Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 18 Nov 2022 12:59:41 +0100
From: Solar Designer <>
Subject: Re: JtR v1.9.0 Jumbo 1 bleeding, fatal error with /dev/shm in Windows CMD and powershell

On Thu, Nov 17, 2022 at 08:26:28PM -0500, Robert Harris wrote:
> JtR cygwin builds with OpenMPI/MPI enabled during compile time, have the 
> "MPI_Init" and "/dev/shm" fatal error when ran outside of cygwin (via 
> CMD or powershell).
> JtR cygwin builds with OpenMPI/MPI disabled during compile time, seem to 
> run without error outside of cygwin.
> ######
> 1. Was this already a known issue?

Not to me.

> 2. Is this the main reason why OpenMPI support is disabled by default 
> during compile time?

No, not this specific Windows-only issue.  However, in general an
MPI-enabled build assumes you'd actually use MPI and run via "mpirun".
So when you build with MPI and then don't use MPI, it's just weird at
best and an unjustified security risk at worst (since the MPI library
gets initialized and may expose itself over network).

magnum, maybe there's room for improvement by us?  Can we detect not
running under "mpirun" and then skip MPI initialization?

> 3. Does JtR jumbo 1 bleeding code need to change to fix this "MPI_Init" 
> issue, or does this issue need to be resolved somewhere else like with 
> cygwin or OpenMPI?

The immediate issue is in third-party libraries.  However, maybe we can
improve things in our project as well.


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.