On 18 March 2026 07:19:53 GMT, David Gow <[email protected]> wrote:
>Le 18/03/2026 à 12:32 AM, Josh Law a écrit :
>>
>>
>> On 17 March 2026 07:19:47 GMT, David Gow <[email protected]> wrote:
>>> Hi Josh,
>>>
>>> Le 16/03/2026 à 5:17 AM, Josh Law a écrit :
>>>> Bug fixes, extensions, and test coverage for lib/glob.c:
>>>>
>>>> Patches 1-3 fix corner cases in the existing glob_match():
>>>> - Inverted character class ranges like [z-a] are now normalized
>>>> instead of silently failing to match.
>>>> - A trailing backslash is treated as a literal '\' rather than
>>>> reading past the end of the pattern string.
>>>> - [^...] is accepted as an alias for [!...] to match the
>>>> regex-style negation syntax documented in glob(7).
>>>>
>>>> Patches 4-5 add two new utility functions:
>>>> - glob_match_nocase() for case-insensitive matching, useful for
>>>> subsystems like ATA that need case-folded denylist comparisons.
>>>> - glob_validate() for checking pattern syntax before use, so
>>>> callers can reject malformed patterns early with a clear error.
>>>>
>>>> Patches 6-7 add kunit test coverage:
>>>> - 47 new test cases for glob_match covering escapes, inverted
>>>> ranges, caret negation, edge cases, and unclosed brackets.
>>>> - 11 test cases for glob_match_nocase.
>>>> - 17 test cases for glob_validate.
>>>>
>>>> Patch 8 adds a real in-tree caller for glob_validate() in the kunit
>>>> executor, validating user-provided filter_glob patterns and returning
>>>> -EINVAL for malformed ones.
>>>>
>>>> Changes since v3:
>>>> - Resend: v3 accidentally included stale v2 patch files in the
>>>> same email thread. No code changes from v3.
>>>>
>>>> Changes since v2:
>>>> - Fixed missing '^' in __glob_match metacharacter comment (patch 4).
>>>> - Dropped unnecessary braces in glob_validate case '[' (patch 5).
>>>> - Added real-world example to glob_validate() commit message (patch 5).
>>>>
>>>> Changes since v1:
>>>> - Added patch 8 (kunit executor caller for glob_validate).
>>>> - Updated glob_match_nocase() commit message to reference the ATA
>>>> denylist as the intended caller (follow-up patch).
>>>>
>>>> Josh Law (8):
>>>> lib/glob: normalize inverted character class ranges
>>>> lib/glob: treat trailing backslash as literal character
>>>> lib/glob: accept [^...] as character class negation syntax
>>>> lib/glob: add case-insensitive glob_match_nocase()
>>>> lib/glob: add glob_validate() for pattern syntax checking
>>>> lib/tests: add glob test cases for escapes, edge cases, and new
>>>> features
>>>> lib/tests: add kunit tests for glob_match_nocase() and glob_validate()
>>>> kunit: validate glob filter patterns before use
>>>>
>>>> include/linux/glob.h | 2 +
>>>> lib/glob.c | 123 ++++++++++++++++++++++++++++++++++++----
>>>> lib/kunit/executor.c | 13 +++++
>>>> lib/tests/glob_kunit.c | 124 +++++++++++++++++++++++++++++++++++++++++
>>>> 4 files changed, 252 insertions(+), 10 deletions(-)
>>>>
>>>
>>> Thanks: always good to have better test coverage (and those are some nasty
>>> corner cases!)
>>>
>>> Series is:
>>>
>>> Acked-by: David Gow <[email protected]>
>>> Tested-by: David Gow <[email protected]>
>>>
>>> I'm making the assumption that this is going in via mm-nonmm, not the
>>> kselftest/kunit tree, as it has dependencies in that branch.
>>>
>>> Cheers,
>>> -- David
>>>
>>
>>
>> Also, David? You mean the *ENTIRE* series as in the code too (the new
>> functions) Or just your kunit changes
>>
>
>I only really need to Ack the KUnit changes, but I did test the entire series,
>so feel free to apply it to all of them.
>
>Cheers,
>-- David
>
Hello David, So the new functions are perfectly fine? I mean code review wise I
think they are, and I have tested them with a stash driver..
So I guess also
Tested-by: Josh Law <[email protected]>
