Add Android platform support by tonyfettes · Pull Request #306 · moonbitlang/async

tonyfettes · 2026-02-28T09:39:19Z

Summary

pidfd fallback: wait_pid now tries pidfd_open first, falls back to blocking waitpid in a worker thread when unavailable (graceful degradation on any Linux kernel < 5.3, not just Android)
addchdir_np runtime guard: On Android, checks availability via dlsym at runtime (API 34+ only), returns ENOSYS if unavailable
Thread stack size: Sets 64KB minimum on Android (the default 512 bytes is below Android's enforced minimum)
TLS/BoringSSL compat: Android-specific libssl.so loading, SSLeay fallback for version detection, SSL_CTX_load_verify_locations for /system/etc/security/cacerts/
Temp directory: Replaces hardcoded /tmp/ with a C function that checks $TMPDIR and falls back to /data/local/tmp/ on Android

Approach

All Android-specific logic is in C via #ifdef __ANDROID__. MoonBit code stays #cfg(not(platform="windows")) for Unix — no MoonBit-level Android guards needed.

Test plan

moon check passes
moon build --target native compiles cleanly
moon test --target native — all tests pass (no regressions)
Cross-compile with Android NDK for aarch64 and verify linking
Run on Android emulator: event loop, TCP sockets, file I/O, process spawn, process wait
Verify TLS loading behavior on Android
Verify posix_spawn with cwd on API < 34 returns clear ENOSYS error

🤖 Generated with Claude Code

coveralls · 2026-02-28T09:42:44Z

Pull Request Test Coverage Report for Build 92

Details

22 of 34 (64.71%) changed or added relevant lines in 3 files are covered.
No unchanged relevant lines lost coverage.
Overall coverage decreased (-0.2%) to 80.683%

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/internal/event_loop/process_unix.mbt	7	13	53.85%
src/process/process.mbt	14	20	70.0%

Totals
Change from base Build 86:	-0.2%
Covered Lines:	2243
Relevant Lines:	2780

💛 - Coveralls

tonyfettes · 2026-02-28T14:26:25Z

src/tls/openssl.c

+  // Android's /system/etc/security/cacerts/ uses MD5-based hash filenames,
+  // but OpenSSL 3.x directory lookup expects SHA1-based hashes. Bypass hash
+  // lookup by iterating the directory and loading each cert directly.
+  if (PEM_read_X509 && X509_STORE_add_cert && X509_free) {
+    X509_STORE *store = SSL_CTX_get_cert_store(client_ctx);
+    DIR *dir = opendir("/system/etc/security/cacerts");
+    if (dir) {
+      struct dirent *entry;
+      char path[512];
+      while ((entry = readdir(dir)) != NULL) {
+        if (entry->d_name[0] == '.') continue;
+        snprintf(path, sizeof(path), "/system/etc/security/cacerts/%s", entry->d_name);
+        FILE *fp = fopen(path, "r");
+        if (fp) {
+          X509 *cert = PEM_read_X509(fp, NULL, NULL, NULL);
+          if (cert) {
+            X509_STORE_add_cert(store, cert);
+            X509_free(cert);
+          }
+          fclose(fp);
+        }
+      }
+      closedir(dir);
+    }


We iterate the directory and loads all certificates here. The "correct" approach would be hook into JNI to get certificates, but it is way to complex here.

tonyfettes · 2026-02-28T15:11:55Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a619abcc2c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

src/internal/event_loop/thread_pool.c

- pidfd fallback: try pidfd_open, fall back to blocking waitpid in worker thread when unavailable (benefits all Linux < 5.3) - guard posix_spawn_file_actions_addchdir_np with runtime dlsym check on Android (only available on API 34+, returns ENOSYS otherwise) - set 64KB thread stack size on Android (512 bytes is below minimum) - TLS: add Android libssl.so dlopen path, BoringSSL SSLeay version compat, SSL_CTX_load_verify_locations for /system/etc/security/cacerts/ - replace hardcoded /tmp/ with C function that checks $TMPDIR and falls back to /data/local/tmp/ on Android Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

openssl.c: Android's /system/etc/security/cacerts/ uses MD5-based hash filenames (OpenSSL 1.0 era), but OpenSSL 3.x directory lookup expects SHA1-based hashes. The previous code relied on SSL_CTX_set_default_verify_paths (which always returns 1) with a fallback to SSL_CTX_load_verify_locations using the hash-based CApath — both fail due to the hash mismatch. Fix by dlopen'ing libcrypto.so for PEM_read_X509/X509_STORE_add_cert/X509_free, then iterating the Android CA cert directory and loading each cert directly into the X509_STORE, bypassing hash-based lookup entirely. thread_pool.c: Guard #include <spawn.h> and all posix_spawn usage with #if !defined(__ANDROID__) || __ANDROID_API__ >= 28. Provide a stub moonbitlang_async_make_spawn_job that returns NULL on older Android, eliminating the need for per-app posix_spawn_compat.c workarounds. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

