Re: [Plplot-devel] Finalized plans for 5.14.0 freeze date, testing period, and release

[Alan W. I. <Ala...@gm...>]

On 2018-10-03 10:18-0700 Alan W. Irwin wrote:

> On 2018-09-23 18:56-0700 Alan W. Irwin wrote:
>
>> I continue have a large list of different topics I would like to work
>> on for PLplot, but, on the other hand, comprehensive tests are looking
>> good (at least on Debian Buster) right now, and it has been much too
>> long since our last release (mostly because of the "new computer" and
>> "new distro version" issues I have been encountering).  So I plan to
>> spend the next several weeks working on the most urgent development
>> topics I have a good chance of finishing before the soft freeze, with
>> a proposed date of the freeze near the end of October followed by a
>> testing and debugging period with the actual release of 5.14.0
>> occurring roughly in mid November.  Please let me know if the general
>> timing of that soft freeze, testing period, and subsequent release
>> will cause you any issues with development topics you would like to
>> push before the release of 5.14.0.  But if I don't hear any strong
>> objections along those lines, then later this week I plan to finalize
>> that soft freeze date as October 27th (to be definite and to place it
>> on the last Saturday in October).
>
> This is official notice that I will be going ahead with this plan.  In
> particular, the soft freeze date has now been finalized as of October
> 27th.

Current status:

This is a reminder that the soft freeze deadline is only 5 days away.
But I am going to stick to it (e.g., some of the small topics I am
currently working on may have to be put off until post release, but
such postponent is convenient to do with git since I am working on
these topics using topic branches.)

Comprehensive testing is going well.  With the current master tip
version I am getting essentially perfect results on Linux (Debian
Buster).  That is there are no obvious configure, build, or run-time
issues and there is a perfect test_diff_device report for the
(default) svg testing device.  Also Arjen is getting promising-looking
results for Cygwin.  There is one regression there (the octave binding
does not build) but a follow-up comprehensive test with octave
disabled worked without issues, and the solution for the octave build
difficulty may simply be a Cygwin system upgrade for one of the
octave-related components.  In any case, I am very happy with this
preliminary result since a lot of build-system development has
occurred since the last time that Cygwin was comprehensively tested.

Despite this complete comprehensive testing success on Linux and
almost-complete preliminary comprehensive testing success on Windows
some additional volunteers to help out with comprehensive testing
would add a lot to the value of this release. For example, we need
someone to comprehensively test the MinGW-w64/MSYS2 Windows platform
if Arjen does not have time to get to it, and we need someone to test
any or all of Mac OS X + fink, MacPorts, and/or HomeBrew.  In addition
"redundant" comprehensive testing on the already tested Cygwin and
Linux platforms would be useful since there are so many different
Windows platforms that underlie Cygwin, and so many different Linux
distributions that do things differently than my Debian Buster
platform.

N.B. the advantages of comprehensively testing this release on as many
different platform variants as possible are obvious from the PLplot
release integrity perspective, but I would also like to emphasize an
additional individual advantage for testers which is testing helps you
learn how to get the most out of PLplot (e.g., which packages to
install to enhance the power of PLplot) for your platform of choice.

Also once your particular system is set up properly for PLplot,
running the comprehensive testing script is pretty trivial if you
specify the "--do_test_interactive no" option so you don't have to
baby-sit the tests generated by the script.  But in any case, please
run "scripts/comprehensive_test.sh --help" to get an idea of what is
possible.

Just to give you an idea of what is involved with the
"--do_test_interactive no" case you do need access to 6 GB of spare
disk space to store the many plot files that are generated by these
tests, and my ~11-year old Intel box took something like ~5 hours to
complete this set of tests while my modern Ryzen 7 1700 box with
faster memory and CPUs and 4 times as many CPUs takes roughly ~one
hour to complete the noninteractive tests.  In my case for
"--do_test_interactive no", I simply use the bash source command to
set relevant environment variables, start the script, and then I do
something else until the tests complete. And it should be just that
simple for you as well once you have installed all relevant platform
packages and decided which components of PLplot have to be disabled if
they are not working on your particular platform.

In sum, comprehensive testing should be straightforward so I hope to
hear from *all* of you lurking on this list who have access to a
reasonably fast modern system where you normally build PLplot in any
case.  And if you want to comprehensively test PLplot on some low-end
hardware (e.g., a raspberry PI system), such comprehensive test
results would be most interesting as well!

Alan
__________________________
Alan W. Irwin

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________