Android library failing to build

[ThreeDeeJay]

Some extra context: #824
I put together a GitHub actions workflow to build the Android library (libopenal.so) and noticed that a commit is causing some compilation errors. After some trial and error, I managed to track down the culprit, from June 1st:

• a0e238f - Last commit to successfully compile
• 2719ce2 - Commit where compilation started failing

[ThreeDeeJay]

On a side note, I'm not sure if it's related, but the the one that did build crashes sView with this error when enabling HRTF in the settings, unlike Centurion96's build which works, though I think it lacks embedded HRTF which might be a clue. 🤔

Here's the log starting right before enabling HRTF:

D com.sview: PlayerBase::stop() from IPlayer
D AudioTrack: stop(186): called with 477126 frames delivered
D openal: [ALSOFT] (II) ALC_OUTPUT_MODE_SOFT = 6578
D openal: [ALSOFT] (II) ALC_HRTF_SOFT = 2
D openal: [ALSOFT] (II) Pre-reset: *Stereo, Float32, 48000hz, 960 / 2880 buffer
D com.sview: PlayerBase::PlayerBase()
D com.sview: TrackPlayerBase::TrackPlayerBase()
I libOpenSLES: Emulating old channel mask behavior (ignoring positional mask 0x3, using default mask 0x3 based on channel count of 2)
D openal: [ALSOFT] (II) Post-reset: Stereo, Int16, 48000hz, 960 / 2880 buffer
D openal: [ALSOFT] (II) Searching / for *.mhr
E openal: [ALSOFT] (EE) Exception enumerating files: filesystem error: in directory_iterator::directory_iterator(...): Permission denied ["/"]
D openal: [ALSOFT] (II) Adding built-in entry "!1_Built-In HRTF"
D openal: [ALSOFT] (II) Loading !1_Built-In HRTF...
D openal: [ALSOFT] (II) Detected data set format v3
W StAudioQueue: type=1400 audit(0.0:9583): avc:  denied  { read } for  name="/" dev="dm-1" ino=2 scontext=u:r:untrusted_app_29:s0:c98,c257,c512,c768 tcontext=u:object_r:rootfs:s0 tclass=dir permissive=0 ppid=1624 pcomm="main" pgid=25685 pgcomm="com.sview" app=com.sview
D openal: [ALSOFT] (II) Loaded HRTF Built-In HRTF for sample rate 48000hz, 64-sample filter
D openal: [ALSOFT] (II) 1st order + Full HRTF rendering enabled, using "Built-In HRTF"
D openal: [ALSOFT] (II) Channel config, Main: 4, Real: 2
D openal: [ALSOFT] (II) Allocating 6 channels, 24576 bytes
D openal: [ALSOFT] (II) Min delay: 7.75, max delay: 32.50, FIR length: 64
D openal: [ALSOFT] (II) New max delay: 24.75, FIR length: 89
D openal: [ALSOFT] (II) Max sources: 256 (255 + 1), effect slots: 64, sends: 2
D openal: [ALSOFT] (II) Dithering enabled (16-bit, 32768)
D openal: [ALSOFT] (II) Output limiter enabled, -0.0005dB limit
D openal: [ALSOFT] (II) Fixed device latency: 1000000ns
D openal: [ALSOFT] (II) Post-start: Stereo, Int16, 48000hz, 960 / 2880 buffer
W openal: [ALSOFT] (WW) pthread_setschedparam failed: Operation not permitted (1)
W openal: [ALSOFT] (WW) D-Bus not available
F libc: Fatal signal 7 (SIGBUS), code 1 (BUS_ADRALN), fault addr 0xe4581a38 in tid 25793 (alsoft-mixer), pid 25685 (com.sview)
I crash_dump32: obtaining output fd from tombstoned, type: kDebuggerdTombstoneProto
I crash_dump32: performing dump of process 25685 (target tid = 25793)
W System: A resource failed to call AbstractCursor.close.
W System: A resource failed to call CursorWrapperInner.close.
W System: A resource failed to call AbstractCursor.close.
