mirror of
https://github.com/github/codeql.git
synced 2026-07-31 23:42:59 +02:00
bfa65301f1
RegexTreeView.qll - hasLocationInfo fix (RULE 4):
- Add regexpContentOffset(RegExp re) private helper that computes the correct
number of source chars before the first content char, from getValueText():
- Plain "...": offset 1
- L"...": offset 2 (L" prefix)
- u8"...": offset 3 (u8" prefix)
- R"(...)": offset 3 (R"( opener)
- R"x(...)x": offset 4 (R"x( with custom delim)
- LR"(...)": offset 4 (LR"()
- Uses vt.matches("%R\"%(%") to detect raw strings; finds '(' position
- For non-raw, finds '"' position
- Replace hardcoded `+ 1` with `+ regexpContentOffset(re)` in hasLocationInfo
- Update comment: documents plain/encoding/raw/combined forms, notes that
escaped non-raw strings are approximate (mirroring Java/Python)
test.cpp:
- Fix r_uni: change "\\u{9879}" to "\\u9879" — braced \u{...} is not standard
C++ regex syntax; the 4-digit \uHHHH form is valid ECMAScript \u escape
Regenerated all 3 .expected files via codeql test run --learn (CodeQL 2.26.1).
All 3 tests pass.
Location test confirms:
- Plain "a+b" (line 144): litStartCol 22 → termStartCol 23 (22+1) UNCHANGED ✓
- Raw R"(a+b)" (line 147): litStartCol 20 → termStartCol 23 (20+3) FIXED ✓
- Raw R"x(a+b)x" (line 156): litStartCol 21 → termStartCol 25 (21+4) FIXED ✓
- L"a+b" (line 159): litStartCol 22 → termStartCol 24 (22+2) FIXED ✓
- LR"(a+b)" (line 162): litStartCol 26 → termStartCol 30 (26+4) FIXED ✓