code review
# [rule.line-length-limit] # too arbitrary
# [rule.max-public-structs] # too arbitrary
# [rule.unnecessary-stmt] # dubious value