Play non-warp ocarina songs from the pause menu

[bassdr]

Closes #6940.

What

Extends the pause-menu song shortcut (formerly "Pause Warp") to the six non-warp songs. On the Quest screen, pressing A on any song you know closes the menu and triggers that song's real gameplay effect: warp songs warp as before, and the zone songs (Zelda's Lullaby, Song of Storms, Song of Time) fire their staff-spot effect when you're standing in range of one.

The old "Pause Warp" toggle is split in two:

• Play Songs from Pause Menu - the shortcut itself, now covering every song.
• Skip Warp Cutscenes - the old animation-skip behaviour, broken out so it also applies when a warp song is played manually.

A config migration carries an existing "Pause Warp" enable onto both new flags.

Why / how

Rather than reimplement each song's effect and its preconditions, the shortcut delegates to vanilla. For a zone song it hands the matching in-range En_Okarina_Tag staff spot straight to its own listening handler and raises OCARINA_MODE_03, so the game's own code owns every effect, cutscene and eligibility check (for example the Ocarina-of-Time gate on the Door of Time). The only decisions the shortcut makes are proximity and song-matches-spot; everything scenario-specific stays in the actor.

Testing

• Opening the Door of Time with the shortcut.
• Playing the wrong song at a spot (no effect, as expected).
• Playing a song away from any spot (nothing triggers).
• Water Temple triforce level-changers, tested with the shortcut and manually playing the song (vanilla behavior), all works.
• Sheik's cutscene in Temple Of Time still plays when having all medallions but not played yet, with "Skip Warp Cutscenes" enabled or disabled, when warping in Temple Of Time
• Big Fairies -- There are many, but tested the one in lost wood's in the secret whole behind Mido, expecting Sun's Song
  • Not working
• Cows -- There are a few, all expecting Epona's Song.
  • I had an issue with a corrupted savegame where the cow did not give me milk, but this caused by a corrupted savegame. See Fix a crash when opening the debug menu with a corrupted savegame #6958
• Song of Time blocks
  • Not working, tested the one in Forest's section of the ganon tower, on the left after entering the room
• Mido -- Lost Woods as an adult expecting Saria's Song
  • WIP, I have a functional build with a specific logic for this one, not in this PR yet
• Darunia -- Goron City as a child expecting Saria's Song
• Skull Kid -- Lost Wood as a child I think expecting Saria's Song (To Be Validated, can't remember this one)
• Malon -- Lon Lon Ranch as an adult expecting Epona's Song
• Learn Saria's song
• Play a song underwater, horsback riding, in air is blocked

Build Artifacts

• soh-windows.zip
• soh-linux.zip
• soh-mac.zip

[djevangelia]

For a zone song it hands the matching in-range En_Okarina_Tag staff spot straight to its own listening handler and raises OCARINA_MODE_03, so the game's own code owns every effect, cutscene and eligibility check (for example the Ocarina-of-Time gate on the Door of Time).

I can't look at the decomp source code or this PR right now, but I think actors like Mido (En_Md) in Lost Woods call Message_StartOcarina or similar on their own without a separate En_Okarina_Tag actor, do they work as well?

• Playing a song away from any spot (nothing triggers).

Do songs with effects liks Sun's or Song of Time (if time travelling) work?

Edit: Had some minutes to look at it, I think it should be possible to incorporate player->ocarinaTalkActor for some checks to not have to go through every possible actor, and also use the Ocarina player states (state group 2 numbers 23-25) to determine whether song should have its generic effect or have an actor-specific effect where one must lookup the actor in question (i.e. is player close to an Ocarina actor or not).