When posix_spawn is unavailable (Android API < 28), make_spawn_job previously returned NULL, causing the worker thread to exit silently without sending a completion notification. This left the caller in perform_job_in_worker waiting forever. Return a proper job struct with a worker function that sets err = ENOSYS, so the caller receives a clear "Function not implemented" error. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

When pidfd_open is unavailable (Android, Linux <5.3), the fallback waitpid in the worker thread reaps the child but discards the exit status. The subsequent get_process_result then calls waitpid again, which fails with ECHILD. Store the exit code in job->ret during the fallback waitpid and propagate it back via wait_pid's return type (Unit -> Int?) so callers skip the redundant second waitpid. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…ory leak The #owned parameters (path, args, envp, cwd) were silently discarded without moonbit_decref(), leaking memory on every Job::spawn call on Android API < 28. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

tonyfettes · 2026-03-02T08:06:51Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c65cd3d809

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-02T08:11:13Z

src/internal/event_loop/process_unix.mbt

+      let exit_code = perform_job_in_worker(job, context~, cancel=_ => {
+        job.cancel_process_waiter()
+        NeedWait


Cancel fallback wait worker when task is cancelled

When register_pid fails (for example on kernels without pidfd_open), this fallback calls perform_job_in_worker with a cancel callback that only sets job.cancel_process_waiter() and returns NeedWait. In EventLoop::wait_for_job, NeedWait suppresses the cancellation until the worker completes, and because this callback never invokes evloop.cancel_job_in_worker(job_id, ...), the blocking waitpid is not interrupted, so @process.run/@process.spawn cancellation handlers do not run until the child exits naturally.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-03-02T08:11:13Z

src/fs/stub.c

+  const char *tmpdir = getenv("TMPDIR");
+  path = tmpdir ? tmpdir : "/data/local/tmp/";


Ensure TMPDIR has trailing slash before using as base path

This uses $TMPDIR verbatim on Android, but tmpdir() builds paths by directly concatenating tmp_base_path with the generated name, so a common value like /data/local/tmp (no trailing slash) produces malformed paths such as /data/local/tmpprefix... and can create directories in the wrong location or fail unexpectedly. Normalize the base path (or append / if missing) before returning it.

Useful? React with 👍 / 👎.

tonyfettes commented Feb 28, 2026

View reviewed changes

chatgpt-codex-connector bot reviewed Feb 28, 2026

View reviewed changes

src/internal/event_loop/thread_pool.c Show resolved Hide resolved

src/internal/event_loop/thread_pool.c Outdated Show resolved Hide resolved

tonyfettes and others added 6 commits March 2, 2026 16:06

chore: format

3e66da6

tonyfettes force-pushed the worktree-android branch from 289a601 to c65cd3d Compare March 2, 2026 08:06

chatgpt-codex-connector bot reviewed Mar 2, 2026

View reviewed changes

tonyfettes force-pushed the worktree-android branch from c65cd3d to 9746562 Compare March 2, 2026 09:09

refactor: use same make/free logic as unix

de8cd65

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add Android platform support#306

Add Android platform support#306
tonyfettes wants to merge 7 commits intomainfrom
worktree-android

tonyfettes commented Feb 28, 2026 •

edited

Loading

Uh oh!

coveralls commented Feb 28, 2026 •

edited

Loading

Uh oh!

tonyfettes Feb 28, 2026

Uh oh!

tonyfettes commented Feb 28, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Uh oh!

Uh oh!

Uh oh!

tonyfettes commented Mar 2, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Uh oh!

chatgpt-codex-connector bot Mar 2, 2026

Uh oh!

chatgpt-codex-connector bot Mar 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		const char *tmpdir = getenv("TMPDIR");
		path = tmpdir ? tmpdir : "/data/local/tmp/";

Conversation

tonyfettes commented Feb 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Approach

Test plan

Uh oh!

coveralls commented Feb 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull Request Test Coverage Report for Build 92

Details

💛 - Coveralls

Uh oh!

tonyfettes Feb 28, 2026

Choose a reason for hiding this comment

Uh oh!

tonyfettes commented Feb 28, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

tonyfettes commented Mar 2, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Mar 2, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot Mar 2, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

tonyfettes commented Feb 28, 2026 •

edited

Loading

coveralls commented Feb 28, 2026 •

edited

Loading