Android buildbot failures after updating to API level 35

[mhsmith]

Bug report

Bug description:

This update happened in #142289, which works fine in GitHub Actions, but caused different problems in the two Android buildbots.

---

On the aarch64 buildbot, the new system image failed to start, with its logs showing a repeated loop of system services crashing and restarting:

12-05 14:50:56.223  4745  4745 F DEBUG   : *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
12-05 14:50:56.223  4745  4745 F DEBUG   : Build fingerprint: 'Android/sdk_slim_arm64/emulator64_arm64:14/UE1A.230829.036.A1/11228894:userdebug/test-keys'
12-05 14:50:56.223  4745  4745 F DEBUG   : Revision: '0'
12-05 14:50:56.223  4745  4745 F DEBUG   : ABI: 'arm64'
12-05 14:50:56.223  4745  4745 F DEBUG   : Timestamp: 2025-12-05 14:50:56.190712855+0000
12-05 14:50:56.223  4745  4745 F DEBUG   : Process uptime: 1s
12-05 14:50:56.223  4745  4745 F DEBUG   : Cmdline: /system/bin/surfaceflinger
12-05 14:50:56.223  4745  4745 F DEBUG   : pid: 4738, tid: 4742, name: surfaceflinger  >>> /system/bin/surfaceflinger <<<
12-05 14:50:56.223  4745  4745 F DEBUG   : uid: 1000
12-05 14:50:56.223  4745  4745 F DEBUG   : tagged_addr_ctrl: 0000000000000001 (PR_TAGGED_ADDR_ENABLE)
12-05 14:50:56.223  4745  4745 F DEBUG   : pac_enabled_keys: 000000000000000f (PR_PAC_APIAKEY, PR_PAC_APIBKEY, PR_PAC_APDAKEY, PR_PAC_APDBKEY)
12-05 14:50:56.223  4745  4745 F DEBUG   : signal 6 (SIGABRT), code -1 (SI_QUEUE), fault addr --------
12-05 14:50:56.223  4745  4745 F DEBUG   : Abort message: 'EGLContext creation failed'
12-05 14:50:56.223  4745  4745 F DEBUG   :     x0  0000000000000000  x1  0000000000001286  x2  0000000000000006  x3  000000785e5ba4a0
12-05 14:50:56.223  4745  4745 F DEBUG   :     x4  0000000000000010  x5  0000000000000010  x6  0000000000000010  x7  7f7f7f7f7f7f7f7f
12-05 14:50:56.223  4745  4745 F DEBUG   :     x8  00000000000000f0  x9  0000007afdc33090  x10 0000000000000001  x11 0000007afdc86058
12-05 14:50:56.223  4745  4745 F DEBUG   :     x12 000000785e5ba470  x13 000000000000001b  x14 000000785e5ba630  x15 00014295acd0344e
12-05 14:50:56.223  4745  4745 F DEBUG   :     x16 0000007afdcf3d08  x17 0000007afdcc7e90  x18 000000785d54c000  x19 0000000000001282
12-05 14:50:56.223  4745  4745 F DEBUG   :     x20 0000000000001286  x21 00000000ffffffff  x22 000000785e5babd8  x23 0000000000000002
12-05 14:50:56.223  4745  4745 F DEBUG   :     x24 0000000000000000  x25 0000000000000000  x26 000000785e5baff8  x27 00000000000fc000
12-05 14:50:56.223  4745  4745 F DEBUG   :     x28 00000000000fe000  x29 000000785e5ba520
12-05 14:50:56.223  4745  4745 F DEBUG   :     lr  0000007afdc779b8  sp  000000785e5ba480  pc  0000007afdc779e4  pst 0000000000001000

I guessed this was caused by the emulator not being compatible with the emulator system image, so I installed the current emulator version 36.3.10, but it made no difference. Neither did 36.2.12, which is what I'd used to test the update on my own aarch64 Mac, or 35.6.11, which is the last in the 35 series. Unfortunately I didn't keep a note of what the previously installed version was.

The same thing happens with several other system images, even the API level 34 one we were previously using, either from the existing snapshot or a freshly created one.

I've taken the aarch64 buildbot offline until this is fixed. The next thing to try might be moving the whole SDK and .android directory out of the way and recreating them from scratch.

---

On the x86_64 buildbot, the tests are running to completion, but giving errors about not being able to contact pythontest.net. It's possible this is caused by an emulator / system image incompatibility, but be careful in case the same thing happens here.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Other