[bassdr]

Thanks for the review!
I'll note here everything I've tested, in the new single-frame hook implementation:

• As a child, open the door of time: played two wrong songs then the song of time, opens only on the right song as expected
• Sun's song effect works anywhere it did in the original game, tested hyruld field and the marken as a child
• Epona's song as a child does nothing in hyruld field
• As an adult, Warp songs still work
• As an adult, epona's song works too, as long as she can reach you
• Water temple triforces:
  • tried playing the right song in the room but not in front of the triforce, does not work as expected
  • tried playing the wrong song in front of the triforce, does not work as expected
  • Playing the Zelda Lullaby in front of the triforce works as expected
• Destroyed the Royal Family Tomb in the Graveyard
• Dry the well in Kakariko Village, works. This one is a bit harder to repro if you want to try yourself, the vanilla cutscene trigggers with 0x65, 0x67 + the windmill per-scene 0x02 switch, all 3 has to be false for the cutsceen to trigger (vanilla or not)

actors like Mido (En_Md) in Lost Woods call Message_StartOcarina

Good catch, I confirm it does not work, but it does work when playing the ocarina the vanilla way.

Edit: Had some minutes to look at it, I think it should be possible to incorporate player->ocarinaTalkActor for some checks to not have to go through every possible actor, and also use the Ocarina player states (state group 2 numbers 23-25) to determine whether song should have its generic effect or have an actor-specific effect where one must lookup the actor in question (i.e. is player close to an Ocarina actor or not).

I'll try to fix it, good idea.

[bassdr]

@djevangelia, I tried your idea and it does not work sadly... I can't get Mido accept the saria song to let me pass. We enter a bogus state where Mido is still listening to us waiting for the Saria Song but it was already played, so we are both playing ocarina and not all at once.

IMO, these are the ways we can solve this:

1. Accept this as a limitation, as-is. With this PR playing in front of the actor does nothing, but you still can take out your ocarina and play the expected song, so the vanilla path is not broken.
2. Continue to investigate to find a "generic" way to achieve this, but I think it'll be 1. hard to find and 2. may break legitimate cases too, where we would inject "playing ocarina" events on actors expecting ocarina but not a specific song. I have the Child Lost Wood's Skull Kids in mind, I'm scared we'll have corner cases with those, like using the menu to play a random song in front of them, and we'd be in a bogus situation.
3. Implement per-actor logic. That is: in the hook, I'd implement a specific logic for each actor triggering exactly what they expect, but this logic would never execute except if we are in front of that specific actor.

I'd definitly go with option 3 myself, there are only 4 of these actor-cases. Worse case if we've missed some things, the vanilla behavior will still work and we can improve this later.

This is the result of Claude's analysis, but I am quite sure it's accurate:

• Mido -- Lost Woods as an adult expecting Saria's Song
• Darunia -- Goron City as a child expecting Saria's Song
• Skull Kid -- Lost Wood as a child I think expecting Saria's Song (To Be Validated, can't remember this one)
• Malon -- Lon Lon Ranch as an adult expecting Epona's Song

Details of why the generic implementation is hard (AI generated):

1. player->ocarinaTalkActor (unk_6A8) can't identify the target from the menu -- it's empty in our context

It's only populated while the player is actively in the ocarina "play-for-actor" flow, which the pause menu never enters. Even standing next to an actor:

• En_Okarina_Tag sets it only for type-7 spots (z_en_okarina_tag.c:138); the cutscene tags (types 1/2/4/6) set only NEAR_OCARINA_ACTOR.
• NPCs like Mido set it only after ATTEMPT_PLAY_FOR_ACTOR (z_en_md.c:757); at rest they set only NEAR_OCARINA_ACTOR (:764).
• And z_player.c:12113 clears unk_6A8 every frame unless PLAY_FOR_ACTOR.

So ocarinaTalkActor is NULL when we run — it doesn't spare the actor lookup.

2. The ocarina states (23-25) give only a boolean "near an ocarina actor," not which one.

NEAR_OCARINA_ACTOR is fine for a generic-vs-specific split, but it can't identify the actor, and ATTEMPT/PLAY_FOR_ACTOR are never set from the menu -- so you still must enumerate to find the target.

3. The only way to deliver via those states (set ATTEMPT_PLAY_FOR_ACTOR, let the actor react) breaks the gameplay

We'd have to make sure we don't break the gameplay behavior

[djevangelia]

Aha, it's only during the play. Makes it difficult.

My suggestion regarding Ocarina states was not for specific actors, but for quickly determining that the song is to be played generic and not for actor. Some players might want to be able to summon Epona from the pause menu etc.

[bassdr]

Aha, it's only during the play. Makes it difficult.

