[MU3] Fix #314899: Identical expression

[Jojo-Schmitz]

Resolves: https://musescore.org/en/node/314899

Same fix needed in master, see #7180, or #7627, both closed by now.
And see #8762, which brought this up again.

[Jojo-Schmitz]

The mtest fail (https://github.com/musescore/MuseScore/pull/7179/checks?check_run_id=1622482883#step:5:761) seems to be the result of this PR's change.

For some strange reason the corresponding PR for master builds cleanly though

[AntonioBL]

The crash is caused by an assert at line 104 of importexport/midiimport/importmidi_simplify.cpp for test voiceSeparationTuple (i.e. importing voice_tuplet.mid)
QFATAL : TestImportMidi::voiceSeparationTuplet() unknown: ASSERT failure in Simplify::lengthenNote: "Too short note durations remaining", file /media/antonio/Mint0/MuseScore/importexport/midiimport/importmidi_simplify.cpp, line 104

[AntonioBL]

And I think that in master importmidi mtests are not being tested at the moment.

[Jojo-Schmitz]

It had to be one of those 4 Q_ASSERT_X, but why?

[AntonioBL]

By using a break at that X_ASSERT (line 104 in 3.x and line 102 in master), the noteDurations list is:

first	second
2/2	1/16 dots 0
2/2	1/128 dots 0
2/2	1/512 dots 0

which gives sum = 37/512, while the desiredLen (offset - DurationStart) is 7/96.

[Jojo-Schmitz]

Close, but not the same, 0.072265625 vs 0.072916666

Guess a case of "never check for error conditions you can't handle"

Need to introduce some kind of an epsilon for this apparent floating point comparison?

[Jojo-Schmitz]

Importing the "tuplet_3_5_7_tuplets.mid" file produces a corrupt MuseScore file (in 3.5.2 and 3.6 RC):
Measure 1, Staff 2 incomplete. Expected: 4/4; Found: 126/72

The reference score though looks very different:

It dies at an entirely different assert:

MuseScore/importexport/midiimport/importmidi_clef.cpp

Line 483 in 7a465c3

Q_ASSERT_X(!doesClefBreakTie(staff), "MidiClef::createClefs", "Clef breaks the tie");

and I don't see how that might be related to this PR here?

Not sure yet what's going wrong with the "voice_tuplet.mid" file/test, this is the one that fails the assertion at

MuseScore/importexport/midiimport/importmidi_simplify.cpp

Lines 104 to 105 in 7a465c3

	Q_ASSERT_X(areDurationsEqual(noteDurations, offTime - durationStart),
	"Simplify::lengthenNote", "Too short note durations remaining");

and as such is related to this PR

[AntonioBL]

tuplet_3_5_7_tuplets.mid is imported exactly as the reference if "Split staff" is unchecked. And indeed in the mtest there is data.trackOpers.doStaffSplit.setDefaultValue(false, false);
Edit:
Actually, not exactly as the reference; in order to have it exactly as the reference, also "clef changes" must be unchecked (in the mtest data.trackOpers.changeClef.setDefaultValue(false, false); )
Edit 2:
But then, when playing with the import options and then exiting without saving, MuseScore 3.5.2 crashes

[Jojo-Schmitz]

For me that .mid file crashes with an assertion failure on load (in a debug build). Only when importing in a non-debug build I could save it as mscz (but without using any of those options) and got that corrupt score.
Anyway; I think this crash might be a red herring, as it crashes in an entirely different place, one that my change can't possibly have caused (famous last words?).

[Jojo-Schmitz]

Want me to port it over to master?

[Jojo-Schmitz]

That "Thumbs up" doesn't notify...
Anyway, see #8849