On Thu, 9 Sep 2021 18:01:38 +0200
Ichthyostega <p...@ichthyostega.de> wrote:
Bottom line: we could propose the Maintainers to use a better test,
e.g. a shell script which feeds some command to the CLI and then
closes yoshimi. Maybe using the new built-in test-invoker ;-)
echo "set test exec" | yoshimi --null --no-gui --cmdline
Am 12.10.21 um 12:08 schrieb Will Godfrey:
I've had a very positive response from one of the debian maintainers :)
He doesn't know who set that test, but agrees it's ridiculous.
He likes the idea of that test, but hit a bug (my fault) - trying to show a
splash on a first time start with no GUI available. So that'll be a bugfix
release in pretty short order :(
Hi Will,
Yes, indeed... I saw already while preparing my DEB package,
that official Debian Packager has integrated the proposed test and even
expanded a bit on that idea
However, while he got similar results on all platforms, he found significant
differences between V2.1.0 and 2.1.1
I can't think of anything we've done that could cause that. Could it be due to
changes made in the test code itself?
My guess is, they see the same differences we recently saw when running
the sound generation on another CPU. We haven't investigated that in more
detail yet, but we probably all suspect that this is due to some kind
of (very minor) rounding error.
Anyway, after building and installing the new Yoshimi version,
I re-run the Testsuite on my local PC, and it was all GREEN.
I.e. on /the same/ PC/System the generated sound is 100% identical.
Obviously, right now we can only tell to the extent which is covered
currently in the Testsuite, which is basically *just one Sine* and
some run time measurements. But IMHO this shows already the potential
of such automated testing, since it's now very easy to perform such
a basic check :-D
-- Hermann
_______________________________________________
Yoshimi-devel mailing list
Yoshimi-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/yoshimi-devel