W System: A resource failed to call CursorWrapperInner.close.
W crash_dump32: type=1400 audit(0.0:9584): avc:  denied  { read } for  name="/" dev="dm-1" ino=2 scontext=u:r:crash_dump:s0:c98,c257,c512,c768 tcontext=u:object_r:rootfs:s0 tclass=dir permissive=0 ppid=1 pcomm="init" pgid=1 pgcomm="init" app=com.sview
W crash_dump32: type=1400 audit(0.0:9585): avc:  denied  { read } for  name="/" dev="dm-1" ino=2 scontext=u:r:crash_dump:s0:c98,c257,c512,c768 tcontext=u:object_r:rootfs:s0 tclass=dir permissive=0 ppid=1 pcomm="init" pgid=1 pgcomm="init" app=com.sview
W crash_dump32: type=1400 audit(0.0:9586): avc:  denied  { read } for  name="/" dev="dm-1" ino=2 scontext=u:r:crash_dump:s0:c98,c257,c512,c768 tcontext=u:object_r:rootfs:s0 tclass=dir permissive=0 ppid=1 pcomm="init" pgid=1 pgcomm="init" app=com.sview
W crash_dump32: type=1400 audit(0.0:9587): avc:  denied  { read } for  name="/" dev="dm-1" ino=2 scontext=u:r:crash_dump:s0:c98,c257,c512,c768 tcontext=u:object_r:rootfs:s0 tclass=dir permissive=0 ppid=1 pcomm="init" pgid=1 pgcomm="init" app=com.sview
F DEBUG: *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
F DEBUG: backtrace:
F DEBUG: #00 pc 000d0b38  /data/app/~~KvOx4ElXIr_v-_FM4vcxzA==/com.sview-Us29JoLSfw4y9PYkPDCKbA==/lib/arm/libopenal.so (void MixDirectHrtf_<NEONTag>(al::span<float, 1024u>, al::span<float, 1024u>, al::span<std::__ndk1::array<float, 1024u> const, 4294967295u>, al::span<std::__ndk1::array<float, 2u>, 4294967295u>, al::span<float, 1024u>, al::span<HrtfChannelState, 4294967295u>, unsigned int, unsigned int)+96) (BuildId: 5929b6a6691adfe4f6e8e384245da208e26cefc1)
F DEBUG: #01 pc 00097951  /data/app/~~KvOx4ElXIr_v-_FM4vcxzA==/com.sview-Us29JoLSfw4y9PYkPDCKbA==/lib/arm/libopenal.so (DeviceBase::ProcessHrtf(unsigned int)+128) (BuildId: 5929b6a6691adfe4f6e8e384245da208e26cefc1)
F DEBUG: #02 pc 000993d3  /data/app/~~KvOx4ElXIr_v-_FM4vcxzA==/com.sview-Us29JoLSfw4y9PYkPDCKbA==/lib/arm/libopenal.so (DeviceBase::renderSamples(unsigned int)+6354) (BuildId: 5929b6a6691adfe4f6e8e384245da208e26cefc1)
F DEBUG: #03 pc 0009985d  /data/app/~~KvOx4ElXIr_v-_FM4vcxzA==/com.sview-Us29JoLSfw4y9PYkPDCKbA==/lib/arm/libopenal.so (DeviceBase::renderSamples(void*, unsigned int, unsigned int)+132) (BuildId: 5929b6a6691adfe4f6e8e384245da208e26cefc1)
F DEBUG: #04 pc 000b743f  /data/app/~~KvOx4ElXIr_v-_FM4vcxzA==/com.sview-Us29JoLSfw4y9PYkPDCKbA==/lib/arm/libopenal.so ((anonymous namespace)::OpenSLPlayback::mixerProc()+326) (BuildId: 5929b6a6691adfe4f6e8e384245da208e26cefc1)
F DEBUG: #05 pc 000b769b  /data/app/~~KvOx4ElXIr_v-_FM4vcxzA==/com.sview-Us29JoLSfw4y9PYkPDCKbA==/lib/arm/libopenal.so (void* std::__ndk1::__thread_proxy[abi:v170000]<std::__ndk1::tuple<std::__ndk1::unique_ptr<std::__ndk1::__thread_struct, std::__ndk1::default_delete<std::__ndk1::__thread_struct> >, std::__ndk1::__mem_fn<int ((anonymous namespace)::OpenSLPlayback::*)()>, (anonymous namespace)::OpenSLPlayback*> >(void*)+54) (BuildId: 5929b6a6691adfe4f6e8e384245da208e26cefc1)
F DEBUG: #06 pc 00044323  /apex/com.android.runtime/lib/bionic/libc.so (__pthread_start(void*)+40) (BuildId: 792a626fd6ec5a902e0f37973978b310)
F DEBUG: #07 pc 0003b42b  /apex/com.android.runtime/lib/bionic/libc.so (__start_thread+30) (BuildId: 792a626fd6ec5a902e0f37973978b310)

