[PATCH] tests: remove seemingly unnecessary "sleep 3"

Jim Meyering jim at meyering.net
Wed Oct 6 15:45:25 UTC 2010


Jeff Darcy wrote:
> On 10/06/2010 09:44 AM, Jim Meyering wrote:
>> Could it have failed with ENOTEMPTY rather than EPERM?
>> It's not easy (maybe not even possible) to make "rm -rf" fail with
>> EPERM when all of the things being deleted were created by the same
>> user running the command.
>
> The error message is "operation not permitted" which maps to EPERM.
> ENOTEMPTY would make more sense to me too, it's pretty clearly the
> actual reason for the failure, but it's not what the logs were showing.

Wow.  Odd.  I really don't see how that could happen
unless some set-uid process created a directory not owned
by you in the hierarchy that rm -rf is trying to remove.
Oh.. one more possibility: something created a directory
unreadable by you.  Possible, if you have an overly restrictive umask.

>> If you can reproduce that, I'd be very interested in any
>> details you can provide, since init.sh is used in many projects,
>> and I'm about to make a coreutils release that includes it.
>
> I'll run it under strace and verify the actual errno.

How do you reproduce it?

>>> If/when I can test that version on my machine and not have it report
>>> errors after all tests have passed, then yes.  Also, whatever fix we
>>> decide on also needs to be applied to other tests.
>>
>> Of course.
>>
>> Here's a patch that does that:
>
> Looks fine, but I'd like a chance to test it at least a few times before
> I ACK and we both have a meeting right now.

What version of mongod are you using?
Here (F13), I have this:

    $ mongod --version
    db version v1.6.2, pdfile version 4.5
    Wed Oct  6 17:03:37 git version: nogitversion


More information about the iwhd-devel mailing list