[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20161203140209.GA20545@openwall.com>
Date: Sat, 3 Dec 2016 15:02:09 +0100
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: JohnTheRipper-bleeding-jumbo, OS X, and error during -test
magnum -
On Sat, Dec 03, 2016 at 02:52:40PM +0100, magnum wrote:
> On 2016-12-03 14:30, websiteaccess@...il.com wrote:
> >Benchmarking: gpg-opencl, OpenPGP / GnuPG Secret Key [SHA1 OpenCL]...
> >OpenCL CL_INVALID_VALUE error in opencl_gpg_fmt_plug.c:300 - Copy result
> >back
> >iMac-de-xxx:run xxx$
>
> I'm pretty sure it's caused by bug(s) in Apple's OpenCL runtime.
Maybe we should make OpenCL errors non-fatal for the "--test" runs,
letting further formats be tested anyway? This will require proper
deallocation of OpenCL resources on such errors, though.
I think it can be useful to quickly see that e.g. 2 out of 400 formats
failed, and 398 worked - just like we have for incorrect computation
errors. I understand we won't always have this anyway, since sometimes
a format will segfault or otherwise crash the process and we only have
one process for all of the tests currently, but for OpenCL errors like
the above, we probably can proceed with further tests.
Alexander
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.
