[BUG] NDK r28 beta1: issues with existing polyfills

[dg0yt]

Description

This is in response to the r28 beta1 release notes.

If your project contains polyfills for any of those APIs, this change may
break your build due to the conflicting declarations. The simplest fix is to
rename your polyfill to not collide with libc. For example, rename
conflicting_api to conflicting_api_fallback and call that instead. Use
#define conflicting_api() conflicting_api_fallback() if you want to avoid
rewriting callsites.

Please open a bug if you run into issues with existing polyfills. We may be
able to add the polyfill directly to the NDK.

A test build in vcpkg CI (which uses android 21) shows a lot of breakage related to __INTRODUCED_IN() with polyfills from gnulib. There is no full picture yet because the build error cascade starts early. The first hit is mempcpy in GNU libiconv. I added a patch to rename the polyfill, but the next failures is already gettext-libintl.

In the non-gnulib group, openssl is affected, removing many ports from the build. For poppler, I assume that GNU libiconv used to be the polyfill, but I didn't verify that. "libiconv as a polyfill" might be a broader pattern.

I grepped the arm64 logs for __INTRODUCED_IN. Transformed and compressed:

gettext-libintl 'feof_unlocked' is unavailable: introduced in Android 23
gettext-libintl 'fgets_unlocked' is unavailable: introduced in Android 28
gettext-libintl 'mblen' is unavailable: introduced in Android 26
gettext-libintl 'mempcpy' is unavailable: introduced in Android 23
gettext-libintl 'nl_langinfo' is unavailable: introduced in Android 26
gsasl 'error_at_line' is unavailable: introduced in Android 23
gsasl 'error' is unavailable: introduced in Android 23
gsasl 'error_message_count' is unavailable: introduced in Android 23
gsasl 'error_print_progname' is unavailable: introduced in Android 23
gsasl 'fflush_unlocked' is unavailable: introduced in Android 28
gsasl 'fputs_unlocked' is unavailable: introduced in Android 28
gsasl '__freading' is unavailable: introduced in Android 28
gsasl '__fsetlocking' is unavailable: introduced in Android 23
gsasl 'getrandom' is unavailable: introduced in Android 28
gsasl 'iconv_close' is unavailable: introduced in Android 28
gsasl 'iconv' is unavailable: introduced in Android 28
gsasl 'iconv_open' is unavailable: introduced in Android 28
gsasl 'mblen' is unavailable: introduced in Android 26
gsasl 'nl_langinfo' is unavailable: introduced in Android 26
libconfuse 'fmemopen' is unavailable: introduced in Android 23
libidn2 'error_at_line' is unavailable: introduced in Android 23
libidn2 'error' is unavailable: introduced in Android 23
libidn2 'error_message_count' is unavailable: introduced in Android 23
libidn2 'error_print_progname' is unavailable: introduced in Android 23
libidn2 'nl_langinfo' is unavailable: introduced in Android 26
libidn2 'strchrnul' is unavailable: introduced in Android 24
liblsl 'pthread_getname_np' is unavailable: introduced in Android 26
libuv 'freeifaddrs' is unavailable: introduced in Android 24
libuv 'getifaddrs' is unavailable: introduced in Android 24
libuv 'preadv' is unavailable: introduced in Android 24
libuv 'pthread_barrier_destroy' is unavailable: introduced in Android 24
libuv 'pthread_barrier_init' is unavailable: introduced in Android 24
libuv 'pthread_barrier_wait' is unavailable: introduced in Android 24
libuv 'pwritev' is unavailable: introduced in Android 24
openssl 'getentropy' is unavailable: introduced in Android 28
poppler 'iconv_close' is unavailable: introduced in Android 28
poppler 'iconv' is unavailable: introduced in Android 28
poppler 'iconv_open' is unavailable: introduced in Android 28

Patching gnutext-libintl would probably add many more.
(All those packages build with NDK r26. NDK r27 is still blocked by its bugs.)

