[danmar]

'str(a/**/b)' gave "ab" instead of "a b". When comments are removed, mark the previous token as followed by whitespace.

[dmcppcheck]

AI review

The PR fixes the reported case, but it breaks stringification when a macro body ends with a comment. In removeComments() a trailing comment in a #define now sets whitespaceahead on the last token of the macro body. expandArg (simplecpp.cpp:2392) passes that flag on to the expansion, so # inserts a space that gcc and clang don't produce:

#define MAJOR 1/*major*/
#define MINOR 2/*minor*/
xstr(MAJOR.MINOR)

gives "1 .2" with the PR. Base, gcc and clang all give "1.2". The same bug already exists in base when there is a real space before the end of the body (#define E3(x) x /**/ gives "a b" where gcc gives "ab"). The PR spreads it to all trailing comments that have no space before them. A trailing comment in a #define is common code, so this is a real regression. Whitespace at the end of a replacement list should not count for stringification. No other problems found.

Testing

Build and test suites

• make -j8 testrunner simplecpp CXXOPTS=-Werror in pr/ built cleanly.
• ./testrunner: 0 failed.
• python3 run-tests.py: 176 tests, 0 failed.
• pytest integration_test.py: 128 passed.
• ./selfcheck.sh: exit 0.

I also built base/ and compared /work/{pr,base}/simplecpp with gcc -E -P and clang -E -P.

Cases that now match gcc (base got them wrong):

• str(a/**/b)
• a block comment that spans lines inside the argument
• str(a/**/+/**/b) gives "a + b"
• str("x"/**/'y')
• a comment inside the macro body: #define F(x) str(x/**/y)
• xstr(H()) with #define H() a/**/b
• xstr(E(a)/**/b)

These already matched gcc in base and still do: str(/**/a), str(a/**/), xstr(CAT(a,/**/b)).

Regression: macro body that ends with a comment

scratch/t4.c:

#define str(x) #x
#define xstr(x) str(x)
#define MAJOR 1/*major*/
#define MINOR 2/*minor*/
xstr(MAJOR.MINOR)

$ /work/pr/simplecpp t4.c    -> "1 .2"
$ /work/base/simplecpp t4.c  -> "1.2"
$ gcc -E -P t4.c             -> "1.2"   (clang: "1.2")

More cases (scratch/t3.c), with #define E(x) x/**/, #define E2 a/**/, #define P(x) (x)/**/:

input	PR	base	gcc / clang
xstr(E(a)b)	"a b"	"ab"	"ab"
xstr(E2;)	"a ;"	"a;"	"a;"
xstr(P(a)b)	"(a) b"	"(a)b"	"(a)b"

The same bug already exists in base with an explicit space: #define E3(x) x /**/ then xstr(E3(a)b) gives "a b" in both pr and base, and "ab" in gcc.

---

Automated review of 2e16061 by Claude. The pull request was built and tested locally.