[SER-1312] Improve crash grouping for partially symbolicated traces#7711
Open
coskunaydinoglu wants to merge 6 commits into
Open
[SER-1312] Improve crash grouping for partially symbolicated traces#7711coskunaydinoglu wants to merge 6 commits into
coskunaydinoglu wants to merge 6 commits into
Conversation
…essing - Remove symbol/frame offsets as one unit before the standalone-number rule, so partially symbolicated iOS frames normalize cleanly - Remove bare hex sequences (4+ chars, digit required) not covered by the 0x rule, e.g. load addresses and <...> address tokens - Fix custom smart_regexes: flags were passed as the replacement string, so only the first match was replaced (case-sensitively) with the literal text "gim"; now all matches are removed with gim flags as the config help text documents - Collapse whitespace runs and trim as the final smart-preprocessing step so column alignment differences no longer change the seed Crashes that differ only by memory addresses, frame offsets, or padding now produce identical grouping seeds (SER-1312, reported for partially symbolicated iOS SIGTRAP crashes).
Unit tests call groupStrategy() directly with explicit override settings: iOS SIGTRAP pair equality, bare-hex normalization, offset removal as a unit, distinct crashes staying distinct, smart_regexes all-match case-insensitive removal without gim injection, invalid regex resilience, smart_preprocessing=false passthrough, stacktrace strategy, and standalone-id regression. Plus an e2e smoke test that two partially symbolicated crashes land in one crash group.
Crash ingestion is fire-and-forget: /i returns 200 before the crash group upsert lands, so asserting immediately after submission is flaky on slow runners. Wait 100*testScalingFactor ms after each submission (suite convention) and poll the crashes query until the group count is stable before asserting. Also URL-encode the crash payloads so the literal '+ 50648' frame offset reaches the server instead of decoding to a space, exercising the offset-removal rule as intended. The strict lengthOf(1) assertion is unchanged.
Contributor
Author
|
CI status summary for reviewers: ✅ 9/10 checks green on ❌ Note on the earlier red runs (21:38–23:00 UTC June 11): the same code failed The two Generated by Claude Code |
ar2rsawseen
approved these changes
Jun 12, 2026
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
3 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.
Jira: https://countly.atlassian.net/browse/SER-1312
Summary
Customers (reported by Vakifbank) saw identical partially symbolicated iOS SIGTRAP crashes split into many separate crash groups. The grouping seed produced by
groupStrategy()inplugins/crashes/api/parts/stacktrace.jsretained per-report noise — frame offsets (+ 50648), bare hex load addresses without a0xprefix (104a30000,<0000000104a3c5d8>), and column-alignment whitespace — so semantically identical traces hashed to different groups.This change, gated behind the existing
smart_preprocessingsetting (no behavior change when it is disabled):+ 600,+0x1a4) before the standalone-number rule0x/UUID rulesBug fix: custom
smart_regexesThe custom-regex loop passed the flags string as the replacement:
seed.replace(new RegExp(reg), "gim"). Only the first match was replaced, case-sensitively, with the literal textgiminjected into the seed. Now:seed.replace(new RegExp(reg, "gim"), "")— all matches removed, exactly as the setting's help text has always documented.Release-notes note: one-time regrouping
For apps with
smart_preprocessingenabled, grouping seeds (and thus hashes) change. New incoming crashes may open new groups instead of joining pre-upgrade groups. Existing groups remain intact and readable; they simply stop accruing new occurrences for affected signatures. No migration or merge of historical groups is performed.Test plan
plugins/crashes/tests.js(groupStrategycalled directly with explicit override settings): iOS SIGTRAP pair equality, bare-hex normalization, offset-as-unit removal, distinct crashes staying distinct, smart_regexes all-match/case-insensitive removal with nogiminjection, invalid-regex resilience,smart_preprocessing: falsebyte-identical passthrough,stacktracestrategy, standalone-id regression — all 9 verified locally against the real module (output in commit d561055's review)crashesplugin suite: runs in CI (no test environment in the authoring checkout)https://claude.ai/code/session_01KV52CRKaxGSbBLugwh7XrJ
Generated by Claude Code