Don't reload sample from disk when reversing

[DomClark]

Currently, when reversing a sample, SampleBuffer reloads it from disk. This is unnecessary and makes it impossible to do so in realtime code. This change still isn't realtime safe, since it locks a mutex, but a proper fix requires much wider changes, and it doesn't read from the disk any more at least.

Blocks #5695.

[LmmsBot]

🤖 Hey, I'm @LmmsBot from github.com/lmms/bot and I made downloads for this pull request, click me to make them magically appear! 🎩

Windows

• Windows 32-bit: lmms-1.2.3-729%2Bgdde5576be-mingw-win32.exe (build link)
• Windows 64-bit: lmms-1.2.3-729%2Bgdde5576be-mingw-win64.exe (build link)

Linux

• Linux (AppImage): lmms-1.2.3-729%2Bgdde5576-linux-x86_64.AppImage (build link)

macOS

• macOS : lmms-1.2.3-729%2Bgdde5576be-mac10.13.dmg (build link)

🤖

{"platform_name_to_artifacts": {"Windows": [{"artifact": {"title": {"title": "32-bit", "platform_name": "Windows"}, "link": {"link": "https://9390-15778896-gh.circle-artifacts.com/0/lmms-1.2.3-729%2Bgdde5576be-mingw-win32.exe"}}, "build_link": "https://circleci.com/gh/LMMS/lmms/9390?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-build-link"}, {"artifact": {"title": {"title": "64-bit", "platform_name": "Windows"}, "link": {"link": "https://9393-15778896-gh.circle-artifacts.com/0/lmms-1.2.3-729%2Bgdde5576be-mingw-win64.exe"}}, "build_link": "https://circleci.com/gh/LMMS/lmms/9393?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-build-link"}], "Linux": [{"artifact": {"title": {"title": "(AppImage)", "platform_name": "Linux"}, "link": {"link": "https://9391-15778896-gh.circle-artifacts.com/0/lmms-1.2.3-729%2Bgdde5576-linux-x86_64.AppImage"}}, "build_link": "https://circleci.com/gh/LMMS/lmms/9391?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-build-link"}], "macOS": [{"artifact": {"title": {"title": "", "platform_name": "macOS"}, "link": {"link": "https://9392-15778896-gh.circle-artifacts.com/0/lmms-1.2.3-729%2Bgdde5576be-mac10.13.dmg"}}, "build_link": "https://circleci.com/gh/LMMS/lmms/9392?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-build-link"}]}, "commit_sha": "eb6b8b3753cd936213ada57cfcb50aee6d7a2f6f"}

[Spekular]

Code looks reasonable to me! Do you want any testing on this? Anything specific?

[Spekular]

Wait, actually, doesn't this assume that every call to setReversed toggles the state? It appears to me that _on is meant to define if the sample should be reversed or not, so repeated calls to setReversed with the same argument should do nothing, right? Whereas here the sample would be reversed each time.

[DomClark]

Good catch! I was looking at similar code by Reflexe, but didn't quite take in that he changed the function from setReversed(bool) to reverse(). Should be fixed now. (Also, m_reversed is now set while the lock is held, which should be safer.)

In terms of testing, I don't think an awful lot is required. Flip some samples back and forth in AFP, make sure they save and load the same way round.

[JohannesLorenz]

LGTM

Still not RT safe, but a great performance increase.

[Spekular]

Did some very quick testing, everything worked as it should. I would probably have wrapped the whole function inside if (m_reversed != on) { ... } to avoid taking the locks if they're not needed, but in practice the if should always be true so it won't make any difference.

[DomClark]

I would probably have wrapped the whole function inside if (m_reversed != on) { ... } to avoid taking the locks if they're not needed

If this function is going to be called from multiple threads, the variable should be read and written under the same lock. Suppose two threads tried to reverse the sample simultaneously - both see that it's not reversed, so both reverse it and it ends up unreversed, even though m_reversed is now true.

Anyway, this has two approvals now, so I will merge it tomorrow if there are no objections.

[Spekular]

Ah, great point.