Repository navigation
Conversation
…FFFF Since the refactor in owasp-modsecurity#3284 (2.9.9) the output pointer always advanced by six bytes after sprintf("%%u%04x"), so a code point that needs five hex digits (anything above U+FFFF) lost its last digit to the next write or to the terminating NUL: U+1F600 came out as "%u1f60". Advance by the number of bytes sprintf actually wrote. While there, fix two adjacent checks in the same function: bytes_left was never decremented, so the "enough bytes left" guards compared the total length instead of the remaining length and a truncated sequence at the end of the value was examined past input_len; and an 0xF4 lead byte with a second byte >= 0x90 (U+110000 and above, outside Unicode) was accepted as a valid four-byte sequence. Add four-byte, above-U+10FFFF and truncated-sequence cases to the utf8toUnicode unit tests, which had no four-byte input at all. Fixes owasp-modsecurity#3684 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Cq9GqRpB4wzTFnEUvQaUJg
Sonar flagged the sprintf call added by the previous commit. The output buffer is sized for the worst case, so this only makes the bound explicit. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Mhspv9mxtMSiWC1U8mVH1H
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configuration
You can disable this status message by setting the Use the checkbox below for a quick retry:
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
…e_ex The two-, three- and four-byte branches repeated the same pattern: check the byte count, check each continuation byte for binary 10, shift in six bits. Move that into utf8_decode_sequence(), which derives the sequence length from the lead byte and loops over the continuation bytes. The NUL special case and the surrogate, range and overlong checks stay in the caller, so the output is unchanged. SonarCloud reported a cognitive complexity of 51 for the function on PR owasp-modsecurity#3685 (accepted there); it is now 17, with 10 for the helper. Verified output-identical to the previous version with a standalone harness under ASan and UBSan: every input of 1 to 3 bytes, every 4-byte input with a high lead byte and a third of the fourth-byte values, and 20 million random 4 to 15 byte inputs biased to high bytes and NUL: 758,263,296 cases, 0 mismatches. tests/tfn/utf8toUnicode.t passes 11/11. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Tz1sULBvqAwnRreKh5eC9h
fzipi
force-pushed
the
v2/refactor-utf8-decoder
branch
from
October 15, 2026 17:18
c96990a to
04b4e47
Compare
|
This branch has not been deployed
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.
Proposed changes
Follow-up to #3685, where SonarCloud flagged
utf8_unicode_inplace_ex()inapache2/msc_util.cfor a cognitive complexity of 51 (limit 25). That was accepted on #3685 because it is a bug fix; this PR does the refactor separately so it can be reviewed as one.The two-, three- and four-byte branches of the decoder repeated the same pattern: check the byte count, check each continuation byte for binary
10, shift in six bits. They move into a small static helper,utf8_decode_sequence(), which derives the sequence length from the lead byte and loops over the continuation bytes. The NUL special case and the surrogate, range and overlong checks stay in the caller. The error codes the old branches distinguished are never observed: the caller only testsunicode_len > 0, and theUNICODE_ERROR_*switch inre_operators.cbelongs to the separate decoder behind@validateUtf8Encoding.Cognitive complexity is now 17 for the function and 10 for the helper. The output is unchanged.
This PR is based on #3685's branch and shows its two commits until that one is merged; only the last commit is the refactor.
Tests
tests/tfn/utf8toUnicode.t, including the four cases added by fix: t:utf8toUnicode drops the last hex digit of code points above U+FFFF #3685: 11 of 11 pass (Ubuntu 24.04 container, Apache 2.4, PCRE2,--enable-assertions). The only failure inmake checkthere is the pre-existingop "ipMatch"assertion on10.0.0.0/100, unrelated to this file.changedflag: every input of 1 to 3 bytes, every 4-byte input with a high lead byte and a third of the fourth-byte values, and 20 million random 4 to 15 byte inputs biased to high bytes and NUL. 758,263,296 cases, 0 mismatches, no sanitizer reports.🤖 Generated with Claude Code