Online Developer Meeting (2025-08-26)

From Octave
Jump to navigation Jump to search

Today's topics

  • Prepare release of Octave 10.3.0? 33 "actual" changes on the stable branch since the release of Octave 10.2.0 (ignoring, e.g., changes to the CI rules or reverted changes), 7 of which are documentation only. Markus will start release process as time permits
    • Any pending changes or bugs that need to be fixed for the stable branch?

bug #67459 (clear ?) jwe has a potential fix. He will test and commit before 10.3 release

  • Should Octave be built without debugging code for parser by default? Yes. Rik will work on this.

The debugging code is always there and always enabled. For every statement parsed there is a check on the the parse_debug_flag to see whether debugging information should be printed. But, this is really infrequently used code. I have turned on the debugger in the parser just once in 10 years. To me, this is specialist code that could be enabled by a developer during configuration or with a `-D` option to `cpp`, but ordinary users will never need this.

Previous topics

  • Replace the abundant files in liboctave/operators/ with modern C++ templates? (Seems like use of modern C++ templates might work) Start a Discourse thread on what code pattern to use
  • Update binutils in Octave for Windows from 2.40 to 2.44 on the release branch of MXE Octave now? Done by Markus
  • How to proceed with updating mod/rem. Current implementation gives nonsensical results for large values (bug #45481) and incorrect results for values > 2/3 flintmax () (bug #67339). Results are also not exactly Matlab compatible for various corner cases (I can't find the exact threshold they are using to determine when something is within round-off error of 0) Arun has fix implemented for bug #67339. Thus, pressing need to resolve other problems with `mod` is gone
  • The F5 key (save and run code) in GUI is broken in Windows because of the operating system's poor timestamping (bug #67340). This issue has come up repeatedly. Can we resolve it, even with a hack, so that we don't get more complaints and bug reports about it? jwe will look at changing GUI code to have save operation emit a signal when file save has finished. Only then will GUI advance to running the file.
  • Should we discontinue NA? Implementation relies on bit signature in value that represents NaN. Not all CPUs seem to respect this. Expand documentation on how to use NA properly. It may not be preserved in numerical context
  • Availability of hg.savannah.gnu.org for buildbots. jwe will point buildbots at hg.octave.org instead

See also