My suggestion regarding Ocarina states was not for specific actors, but for quickly determining that the song is to be played generic and not for actor. Some players might want to be able to summon Epona from the pause menu etc.

The epona from pause works as is, of course.

Just to be clear, in the latest commit of this pr, everything I've tested works as you expected it, is it that you play the song from the pause menu or play it manually. Most gameplay events work too, except 4 gameplay events won't trigger with the pause menu, but I did not break the vanilla behaviour, you still can use the Ocarina to pass mido. Everything else I've tested works as expected.

The last message is about finding a way to fix this 4 corner cases and let us pass mido with the pause menu's Saria's song

[serprex]

Feel like a gameplay macro system would address this pretty well, maybe some trickery to force using ocarina out of pause

[djevangelia]

Cows also respond to Ocarina but seem to have a different system for activation relying on ocarinaMode and R_EPONAS_SONG_PLAYED set through z_message.c

[bassdr]

Cows also respond to Ocarina but seem to have a different system for activation relying on ocarinaMode and R_EPONAS_SONG_PLAYED set through z_message.c

Right, they answer to epona's song if I remember well, I'll test this case too

[bassdr]

The big-fairy spawning on some songs (usually sun's or storms) won't work either with this last commit, the fairy won't spawn when the right song is played from the menu, they do when playing the ocarina. Still looking for a cow, but my guess is that they won't work either.

[bassdr]

Cows also respond to Ocarina but seem to have a different system for activation relying on ocarinaMode and R_EPONAS_SONG_PLAYED set through z_message.c

Right, they answer to epona's song if I remember well, I'll test this case too

Confirmed this one somewhat works with my local changes... But the cow won't give me what I want for some reason even if I play the vanilla song... I'm re-testing without these changes to compare...

I also confirmed playing the song of storm in the Kokiri forest near the Lost Wood entrance works to open the secret whole... so really the only remaining corner cases to test are

• Big Fairies -- There are many, but tested the one in lost wood's in the secret whole behind Mido, expecting Sun's Song
• Cows -- Partially working, but ocarina play broken too, TBC if a regression
• Mido -- Lost Woods as an adult expecting Saria's Song
• Darunia -- Goron City as a child expecting Saria's Song
• Skull Kid -- Lost Wood as a child I think expecting Saria's Song (To Be Validated, can't remember this one)
• Malon -- Lon Lon Ranch as an adult expecting Epona's Song

[djevangelia]

The big-fairy spawning on some songs (usually sun's or storms) won't work either with this last commit, the fairy won't spawn when the right song is played from the menu, they do when playing the ocarina. Still looking for a cow, but my guess is that they won't work either.

They are from z_shot_sun.c which seem to work similarly to z_obj_timeblock.c (most SoT blocks) and z_obj_warp2block.c (some SoT blocks such as SoT block room in MQ Spirit) which also should be tested. Sorry for coming with new actors all the time, I don't barely remember any of them as I don't play. Do you have an empty bottle for cows?

[bassdr]

The big-fairy spawning on some songs (usually sun's or storms) won't work either with this last commit, the fairy won't spawn when the right song is played from the menu, they do when playing the ocarina. Still looking for a cow, but my guess is that they won't work either.

They are from z_shot_sun.c which seem to work similarly to z_obj_timeblock.c (most SoT blocks) and z_obj_warp2block.c (some SoT blocks such as SoT block room in MQ Spirit) which also should be tested. Sorry for coming with new actors all the time, I don't barely remember any of them as I don't play. Do you have an empty bottle for cows?

Not a problem! Thanks for helping!

So I actually forgot to test the SoT blocks, I'll have to do it. I'll edit the PR description with a list of things I've tested vs. know aren't working.

For the cow, I suspect I either have a corrupted data, or one of my mod broke it. I'm currently testing in develop to see if I have a repro. High chances I found an unrelated bug, and the feature works as expected. Yes, I have an empty bottle, and the sound I hear is the 5-rupy sound, not the usual Fanfare, very weird.

[bassdr]

Not sure why, but develop (8481060) is crashing a lot for me... I turned off my HD texture mods but it still crashes. 29285ed does it too.