[kcat]

Some extra context: #824 I put together a GitHub actions workflow to build the Android library (libopenal.so) and noticed that a commit is causing some compilation errors. After some trial and error, I managed to track down the culprit, from June 1st:

* [a0e238f](https://github.com/kcat/openal-soft/commit/a0e238f54dbf57a2c88f2b45a0c49f6a13e0784c) - Last commit to [successfully compile](https://github.com/ThreeDeeJay/openal-soft/actions/runs/9703301432/job/26781110464#step:2:7)

* [2719ce2](https://github.com/kcat/openal-soft/commit/2719ce2c0aaf40bdf2db1c3e579b3b76e9cbcf14) - Commit where compilation started [failing](https://github.com/ThreeDeeJay/openal-soft/actions/runs/9703265481/job/26780989241#step:5:107)

Odd. Seems to be failing because libopenal.version specifies functions to export/make global that were deprecated. They still exist and should be exportable, they're marked with __attribute__((visibility("protected"))) to be globally visible (and not internally overridable) as they should be, but for some reason the linker isn't seeing them. Perhaps the linker is discarding them since they're not referenced elsewhere, though they're not the only functions like that. I'll see what's different about them and try something.

On a side note, I'm not sure if it's related, but the the one that did build crashes sView with this error when enabling HRTF in the settings, unlike Centurion96's build which works, though I think it lacks embedded HRTF which might be a clue. 🤔

It's not related to the build failure, but it is related to HRTF. It seems to crash in MixDirectHrtf_<NEONTag> as a result of a bad address alignment, given the address 0xe4581a38. Unfortunately it doesn't say what that address belongs to, or the faulting instruction. AFAIK, NEON should be more robust with address alignment, and the given address is 8-byte aligned that loads and stores should be able to use. I'm not sure why it's crashing like that.

[ThreeDeeJay]

It's not related to the build failure, but it is related to HRTF. It seems to crash in MixDirectHrtf_<NEONTag> as a result of a bad address alignment, given the address 0xe4581a38. Unfortunately it doesn't say what that address belongs to, or the faulting instruction. AFAIK, NEON should be more robust with address alignment, and the given address is 8-byte aligned that loads and stores should be able to use. I'm not sure why it's crashing like that.

Interesting, setting ALSOFT_EMBED_HRTF_DATA=FALSE prevents the crash, which I guess explains why Centurion's build (which also lacked embedded HRTF) didn't crash either.

Here's a more verbose log, comparing a0e238f with and without ALSOFT_EMBED_HRTF_DATA where I just started sView and enabled HRTF: https://www.diffchecker.com/zpmHlWaC/

BTW in case it's relevant, I'm running the 32-bit version of sView on a 64-bit OS/device.

[kcat]

Yes, disabling ALSOFT_EMBED_HRTF_DATA prevents the library from having a built-in HRTF dataset, which prevents HRTF from working if there's no external mhr files it can load load. The app resets the device setting ALC_OUTPUT_MODE_SOFT, ALC_STEREO_HRTF_SOFT to try to enable HRTF, which fails and it falls back to normal stereo. With ALSOFT_EMBED_HRTF_DATA on, it can enable HRTF processing using the built-in dataset, which eventually calls MixDirectHrtf_<NEONTag>, and then crashes for some reason.

[kcat]

Is the HRTF crash fixed now with commit a56689d?

[ThreeDeeJay]

Thanks, it's not crashing anymore! 👌
https://youtu.be/ssmLgkbRMh0