[Bug] - Android 8 [end of support?]

[vitorsaldiva]

Description

Any reason to stop supporting the Android 8 from 1.11 onwards?

[ExTV]

Hey! Not a deliberate deprecation — it was a side-effect of fixing a QEMU crash on Android 12 (OnePlus 7 Pro). QEMU was built against the API-34 sysroot, which made it link against copy_file_range, a symbol missing from older Bionic. The fix was to drop the QEMU build sysroot to API 28, and the app's minSdk had to follow it up from 26 → 28 to match.

In theory we could rebuild QEMU against the API-26 sysroot and bring Android 8 back, but it's untested since v1.0.8 and I don't have an Android 8 device to validate. If you (or anyone) can test on real Android 8 hardware, happy to look at it.

[vitorsaldiva]

Hey @ExTV, I have an Asus Zenfone 4 (ZE554KL), a real Android 8, and we can test it on it. Let me know when we can start working on this!

[ExTV]

Tried again in 1.2.0: dropped minSdk 28 → 26 and re-targeted the NDK sysroot to aarch64-linux-android26 in Dockerfile. Build fails in the glib step:

meson.build:2200:13: ERROR: Dependency "iconv" not found (tried builtin and system)

Bionic only exposes iconv_open / iconv / iconv_close from API 28+, so glib's built-in detector finds nothing on the API-26 sysroot. Adding GNU libiconv statically as an upstream dep doesn't work either — tried both 1.16 and 1.17 and both fail to compile against API 26 because gnulib's srclib/stdint.h defines _STDINT_MAX(signed, bits, zero), and recent Clang refuses (signed) as a value (it parses as a parenthesised type cast → 6–12 errors per TU in careadlinkat.o).

So API-26 support needs one of: a patched gnulib (sed-rename the signed macro parameter), a hand-rolled iconv shim, or musl-iconv vendored in. Deferring for 1.2.0 — it ships with minSdk = 28 unchanged. Leaving this open; happy to look at it again once one of those workarounds is sketched.