I can't think of any valid reason whatsoever for doing so. PIKE_DEBUG should be side-effect free, with the exception of Pike_fatal calls, or else it'll become useless as a debug tool.
I know there are several occasions where that principle is violated, for instance you'll get a more accurate __FILE__ if you compile pike with --with-rtldebug. I think that too is wrong.
Are there any objections to this principle regarding PIKE_DEBUG? If not we can settle on it and regard the offending places as simple bugs.
/ Martin Stjernholm, Roxen IS
Previous text:
2002-09-30 14:39: Subject: Gz testsuite memory usage
Um, program.c line 5374:
#ifdef PIKE_DEBUG if (resolve_cache) free_mapping(resolve_cache); #endif resolve_cache = c->resolve_cache_save;
Why is there a #ifdef PIKE_DEBUG around the freeing of the old cache?
/ Marcus Comstedt (ACROSS) (Hail Ilpalazzo!)