Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 2 Jan 2014 16:23:07 +0100
From: Frank Dittrich <frank_dittrich@...mail.com>
To: john-dev@...ts.openwall.com
Subject: Problem loading dynamic formats with TestD=... lines (was: Segfault
 with linux-x86-64-clang-debug)

>> See attached file backtrace.txt
> 
> Ok... Not sure what to make out of that. It's dynamic crashing, not crypt.

Yes, any command which causes dynamic formats being loaded causes the
same segfault with linux-x86-64-clang-debug.

I tracked this down to the
TestD=$dynamic20nn$ lines in run/dynamic_flat_sse_formats.conf

For some reason, these lines are not processed correctly, so that
Setup->pPreloads[i].ciphertext[0] is NULL in dynamic_SETUP at
dynamic_fmt.c:7224
 if (Setup->pPreloads[i].ciphertext[0] == 'A' &&
Setup->pPreloads[i].ciphertext[1] == '=') {

I wonder how gcc builds survived this without a segfault.

Jim, the backtrace has been posted in my previous message:
http://www.openwall.com/lists/john-dev/2014/01/02/22

Commenting out all the TestD= lines in run/dynamic_flat_sse_formats.conf
makes the bug disappear.

Changing the TestD= lines into Test= lines also avoids the segfault.
And --format=dynamic_2014 still passes self test with that change.

What is the purpose of TestD?

Frank

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.