The click plays every meter as if it were even, so odd and compound meters do not sound like themselves #595

The click marks exactly one beat per bar. Both implementations use the same rule -- (index - mark.beat) % per_bar == 0 in app/pipeline/click_render.py and the equivalent in static/js/metronome.js -- and there are only two click voices, an accent and a plain beat.

That is correct for 2/4, 3/4 and 4/4. It is wrong for every meter that carries internal stresses.

What it gets wrong

7/8 comes out as seven even clicks. In practice 7/8 is grouped, usually 3+2+2 or 2+2+3, and a player follows those group starts, not a bar of seven equal pulses. Without them the click gives no usable pulse to lock onto.

6/8 is already shipping wrong, and it is a preset. It renders as six straight eighths with a single accent. 6/8 is felt in two dotted-quarter groups, so it wants a stress on 1 and a lesser one on 4. The same applies to 9/8 and 12/8.

5/4 has the same problem, being 3+2 or 2+3 rather than five even beats.

Note this is not the same gap as #593. Adding more meters to the selector lets a user pick 7, but the click still plays those seven beats evenly, so an odd meter is selectable and still unusable.

What a fix needs

Grouping is the feature. A bar needs a subdivision -- 7 as 3+2+2, 6 as 3+3, 5 as 3+2 -- and an accent at each group start. A denominator on its own buys nothing: 7/8 and 7/4 are the same seven pulses grouped the same way, so adding "/8" to a label would not change a single click.

Two click voices are not enough. With only accent and plain, a 3+2+2 bar has three identical accents and the downbeat stops being findable. Grouping needs three levels: strong on beat 1, medium at each group start, weak elsewhere.

Constraints

• The live click (static/js/metronome.js) and the rendered one (app/pipeline/click_render.py) must agree beat for beat. They are held to each other by tests/js/count-in.test.mjs and tests/test_click_render.py, and a player who monitors one and exports the other must not hear a difference.
• The count-in runs through the same accent logic on both sides and has to group identically.
• Changing 6/8 to its correct feel changes what existing users hear from a preset they already use. That is a deliberate behaviour change and should be called out, not slipped in.
• Defaults for 2, 3 and 4 should stay exactly as they are. 4/4 has a real secondary stress on beat 3, but adding it by default would change the most common meter in the app for everyone.
• Any grouping the user supplies has to sum to the bar length, and the export needs it too or a rendered file will not match playback.