Fix #15183 Syntax error with -!! - #9005

AI review

The fix works for ! and ~ after a unary +/-. It is incomplete, though: the same check in concatenateNegativeNumberAndAnyPositive() still reports a false syntaxError on valid code when another prefix operator follows the sign. Examples are return - --i;, return - ++i; and return +&x;. They are the same kind of bug as #15183 and fail the same way on base and on the PR. I found no regressions: invalid code such as return -!; is still rejected, now as internalAstError. return -~; is now accepted silently, but return ~; and return x + ~; are already accepted on base, so that is not new.

Testing

• I built cppcheck and testrunner in pr/, and cppcheck in base/.
• ./testrunner TestTokenizer: 392 tests, 0 failures.
• I compared base and PR on small C files of the form int x; int f(int i) { <stmt> }, run with cppcheck -q --enable=all.
• Fixed by the PR. Base reported syntaxError on each of these and the PR reports nothing wrong:
  • return -!!i;
  • return -~i;
  • return +!i;
  • return x * -!i;
  • return x == -!i;
  • return x < -~i;
  • return i ? -!i : -~i;
  • int a[] = {-!i, +~i};
  • case -!0:
  • return sizeof -!i;
• Still a false syntax error on valid code, the same on base and PR:

int x; int f(int i) { return - --i; }
int x; int f(int i) { return - ++i; }
int x; int f(int i) { return +&x; }

$ ./cppcheck -q u.c
u.c:1:23: error: syntax error [syntaxError]

• Invalid code:
  • return -!; now gives internalAstError; base gave syntaxError.
  • return -~; is now accepted without any error. Base also accepts return ~; and return x + ~;, so this is not a new kind of gap.

Automated review of a05990f by Claude. The pull request was built and tested locally.

AI review

I found no new problems. The PR now accepts !, ~, &, ++ and -- after a unary +/-. That fixes #15183 and the cases raised in the earlier review (- --i, - ++i, +&x), and the new tokenizer test covers them. Code that is still invalid, such as return -!;, return - ++; and return -!-;, is still rejected, now as internalAstError or syntaxError. One kind of invalid code is newly accepted without an error: return -&;. Base already accepts return &; and return x + &;, so this is the same known gap as -~; from the earlier review and not a new problem.

Earlier comments

Testing

• I built cppcheck and testrunner in pr/, and cppcheck in base/.
• ./testrunner TestTokenizer: 392 tests, 0 failures. ./testrunner TestGarbage TestSimplifyTokens TestSymbolDatabase TestValueFlow: 0 failures.
• I ran cppcheck -q --enable=all on 20 small files of the form int x; int f(int i) { <stmt> }.
• Valid code, no syntax error now:
  • -!!i, - --i, +&x
  • x * +~i, x * + !i
  • - - -i, x - - - !i, - *p, -&x == 0
  • -!i ? -~i : +!i, x = -!-!-~i
• Invalid code, compared with base:

Base already accepts the last three, so the PR does not create a new kind of gap.

• The GCC extension return -&&lbl; gives syntaxError on both base and the PR.

Automated review of f623f9a by Claude. The pull request was built and tested locally.