Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 18 Sep 2015 18:03:44 +0200
From: Frank Dittrich <>
Subject: Re: multi-threaded hash table initialization

On 09/18/2015 05:45 PM, Solar Designer wrote:
> On Thu, Sep 17, 2015 at 12:53:23PM +0200, Frank Dittrich wrote:
>> It also causes a warning for non-OMP builds:
> I didn't expect magnum would commit this patch after I had called it in
> here a "failed experiment".  Now that it's in there, we should either
> revert it or fix it for non-OpenMP and benchmark on more systems.
> I see you're reporting issues with non-OpenMP builds here:
> but it makes no sense to try matching any non-OpenMP build issue to any
> other patch until the obvious bugs in the "multi-threaded hash table
> initialization" commit are fixed (or it is reverted).

For some strange reason, git bisect pointed to that unrelated commit
which must have exposed a problem with a previous commit.
For JomF, reverting a different unrelated commit makes the bug go away.

When I just revert 4c541b32c3a02a1fc4981879359307d273c9b9ca, I still
have all jtrts tests fail (with non-OMP).
After reverting 4c541b32c3a02a1fc4981879359307d273c9b9ca and
4e39b7e608e5bb09624c0b455561c73700e6f24f, the problem goes away.

If instead ov reverting both of these commits, I just revert
4e39b7e608e5bb09624c0b455561c73700e6f24f, the bug is still there.


Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.