Affected versions

r28

Canary version

beta1

Host OS

Linux

Host OS version

vcpkg CI image

Affected ABIs

arm64-v8a

Build system

Other (specify below)

Other build system

No response

minSdkVersion

21

Device API level

No response

[dg0yt]

In gettext libintl, the mempcpy related errors also include:

[..]gettext-runtime/intl/dcigettext.c:1723:1: error: static declaration of 'mempcpy' follows non-static declaration

[DanAlbert]

Ooh, if vcpkg's CI has issues that gives us a great place to go look at the scope of the problem. Thanks! I should remember to try that whenever I'm making ambitious changes in the future.

We'll have a look and see what can be done. My gut says that we'll re-hide the stuff that's causing problems for r28 and work on upstreaming fixes. If it were a shorter list of projects I'd say we should just fix those projects and ship r28 as-is, but if it's hit a non-trivial number of projects just in vcpkg, that's a lot of patches to upstream, and a lot of package updates that developers would need to take to get those fixes.

A complete rollback is an option, but it's sort of the last resort. For the reasons mentioned in the changelog, this is a step in the right direction that we want to take, so I'd like to see us make progress wherever we can do so safely. I'm okay with gradual progress though, so if we need to re-hide the stuff that's broadly polyfilled already we can do that to avoid breaks while we fix the various open source projects to be compatible with the change.

[DanAlbert]

A test build in vcpkg CI (which uses android 21) shows a lot of breakage related to __INTRODUCED_IN() with polyfills from gnulib.

How did you do this? I was expecting to find a GitHub workflow for this, or instructions on how to build all the ports in their docs, but I'm not finding either.

[DanAlbert]

I eventually figured this out: ANDROID_NDK_HOME=/path/to/ndk vcpkg ci --triplet=arm64-android, which seems to do what I need.

[DanAlbert]

This is very slow though so if there's a smarter way to do this please lmk.

[enh-google]

i agree with danalbert's plan to quickly fix all this by just adding #if version checks...

...but for the longer-term solution, here's a quick brain dump in case i go under a bus:

gettext-libintl 'feof_unlocked' is unavailable: introduced in Android 23
gettext-libintl 'fgets_unlocked' is unavailable: introduced in Android 28
gsasl 'fflush_unlocked' is unavailable: introduced in Android 28
gsasl 'fputs_unlocked' is unavailable: introduced in Android 28
gsasl '__freading' is unavailable: introduced in Android 28
gsasl '__fsetlocking' is unavailable: introduced in Android 23

i don't really see any way to polyfill any of the unlocked stuff unless we re-expose the FILE implementation details. which maybe we should, but (a) that's a big decision and (b) we should definitely see what the wasm plan looks like in that area first. (since they might be the first actual beneficiaries of the increased opacity.)

libconfuse 'fmemopen' is unavailable: introduced in Android 23

large but doable, modulo similar FILE-internals leakage.

gettext-libintl 'mempcpy' is unavailable: introduced in Android 23

trivial one-liner. we should probably have done this as a polyfill in the first place and stopped there given that memcpy() can even be optimized to direct loads/stores, but mempcpy() is definitely going to cost you two function calls. where it should really just cost you an add.

gettext-libintl 'mblen' is unavailable: introduced in Android 26

so believe it or not, mblen() is a trivial wrapper for mbrlen() which is freely available. so surprisingly, this should be a trivial polyfill too.

gettext-libintl 'nl_langinfo' is unavailable: introduced in Android 26

this might seem like an obvious "no", but it's one where i wonder whether this is actually a great candidate for "let's just polyfill all the time". why? because Android has hard-coded answers to all the questions. and because every call i've ever seen asks a hard-coded question. so i'd expect that clang would be able to turn nl_langinfo("AM_STR") into "AM" or whatever. (of course, this probably isn't as exciting as it sounds, because in practice nothing is using this for anything that actually matters.)

