Eaten Dashdance Inputs - Bisected r18 Regression about ishiiruka

[NikhilNarayana]

Can you revert that commit and test?

Looks like it causes conflicts, might need to do it right at a 2.0.x rather than the most recent commit but there aren't tags for those so I'll have to look through the history. I'll see if a can resolve them later but I don't really know anything about the code base.

On branch slippi
Your branch is up to date with 'origin/slippi'.

You are currently reverting commit 078aa1e.
(fix conflicts and run "git revert --continue")
(use "git revert --abort" to cancel the revert operation)

Changes to be committed:
(use "git reset HEAD ..." to unstage)

    modified:   Source/Core/DolphinWX/Frame.h
    modified:   Source/Core/DolphinWX/FrameTools.cpp

Unmerged paths:
(use "git reset HEAD ..." to unstage)
(use "git add ..." to mark resolution)

    both modified:   Source/Core/Core/HW/EXI_DeviceSlippi.cpp
    both modified:   Source/Core/Core/HW/EXI_DeviceSlippi.h
    both modified:   Source/Core/Core/HW/SI.cpp

Well, so far I've isolated the issue to Source/Core/Core/HW/SI.cpp 6128f83.

All the conflicts were completely innocuous other than different values for ROLLBACK_MAX_FRAMES (7 vs 5) in EXI_DeviceSlippi.h, which I verified had no impact on the issue. So I did a partial revert where I only reverted the changes with an impact in SI.cpp, and it feels much better.

I haven't bothered trying to alter the remaining changes for a fix at all since I have basically no knowledge of the inner workings of this code.

[JLaferri]

The commit you linked, does it revert that section to what it was before?

Other than the bit of code here 078aa1e#diff-b62f4664b164e36115c14f8f55559034R684, which is commented out in the current revision, yes, it reverts SI.cpp to 078aa1e.

[JLaferri]

Gotcha, that should all be safe. You can probably just revert anything that references the g_BButtonPressTime variable. Mind doing that?

[JLaferri]

Afterwards feel free to submit the PR, I'll look it over and merge if it looks good.

C++ isn't my expertise but I can probably manage. These are the only references to g_BButtonPressTime:

./Source/Core/Core/HW/EXI_DeviceSlippi.cpp:76:extern u64 g_BButtonPressTime;
./Source/Core/Core/HW/EXI_DeviceSlippi.cpp:1959:        //      NOTICE_LOG(SLIPPI, "%llu", Common::Timer::GetTimeUs() - g_BButtonPressTime);
./Source/Core/Core/HW/SI.cpp:683:               g_BButtonPressTime = Common::Timer::GetTimeUs();
./Source/Core/Core/HW/SI.cpp:684:               ERROR_LOG(SLIPPI, "B Press: %llu", g_BButtonPressTime);

You just want me to rip that stuff out?

[JLaferri]

Yeah any reference to the variable and you can get rid of the commented code blocks that reference it too

[JLaferri]

Merged

Yes, thank you both for working with me on it. Does this affect both rollback online and old netplay lobbies, or just old netplay lobbies? I haven't tried to test the changes with rollback online at all because I assumed it would only allow official release clients.

[NikhilNarayana]

Yes, thank you both for working with me on it. Does this affect both rollback online and old netplay lobbies, or just old netplay lobbies? I haven't tried to test the changes with rollback online at all because I assumed it would only allow official release clients.

you can try playing with the latest commit, just no guarantees it won't desync.

[JLaferri]

@reinhard33 how were you able to confirm that this fixed your issue btw? Just testing it locally? Also do you have discord? would be easier to communicate more quickly there, feel free to DM after joining the Slippi Discord

I confirmed that this PR fixed the issue with the local netplay lobbies. I think the slowness online I mentioned in the first post was something else because I did my first direct connect with a friend on 2.0.3 with known good connections, and that was perfect.