fix(dogstatsd): reject tag_length.end() <= MIN_TAG_LENGTH upfront#1875
Merged
Conversation
The dogstatsd `common::tags::Generator` wrapper subtracts one from
`tag_length.end()` to account for the on-wire `:` separator before
forwarding to the inner generator. The inner generator requires
`start() >= MIN_TAG_LENGTH = 3`. When the caller's `tag_length.end()`
equalled `MIN_TAG_LENGTH` -- e.g. `Constant(3)` or `Inclusive { 3, 3 }` --
the wrapper produced `Inclusive { 3, 2 }` and the inner rejected it on
`min <= max`. That rejection surfaced upstream as `Error::StringGenerate`,
which dropped the underlying message and pointed callers at pool
generation rather than the config range that was actually at fault.
Validate the constraint at the wrapper entrypoint and return a dedicated
`Error::TagLengthEndTooSmall` variant naming the offending value and the
minimum required. Tests cover both the rejection cases and the smallest
accepted `end` value.
blt
force-pushed
the
blt/fix-dogstatsd-tag-length-wrapper-underflow
branch
from
a151d21 to
8f2643f
Compare
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8f2643f332
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
The dedicated Error::TagLengthEndTooSmall added by the parent commit was
still swallowed at the public DogStatsD::new boundary: MemberGenerator::new
mapped every tags::Generator::new failure to crate::Error::StringGenerate,
so a tag_length of Constant(3) or Inclusive { min: 3, max: 3 } continued to
reach external callers as the misleading StringGenerate.
Match the TagLengthEndTooSmall variant explicitly and convert it to
crate::Error::Validation, preserving its message. Other tag-generator
failures keep the existing StringGenerate behaviour. Adds a regression test
exercising the public path.
Addresses PR review feedback.
The three tag_length boundary tests repeated the same 7-arg tags::Generator::new call, varying only the tag_length range. Replace the small_string_pool helper with generator_with_tag_length, which holds every other argument fixed and returns the construction result, so each test states only the range it exercises.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3a38d74737
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
…subtract
The upfront check only rejected tag_length.end() <= MIN_TAG_LENGTH, but the
wrapper subtracts one from end (not start) when forwarding to the inner
generator. Any range with start == end above the minimum -- Constant(4),
Inclusive { min: 100, max: 100 }, etc. -- therefore still produced an adjusted
Inclusive { min: N, max: N - 1 }, which the inner generator rejected as
InvalidConstruction and MemberGenerator mapped to the misleading StringGenerate.
Generalize the check to reject when end is not strictly greater than both
start and MIN_TAG_LENGTH, renaming the variant to TagLengthRangeTooNarrow to
reflect the broader constraint. Extend the unit tests with the constant-above-min
and single-value cases and make the public-path test table-driven over the
collapsing configs.
Addresses PR review feedback.
blt
force-pushed
the
blt/fix-dogstatsd-tag-length-wrapper-underflow
branch
from
45fb439 to
509221f
Compare
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
2 participants
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.
Summary
The dogstatsd
common::tags::Generatorwrapper reserves one byte of the on-wiretag_lengthfor the:separator before forwarding to the innercommon::tags::Generator, by buildingInclusive { min: start, max: end - 1 }. The inner generator requires a non-empty range whosestart() >= MIN_TAG_LENGTH = 3.Because only
maxis decremented (notmin), any range wherestart == endcollapses — the adjusted range becomesInclusive { N, N - 1 }(min > max). That includesConfRange::Constant(N)for anyN, andInclusive { min: N, max: N }. The inner generator then rejected the range, and the failure bubbled up asError::StringGenerate, which drops the message and points callers at pool generation rather than at the tag-length range that was actually at fault.This change:
tag_lengthis rejected whenendis not strictly greater than bothstartandMIN_TAG_LENGTH, via a dedicatedError::TagLengthRangeTooNarrowvariant that names the offending range. This catches every collapsing case — values at/below the minimum (Constant(3),Inclusive { 3, 3 }) and constant/single-value ranges above the minimum (Constant(4),Inclusive { 100, 100 }).DogStatsD::newboundary ascrate::Error::Validation(message intact), instead of the previous catch-allStringGenerate. Other tag-generator failures keep the existingStringGeneratebehaviour.Generation semantics are intentionally unchanged (no fingerprint churn): collapsing ranges are rejected with a clear error rather than being silently reinterpreted.
Discovered while fuzzing
DogStatsD::newfrom an external rig: the symptom was aStringGeneratefailure at specific seeds, with no indication that the root cause was the configured tag-length range collapsing.Test plan
cargo test -p lading-payload— passes, including:tag_length_constant_at_min_rejected,tag_length_inclusive_min_equals_max_at_min_rejected,tag_length_constant_above_min_rejected,tag_length_inclusive_single_value_above_min_rejected;tag_length_at_min_plus_one_accepted;collapsing_tag_length_surfaces_as_validation_error.cargo clippy -p lading-payload --all-targets— clean.cargo fmt --check— clean.Unreleased.