stillwater-sc / universal / 30597794219

85%

master: 84%

Build Type

push

github

Committed by web-flow

Commit Message

fix(numeric): decimal integer conversion is UB (and hangs) for LLONG_MIN (#1273) (#1275)

number/support/decimal.hpp negated a signed long long to get its magnitude, which
is undefined behavior for LLONG_MIN (magnitude 2^63 is not representable in
long long):

- convert_to_decimal: `v *= -1` left v == LLONG_MIN (still negative); the bit loop
  then did `v >>= 1` on a negative value, which sticks at -1 forever -> INFINITE
  LOOP (a hang, confirmed).
- setvalue: `-v` is the same overflow UB (its fixed 64-iteration loop masked the
  symptom, but the negation is still UB and UBSan-flagged).

Negate in the unsigned domain instead, which is well-defined and yields 2^63 for
LLONG_MIN. convert_to_decimal now loops over a uint64_t magnitude (which also
retires the -Wsign-conversion cast from #1270 at that line).

Regression: static/conversions/decimal_boundary.cpp checks convert_to_decimal and
setvalue for LLONG_MIN / LLONG_MIN+1 / LLONG_MAX / 0 / +-small -- it hangs on the
old code and PASSes on the fix.

Verified: correct decimal for LLONG_MIN, terminates, and UBSan-clean (no
signed-overflow). gcc + clang: conversions_decimal_boundary, bt_decimal PASS.

Resolves #1273
Relates to #1270

Coverage Stats

6 of 6 new or added lines in 1 file covered. (100.0%)

8 existing lines in 1 file now uncovered.

41671 of 48905 relevant lines covered (85.21%)

7268704.24 hits per line