[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Accepted octave-signal 1.4.4-1 (source) into experimental



* Sébastien Villemot <sebastien@debian.org> [2023-06-01 10:22]:

Le jeudi 01 juin 2023 à 05:05 +0200, Rafael Laboissière a écrit :
* Sébastien Villemot <sebastien@debian.org> [2023-05-31 22:52]:

Le lundi 22 mai 2023 à 07:03 +0000, Debian FTP Masters a écrit :

Date: Mon, 22 May 2023 03:05:20 -0300
Source: octave-signal
Architecture: source
Version: 1.4.4-1
Distribution: experimental

Note that it failed to build on many architectures.

Some unit tests for firpm are failing on those architectures. The function firpm has been introduced in version 1.4.4. It is implemented in file src/firpm.cc, which calls the C file src/mmfir.c. This is a new implementation of McClellan-Parks-Rabiner method for designing FIR filters using the minimax technique. It is intended to replace the remez function, which is also on the signal package.

These files were proposed by Rob Sykes back in 2014 and have only been integrated into the signal package in 2023 (see [1]). I would guess that the author of the files, as well as the maintainers of the signal package, did not test the new code on the problematic architectures.

I do not have time to debug this issue.

Any ideas on how to proceed ?

I would suggest to open a bug upstream, possibly giving details about the test failures. And then if upstream does not provide a fix, simply disable the tests.

I just noticed this: https://savannah.gnu.org/patch/?10351

I will apply the patch from the Mercurial upstream repository and release version 1.4.4-2 to experimental. Let us see how the autobuilders will behave.

Best,

Rafael Laboissière


Reply to: