This commit allows -flto=thin to be added to the global CFLAGS, though
we don't enable it by default for the time being.

This patch enables certain wasm feature flags and optimizations that
we approve of.

This reverts commit 48eb92c1.

Update @bjorn3/browser_wasi_shim to version 0.4.2. This release includes
a limited polyfill for `poll_oneoff`, which is necessary for the
`pthread_cond` test to pass in the browser.
https://github.com/bjorn3/browser_wasi_shim/pull/88

The `pthread_cond` test uses `nanosleep` which eventually calls
`poll_oneoff` but `@bjorn3/browser_wasi_shim` does not implement it. So
we skip the test for now, but we will revisit it later once
https://github.com/bjorn3/browser_wasi_shim/pull/88 is merged.

We are heavily using wasi-libc + wasip1-threads on browsers. Since
wasi-libc uses `memory.atomic.wait32` to implement futex, and the usage
of `memory.atomic.wait32` on the main thread is prohibited on browsers,
we currently need to write code carefully not to reach the instruction.

Emscripten addresses this limitation by [employing a busy-wait on the
main
thread](https://github.com/emscripten-core/emscripten/blob/058a9fff/system/lib/pthread/emscripten_futex_wait.c#L111-L150)
as a workaround. Rust has [always employs busywait regardless of the
current thread is
main](https://github.com/rust-lang/rust/blob/a47555110cf09b3ed59811d9b02235443e76a595/library/std/src/sys/alloc/wasm.rs#L75-L144).
This approach, while effective for browsers, introduces unnecessary
overhead in environments where memory.atomic.wait32 is permitted.

This change provides a similar solution by introducing an opt-in
busy-wait mode for futexes. By making this an optional feature, we avoid
imposing the overhead of busy-waiting in non-browser environments while
ensuring compatibility with browser-based restrictions.

```c
#include <wasi/libc-busywait.h>

/// Enable busywait in futex on current thread.
void __wasilibc_enable_futex_busywait_on_current_thread(void);
```

This change slightly adds some runtime overheads in futex to check if we
should use busywait, but it can be optimized away as long as
`__wasilibc_enable_futex_busywait_on_current_thread` is not used by user
program and LTO is enabled.

This is a follow-up PR for #524 by implementing the following features:

- Embeds a minimal network services database with 17 common protocols,
suggested by @badeend. The database array is a weak symbol allowing
applications to override the default database at link time.
- Updates `getaddrinfo` to resolve named services in the address info.
For example `getaddrinfo("google.com", "https", NULL, &res);`
- Implements the `getservbyname` and `getservbyport` functions. These
functions are implemented using a static variable (`global_serv`), which
holds the returned service entry. This approach is acceptable because
these functions [are defined as not being
thread-safe](https://man.archlinux.org/man/getservbyname.3.en).

~~Additionally, this PR introduces an optional, more comprehensive
services database (`sockets_full_services_db.c`), based on Debian
Bookworm's `/etc/services` file (320 entries). To use this database,
link with the `-lc-full-services-db` flag.~~

There are a number of changes in this commit aimed at addressing #587
and making it easier to test #586 in CI. Notable changes here are:

* The matrix of what to test is much different from before. One matrix
entry now builds just one target and optionally tests that target.

* The CI matrix ensures that wasi-libc builds on a variety of platforms,
e.g. Windows/macOS/Linux as well as Linux aarch64.

* The CI matrix has one entry for building with an older version of
LLVM. This version was bumped from LLVM to LLVM 11 since LLVM is
installed through `apt-get`, not through downloads any more.

* On Linux LLVM/Clang are downloaded through `apt-get` instead of from
llvm-project binaries to avoid dealing with `libtinfo5` and
dependencies.

* The CI matrix has a test job per-target. This can be expanded/shrunk
as necessary but currently everything is tested with LLVM 16 (as before)
and only on Linux (also as before). The test run is seqeunced to happen
after the build of libc itself.

* The CI matrix has split out V8 headless tests into their own job to
avoid running multiple suites of tests in a single job.

* Installation of LLVM is refactored to a separate action to reduce the
noise in `main.yml`.

* Setting `TARGET_TRIPLE` can now be done through environment variables
as opposed to only through arguments to `make`.

* Handling of `BULITINS_LIB` has improved. Previously the build target
for `libc_so` would modify the compiler's resource directory and this is
updated to use a custom directory in `OBJDIR`.

* Arranging compiler-rt for tests is now done with `-resource-dir`
instead of copying the directory into the system compiler's location.

Overall it's the intention that no amount of testing is lost in this PR.
The goal is to expand things out in such a way that it's much easier to
add one-off tests of wasi-libc in various build configurations and such.
The theory is that this is as "simple" as adding a new matrix entry,
copied from previous ones, customized with various variables and
environment variables to affect the build (e.g. `CFLAGS`).

Closes #587

See the WHATSNEW file for details.

As described on the [musl homepage](https://musl.libc.org/), this also includes the two patches ([1](https://www.openwall.com/lists/musl/2025/02/13/1/1), [2](https://www.openwall.com/lists/musl/2025/02/13/1/2)) fixing  CVE-2025-26519. This fixes #577.

clang introduced a new warning (following gcc) for string initialization
to char arrays that do not fit the null terminal.

Introduced here: https://github.com/llvm/llvm-project/pull/137829
Included in LLVM version [LLVM
20.1.4](https://github.com/llvm/llvm-project/releases/tag/llvmorg-20.1.4)+

More detail:
-Wunterminated-string-initialization warns for c string literals that
are written to a buffer too small to include a null terminator. This
warning is also included in -Wextra which is enabled for wasi-libc and
fails compilation due to -Werror.

Possible fixes:
 1. deactivate the warning
2. add `__attribute__ ((nonstring))` to the three occurrences where this
is a problem.

Chosen solution: 1.
Since the attribute is not available on older clang versions, it would
trigger another warning there.

(included a small formatting change in the commit - pls ignore)

My motivation is to allow -mcpu=lime1 build with LLVM 20.

Note that extended-const is a part of lime1.
https://github.com/WebAssembly/tool-conventions/blob/main/Lime.md#lime1

cf. https://github.com/WebAssembly/wasi-sdk/pull/527

This change makes it so that the timestamps of installed header files
are only updated when the contents of the header files change. This
change is beneficial for incremental builds of projects depending on
wasi-libc. For example, the Swift project builds wasi-libc as part of
its build process, and this change reduces the number of build products
in the Swift build that are unnecessarily rebuilt.

This is a 2nd attempt of
https://github.com/WebAssembly/wasi-libc/pull/530 but much simpler
thanks to @abrown 's build work

https://github.com/WebAssembly/wasi-libc/pull/563#discussion_r1940291020

This PR updates ADAPTER_URL to reference the latest release version
(versions 29.0.1)

This PR updates WASM_TOOLS_URL to reference the latest released version
(version 1.224.0)

This PR updates WASMTIME_URL to reference the latest release version
(29.0.1)

This PR updates LIBRT_URL to reference the latest release version
(wasi-sdk-25).

This commit adds a browser test harness to run the tests in the browser.

## Motivation

We are heavily using wasi-libc on browsers but we don't have any test
for the case in wasi-libc. In theory, there should be no behavioral
difference between wasmtime and browsers (as long as WASI
implementations are correct) but browsers have their own limitations in
their Wasm engines. For example, memory.atomic.wait32 is not permitted
on the main thread.

We are working on adding some mitigation for such browser-specific
issues (https://github.com/WebAssembly/wasi-libc/pull/562) and this test
harness will help to validate the fix.

Rather than using Macos-latest this makes the ci used a pinned version
of Macos (set to the latest version available). This is so if you happen
to have an issue with the MacOS build you know exactly which version of
MacOS it is running, rather than having to dig into the ci, or Github
runner documentation.

Rather than using Windows-latest this PR updates the ci to use a pinned
version of Windows (set to the latest version available). This is so if
you happen to have an issue with the Windows build you know exactly
which version of Windows it is running, rather than having to dig into
the ci, or Github runner documentation.

https://github.com/llvm/llvm-project/commit/33ad474c45e6d7a0de7bc75e15e27cf6cb9ff895

This change enables the pthread-related tests in the libc-test suite.
The tests are enabled only for the `wasm32-wasip1-threads` target, which
is the only target that supports threads at the moment.

The following pthread tests are still disabled:
- pthread_cancel-points.c
- pthread_cancel.c
- pthread_robust.c

This is a preparative change for
https://github.com/swiftwasm/swift/issues/5598

Follow-up to 2b853ff0

The wrong shebang caused `&>` to be interpreted as a background job

GitHub Actions is migrating ubuntu-latest to Ubuntu 24.04, which no
longer provides the libtinfo5 package. Since older LLVM versions depend
on libtinfo5 and do not offer pre-built binaries for recent Ubuntu
versions, we explicitly specify ubuntu-22.04 in the workflow to maintain
compatibility.

https://github.com/llvm/llvm-project/commit/c3536b263f253a69fb336fb0617ee33a01a5c5dd

This change represents a rather large re-design in how `wasi-libc`
builds and runs its tests. Initially, #346 retrieved the `libc-test`
repository and built a subset of those tests to give us some amount of
test coverage. Later, because there was no way to add custom C tests,
#522 added a `smoke` directory which allowed this. But (a) each of these
test suites was built and run separately and (b) it was unclear how to
add more tests flexibly--some tests should only run on `*p2` targets or
`*-threads` targets, e.g.

This change reworks all of this so that all tests are built the same
way, in the same place. For downloaded tests like those from
`libc-test`, I chose to add "stub tests" that `#include` the original
version. This not only keeps all enabled tests in one place, it also
allows us to add "directives," C comments that the `Makefile` uses to
filter out tests for certain targets or add special compile, link or run
flags. These rudimentary scripts, along with other Bash logic I moved
out of the Makefile now live in the `scripts` directory.

Finally, all of this is explained more clearly in an updated
`README.md`. The hope with documenting this a bit better is that it
would be easier for drive-by contributors to be able to either dump in
new C tests for regressions they may find or enable more libc-tests. As
of my current count, we only enable 40/75 of libc-test's functional
tests, 0/228 math tests, 0/69 regression tests, and 0/79 API tests.
Though many of these may not apply to WASI programs, it would be nice to
explore how many more of these tests can be enabled to increase
wasi-libc's test coverage. This change should explain how to do that
and, with directives, make it possible to condition how the tests
compile and run.

This change is not meant to change any functionality, only move some
Bash-specific logic out of the `Makefile` into its own script:
`install-include-headers.sh`. This reduces the perceived complexity of
the `Makefile` but the complexity is still there, tucked away in this
script. This script also has the advantage that it can be run separately
if needed.

This commit comes after a few different attempts at building up
`Makefile` lists of headers to copy over along with the various
locations they must be copied from. It is certainly possible to do this,
but due to how we need to remove some headers from the list, it ends up
being easier to just `cp` and then `rm`, which this script retains.

This change is an effort to move all dependencies of the phony
`bindings` target into one place, for clarity.

This adds a `.stamp` file to the process of copying over the `crt` files
to the sysroot directory to avoid re-copying on each `make` invocation.

This adds a `.stamp` file to the build directory that tracks when all
headers have been copied over to the sysroot `include` directory.
Previously, the `include_dirs` target was phony, which `make` will run
every time, regardless of whether it is needed. This was part of the
problem with all of libc being rebuilt at each `make` invocation.

Previously, these were rebuilt each time as a part of the phone
`dummy_libs` target. This change only rebuilds them if they don't exist.