Eric Blake wrote: > According to Eric Blake on 11/13/2009 11:35 AM: >> I also discovered that >> FreeBSD getgroups(-1,NULL) fails with EFAULT instead of EINVAL, which >> means that it mistakenly populates a valid pointer rather than rejecting a >> negative argument; should be easy to fix today. > > Like so. And I feel silly for the buffer overrun in the test; I was > originally using xnmalloc, but didn't want drag in "xalloc.h" for the > test, but converted it wrong. ... > Subject: [PATCH] getgroups: work around FreeBSD bug > > FreeBSD 7.2 mistakenly succeeds on getgroups(-1,ptr) (POSIX > requires EINVAL failure since -1 is less than the proper result). > > * lib/getgroups.c (rpl_getgroups): Work around the bug. > * m4/getgroups.m4 (gl_FUNC_GETGROUPS): Detect the bug. > * doc/posix-functions/getgroups.texi (getgroups): Document it. > * tests/test-getgroups.c (main): Fix buffer overrun.
The patch looks fine. For some reason (probably due to the dichotomy of success/failure in shell vs. true/false in C) I alway have to stare hard at code like this to be sure the returned expression is not reversed: + [AC_RUN_IFELSE([AC_LANG_PROGRAM([AC_INCLUDES_DEFAULT], + [[int size = getgroups (0, 0); + gid_t *list = malloc (size * sizeof *list); + return getgroups (-1, list) != -1;]])], + [gl_cv_func_getgroups_works=yes], + [gl_cv_func_getgroups_works=no], + [gl_cv_func_getgroups_works="guessing no"])])