Anyway, I tried to go to the exact same cow, with exact same behavior on the develop branch: the cow speaks to me about its childhood, then 5 rupee song, then no milk in my empty glass. Unrelated issue. I'll investigate it separately.

I think my savegame is corrupted so the cow thinks I don't have an empty bottle.

This is the callstack I get:

Traceback:
1 /usr/lib64/libc.so.6(+0x41ad0) [0x7f569053cad0]
2 Fast::Interpreter::GfxSpVertex(unsigned long, unsigned long, Fast::F3DVtx const*) (+0xE3)
3 Fast::gfx_vtx_handler_f3dex2(Fast::F3DGfx**) (+0xE2)
4 Fast::Interpreter::Run(Gfx*, std::unordered_map<MtxS*, MtxF, std::hash<MtxS*>, std::equal_to<MtxS*>, std::allocator<std::pair<MtxS* const, MtxF> > > const&) (+0x575)
5 Fast::Fast3dWindow::DrawAndRunGraphicsCommands(Gfx*, std::unordered_map<MtxS*, MtxF, std::hash<MtxS*>, std::equal_to<MtxS*>, std::allocator<std::pair<MtxS* const, MtxF> > > const&) (+0x186)
6 RunCommands(Gfx*, int, int, int, int) (+0x3C5)
7 Graph_ProcessGfxCommands (+0x28C)
8 Graph_ThreadEntry (+0x5C)
9 Main (+0x7D2)
10 main (+0x11B)
11 /usr/lib64/libc.so.6(+0x2b47e) [0x7f569052647e]
12 __libc_start_main (+0x8B)
13 _start (+0x25)

[bassdr]

OK tested with a fresh savegame, and it works with both the ocarina or using the menu to play the song. I used the cow in the house as a kid (forcing the save to have the bare minimum) and it works. So I'll mark this one as working.

I also fixed my savegame with #6958 and after the cows were giving me some milk as expected.

[TheOneTrueSpoon]

uh for some reason when learning songs it displays 2 notes then doesn't let me repeat the song to learn it, don't know if you are just starting with the songs but for zelda's lullaby i had to debug warp out to fix, it doesn't seem to work with epona's song though so i have to edit my save to add it

[bassdr]

Latest version added a check so we can't play a song where the ocarina cannot be used (horseback riding, in the air, underwater, etc.) to avoid gameplay bugs; I reused the in-game logic. I tested Mido, Darunia and the SoT blocks, they now work. Big Fairies are using similar mechanism than the SoT block but I did not have time to re-test, same for Skull Kid and Malon, they should work but I did not test.

Feel like a gameplay macro system would address this pretty well, maybe some trickery to force using ocarina out of pause

I did consider this btw @serprex, I just felt I was almost there with the current approach of sending gameplay events directly, but using a macro system to send actual song notes would work to, pretty sure there are corner cases also (detect if the actor is listening before sending the notes, etc.)

[bassdr]

uh for some reason when learning songs it displays 2 notes then doesn't let me repeat the song to learn it, don't know if you are just starting with the songs but for zelda's lullaby i had to debug warp out to fix, it doesn't seem to work with epona's song though so i have to edit my save to add it

I could not repro this on my end... maybe you had a corrupted savegame like I did at some point? I did test learning the saria's song near the forest temple entrance as a kid.

[bassdr]

I did another testing pass on this one and fixed the last remaining bugs I could spot. @serprex, @djevangelia, ready for another review pass.

Tested all actors that interact with the ocarina: Darunia, Skull Kid, Mido, and Malon, plus Song of Time blocks, cows, etc. — everything is now in the PR description. One caveat with Malon: I can't remember what was the expected behavior, but I do see the exact same dialogue after playing the song manually and with the Pause menu.

I also added a check that the ocarina can actually be played before playing the song: if it can't, a dialogue shows up saying you can't play the ocarina here. The message is translated in English (fluent), French (native), and German (AI-generated). Tested many cases to make sure it doesn't break gameplay. This check also applies to the old warp songs — it would be easy to gate behind an option to bring back the old behavior (no check was done before warping), but I didn't.