On 2018/10/18 12:37, Amir Goldstein Wrote:
On Thu, Oct 18, 2018 at 5:37 AM zhangyi (F) <yi.zhang@xxxxxxxxxx> wrote:
On 2018/10/16 17:45, Amir Goldstein Wrote:
On Tue, Oct 16, 2018 at 10:32 AM zhangyi (F) <yi.zhang@xxxxxxxxxx> wrote:
[...]
Please consider the name _overlay_repair_dirs() with reference to
_repair_scratch_fs(), which is used for roughly the same purpose.
_run_check_fsck() is helper used to test fsck and may expect to return
"error" exit code when we do exception tests(see patch 4), but
_repair_scratch_fs() always want to return "correct" exit code.
Right. so perhaps we need _overlay_repair_dirs() as a convenience
helper for things like the stress test and also related to another
comment about expecting return 0 is too fragile.
BTW, the tests generic/330 generic/332 when run with -overlay
over xfs with reflink support seem to call _repair_scratch_fs() which does:
# Let's hope fsck -y suffices...
fsck -t $FSTYP -y $SCRATCH_DEV 2>&1
local res=$?
case $res in
0|1|2)
res=0
;;
*)
_dump_err2 "fsck.$FSTYP failed, err=$res"
How come fsck.overlay is not complaining about missing arguments??
The rest of the generic tests that call _repair_scratch_fs() have
_require_dm_target() which _require_block_device(), so don't run with overlay.
If you do sort this out and add overlay support to
_repair_scratch_fs() its probably
worth replacing 0|1|2) with FSCK_ constants.
Thanks for pointing this out, I think we do something like below can fix
this problem, what do you think?
I am puzzled about why those tests do NOT fail when fsck.overlay is given
incorrect args??
Oh, it's _dump_err2() helper's fault, it should be
--- a/common/rc
+++ b/common/rc
@@ -98,7 +98,7 @@ _dump_err_cont()
_dump_err2()
{
_err_msg="$*"
- >2& echo "$_err_msg"
+ >&2 echo "$_err_msg"
}
Thanks,
Yi.
