[Octave-bug-tracker] [bug #58800] BIST for rng sometimes fails

[Markus Mützel]

Follow-up Comment #10, bug #58800 (project octave):

@Michael Leitner: Thanks for the feedback.
The C++11 <random> header seems to provide an interface to a "real" (possibly
non-deterministic) random source:
http://www.cplusplus.com/reference/random/random_device/
https://en.cppreference.com/w/cpp/numeric/random/random_device/random_device
IIUC, it uses /dev/urandom on Linux.
For mingw-w64, it should produce non-deterministic values (using rand_s)
starting with gcc 9.2 (MXE Octave currently uses 9.3):
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85494
https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/rand-s?view=vs-2019

The attached patch adds the C++11 "entropy source" to the Mersenne-Twister
initialization.

It might also be possible that that code block could be used as a replacement
for the code block (just above in the same function) that reads from
/dev/urandom.

If I correctly understand the documentation, the constructor of random_device
might throw an exception. That isn't handled yet. The patch is only for
testing.

@hg200: Could you try if you still see the issue with this patch? (I haven't
tested yet. So I hope it cross-builds and doesn't fail.)

Doing something different on a reset (instead of re-initializing) also sounds
like a good idea to me.

(file #49535)
    _______________________________________________________

Additional Item Attachment:

File name: bug58800_random_device.patch   Size:0 KB
   
<https://file.savannah.gnu.org/file/bug58800_random_device.patch?file_id=49535>



    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?58800>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/