gsasl 'error_at_line' is unavailable: introduced in Android 23
gsasl 'error' is unavailable: introduced in Android 23
gsasl 'error_message_count' is unavailable: introduced in Android 23
gsasl 'error_print_progname' is unavailable: introduced in Android 23

the reliance on globals here makes this a "no".

gsasl 'iconv_close' is unavailable: introduced in Android 28
gsasl 'iconv' is unavailable: introduced in Android 28
gsasl 'iconv_open' is unavailable: introduced in Android 28

even our limited set of supported encodings is probably too big for this to make sense.

libidn2 'strchrnul' is unavailable: introduced in Android 24

as long as we don't care about not having the optimized arm64 implementation (which is kind of a moot point if your alternative is "nothing", as here), there's a trivial implementation of this.

liblsl 'pthread_getname_np' is unavailable: introduced in Android 26

certainly doable, though it's a weak symbol so that native bridge can override it, so doing so would likely cause a bit of trouble there.

libuv 'freeifaddrs' is unavailable: introduced in Android 24
libuv 'getifaddrs' is unavailable: introduced in Android 24

nightmare.

libuv 'pthread_barrier_destroy' is unavailable: introduced in Android 24
libuv 'pthread_barrier_init' is unavailable: introduced in Android 24
libuv 'pthread_barrier_wait' is unavailable: introduced in Android 24

this doesn't seem to touch any internal pthread state, and it's all relatively self-contained and based on <stdatomic.h>, so i think this is something we could do ... it would leak the currently-private internal implementation of pthread_barrier_t, though i suspect we could always cheat there by giving the polyfill struct and functions slightly different names?

libuv 'preadv' is unavailable: introduced in Android 24
libuv 'pwritev' is unavailable: introduced in Android 24

conceptually easy, but really annoying as long as we're still supporting ILP32/32-bit off_t. which, realistically, is something we'll be doing for longer than we're still supporting API 24. certainly doable though. just gross.

gsasl 'getrandom' is unavailable: introduced in Android 28
openssl 'getentropy' is unavailable: introduced in Android 28

getentropy() relies on getrandom() under the covers, and although that's been around since 3.17, it was only added to bionic at the end of 2017, so only 2018+ devices will have a seccomp policy that will let you use these. TL;DR: not likely.

[dg0yt]

A test build in vcpkg CI (which uses android 21) shows a lot of breakage related to __INTRODUCED_IN() with polyfills from gnulib.

How did you do this? I was expecting to find a GitHub workflow for this, or instructions on how to build all the ports in their docs, but I'm not finding either.

It a draft PR, microsoft/vcpkg#41293. It is in the timeline here. The failure logs are available as "assets" from Azure Pipelines.

I eventually figured this out: ANDROID_NDK_HOME=/path/to/ndk vcpkg ci --triplet=arm64-android, which seems to do what I need.

Basically yes, but vcpkg CI also skips known failures with --ci-baseline=scripts/ci.baseline.txt.

This is very slow though so if there's a smarter way to do this please lmk.

Of course ... vcpkg builds debug+release, and you start with empty caches for assets and artifacts.

You can look at individual ports with vcpkg install <port> --triplet=arm64-android.

Maybe you can limit the build to the release config with --cmake-args=-DVCPKG_BUILD_TYPE=release.

[DanAlbert]

I definitely looked for PRs... I must have accidentally done that in the wrong tab with a different repo. Thanks.

This is very slow though so if there's a smarter way to do this please lmk.

Of course

Ah, okay. If this is just a thing that's slow, alright.

[dg0yt]

Note that when the Microsoft team added Android, they didn't fix all the existing ports. scripts/ci.baseline.txt really must be applied. Otherwise you will meet ports which will make you cry.

[DanAlbert]

Oh, I'd misunderstood you before and thought that was applied automatically. That explains why I've found mostly stuff that didn't work with r27 either.