Pádraig Brady wrote:
This does add overhead though it's less of an issue at configure time and tests, and so probably not worth complicating with #ifdef lint as is done in coreutils and a few places in gnulib.
I didn't know that there were already workarounds for this problem with #ifdef lint. I prefer this sort of thing without #ifdef lint, so that the configure-time tests and the test cases also test that the relevant objects can be freed. Since these tests don't affect production runtime, and have only an insignificant effect on configure and testing runtime, I'm not worried about their runtime overhead.
