[danmar]

No description provided.

[dmcppcheck]

AI review

I found no problems. When combineOperators() merges tokens, the PR now copies the whitespaceahead flag from the last merged token to the combined one. That covers every merge in the function: ..., float literals, exponents, compound assignments, &&, ||, ::, ->, <<, >>, <<=, >>=, ++ and --. With the PR, #x keeps the space after these operators, as gcc and clang do. The previous TODO_ASSERT_EQUALS in macro_c11_example4 is now a real assertion and passes. The change adds no cost to preprocessing and does not touch the API.

Testing

Build and test suites

• make -j8 testrunner simplecpp CXXOPTS=-Werror in pr/ built with no warnings.
• ./testrunner: 0 failed.
• python3 run-tests.py: 176 tests, 21 skipped, 15 todos, 0 failed.
• python3 -m pytest integration_test.py: 128 passed.
• ./selfcheck.sh: exit code 0.

My own tests
I wrote stringification tests in /work/scratch and ran them through pr/simplecpp, base/simplecpp, gcc -E -P and clang -E -P. They covered:

• s(a == b), s(a ==b), s(a<<= 2) and s(a >>== b)
• s(1.5 f), s(.5 x), s(1.e x), s(1e+5 x) and s(0x1p+1 x)
• s(x ... y), s(f(a ... )), s(a .. .b) and s(a+++ b)
• s(a:: b), s(a&& b) and s(a-> b)
• Operators followed by a comment, by a backslash-newline or by a newline inside the macro arguments
• Arguments built by ##, nested macro expansion (str(P(a ->) b)) and #__VA_ARGS__

Results

• The PR's output matches gcc and clang in all of these cases.
• Base lost the space in many of them, for example "a ==b", "1.5f", "x ...y", "a ->b" and "a ==b , c ->d".
• One mismatch is still there, and base gives the same result, so the PR did not cause it: s(1e+ 5) gives "1e+5", while gcc and clang give "1e+ 5".

---

Automated review of 8f6fc00 by Claude. The pull request was built and tested locally.

[danmar]

This fixed #754