Update: abseil-cpp, rnnoise, easyeffects; new package: webrtc-audio-processing-2

[JkktBkkt]

Testing the changes

• I tested the changes in this PR briefly:
  • manually used: vlc, telegram-desktop, easyeffects, android-tools
  • as dependency in above: tg_owt, re2, protobuf, rnnoise
  • as dependency in only builds: grpc
  • builds only: sysdig, python3-grpcio, usbguard, protonmail-bridge, protobuf-c, nsjail, netdata, mosh, libphonenumber, libcompizconfig, kicad, justbuild, Bear, alsa_rnnoise

New package

• This new package conforms to the package requirements: YES
  • easyeffects requires v2 of webrtc-audio-processing
  • pulseaudio still requires v1

Local build testing

• I built this PR locally for my native architecture, x86_64 (and tested as described above)
• Built native but haven't manually tested packages for: i686, x86_64-musl
• I built this PR locally for these architectures (all crossbuilds):
  • aarch64
  • aarch64-musl
  • armv7l
  • armv7l-musl
  • armv6l
  • armv6l-musl

Native builds on aarch64 and aarch64-musl were attempted. Failed only with telegram-desktop, see comment for details: #57903 (comment)
I've tried to disentangle their build system but couldn't. Cross builds just fine.

Notable changes:

• abseil-cpp has been chosen of the version that doesn't have newest breaking changes (from May of this year instead of more recent, August release).
• Bear kept trying to build and run tests in cross (due to etc/conf having enabled tests or flag passing), disabled.
• telegram-desktop has libdispatch added (it'll fallback to a qt implementation for similar functionality instead, but libdispatch seems to be preferred by upstream), it consistently failed to build cross to aarch64-musl if done without including glib-devel in makedepends instead of only libglib-devel. Not sure why this works in ci.
• easyeffects is no longer available for armv6l and armv7l (see comment)
• marble failed two tests on i686. Added exclusion for both, as I've been unable to figure out the culprit. FWIW, it fails same tests without the changes introduced in this PR. added sse2 flags, seems to pass tests now.
• rnnoise reverts 0.4.1 because upstream didn't have versions, so it was invented based on SONAME: 04b2182#r659096017
  The commit used in current in-repo template is 4 years old at this point and that version lacks a function that easyeffects have just decided to use, so builds failed without the update. I tested building rnnoise's rev-deps: alsa_rnnoise works, telegram-desktop works, easyeffects works. pulseeffects-legacy currently fails to build due to an unrelated issue, marked broken for now.

ci ran succeeded prior to additional changes applied to: rnnoise, easyeffects, marble, pulseffects-legacy, alsa_rnnoise, sysdig (has been changed to only revbump after recent update)
[ci skip] due to being too big

[zen0bit]

Why not easyeffects 8.0.3?

[zen0bit]

Update commit message of easyeffects to easyeffects: update to 8.0.3, change to qt6

Also re2 could be bumped to 2025.11.05?

[JkktBkkt]

commit message

Fixed with ci skipped push above, ty

re2

I think avoiding growing in complexity too much and instead updating it in a separate PR (if necessary) would be better though. grpc, python3-grpcio, protonmail-bridge, protobuf-c, libphonenumber, netdata (repo version of it is >2 years old) all have updates as well, but can't really vouch since didn't test them personally.

[JkktBkkt]

Most recent push changed telegram-desktop to be just a revbump + cross deps fix, since update to 6.3.1 has been merged separately.

[JkktBkkt]

Re-ordered commits a bit, fixed common/shlibs merge conflict

[JkktBkkt]

Found something to fix marble tests on i686, unsure if this is the right path to take, but here we go
And more re-ordering of some commits

[JkktBkkt]

Easyeffects 8.0.6 is out, rebuilt for all, tested on x86_64
removed the check to not build tests from Bear template (changes for it are now just a revbump)