[bug #68436] test fail in function fminunc

"Dmitri A. Sergatskov" <[email protected]> Tue, 9 Jun 2026 08:22:15 -0400 (EDT)

Newsgroups	gmane.comp.gnu.octave.bugs
Message-ID	<[email protected]>

Please use the bug tracker to post updates to a bug report. The mailing list is intended as a read-only notification stream. Info posted to this mailing list address won't appear in the tracker database where it is most useful.

Follow-up Comment #3, bug #68436 (group octave):

I was getting this error which Intel MKL, but never with anything else
(openblas, netlib, Apple's Veclib). So I have a fix
but never bother to file the bug report since no-one seems to be able to
reproduce it.
The file that worked for me is attached
(file fminunc.m )

There are essentially two differences:
1) (to fix the actual bug) Before failing with info = -3 when the trust region
becomes too small, check if we're already close to convergence.

2) (discovered when I was debugging the first one) We need to double the
default MaxFunEvals when using central differences (200*n vs 100*n), since
central differences require twice as many function evaluations per gradient.

Let us know if that works for you too. Since it is an M-file you do not need
to recompile anything just replace the old one.

Dmitri.
--

(file #58695)

_______________________________________________________

Additional Item Attachment:

Name: fminunc.m Size: 16KiB
<https://file.savannah.gnu.org/file/fminunc.m?file_id=58695>

AGPL NOTICE

These attachments are served by Savane. You can download the corresponding
source code of Savane at
https://git.savannah.nongnu.org/cgit/administration/savane.git/snapshot/savane-c36938be85ff6c1b727bc7dd7fd30e48f9142870.tar.gz

_______________________________________________________

Reply to this item at:

<https://savannah.gnu.org/bugs/?68436>

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

signature.asc (application/pgp-signature, 228 B)