On Fri, May 17, 2019 at 9:13 AM David Howells [off-list ref] wrote:
Christian Brauner [off-list ref] wrote:
quoted
If you still prefer to have cloexec flags
for the 4 new syscalls then yes,
if they could at least all have the same name
(FSMOUNT_CLOEXEC?) that would be good.
They don't all have the same value (see OPEN_TREE_CLOEXEC).
Note that I also don't want to blindly #define them to O_CLOEXEC because it's
not necessarily the same value on all arches. Currently it can be 02000000,
010000000 or 0x400000 for instance, which means that if it's sharing a mask
with other flags, at least three bits have to be reserved for it or we have to
have arch-dependent bit juggling.
One thing I like about your approach of just making them O_CLOEXEC by default
and removing the constants is that it avoids this mess entirely.
+1.
Confusion caused by inconsistency of naming is going to hurt more than
inconsistency of semantics wrt. open(2).
Thanks,
Miklos