On Sat, 14 Jul 2007, Andrey Chernov wrote: *snip* > BTW, I fear about another place on this matter, merge_environ(). > Look at this example: > > ... some setenv's called before > > nenv[0] = NULL; > environ = nenv; > unsetenv("NONEXISTEN"); > > What happens to environ pointer _after_ that? > > unsetenv() leads to environ = watchEnviron; in the merge_environ() but > this environ is never rebuilded because no single setenv() is called > from merge_environ() and no rebuild_environ() is called from > unsetenv() due to "NONEXISTENT" value not found. It means reassigned > environ inherits stale pointers from non-rebuilded watchEnviron which > not equals to environ[0] == NULL > > Did I miss something? No. I see it too. Here are the changes: 1. __merge_environ no longer touches environ. Only __rebuild_environ(), __build_env() and __clean_env() alter environ. 2. watchEnviron renamed to intEnviron. 3. __rebuild_environ() uses intEnviron when cycling through the environment; it only sets environ. This way environ is not changed unless it has at least one name-value pair. The new patch is here: http://www.farley.org/freebsd/tmp/setenv/clearenv/patch2 Sean -- scf_at_FreeBSD.orgReceived on Sat Jul 14 2007 - 13:25:21 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:14 UTC