Alexander J. Semenuk
5d5e35fb9b
## Problem Windows builds break with a current local toolchain (Scoop LLVM 22.1.8, CMake 4.4.0, VS 2026), in two independent ways: 1. The build stops at curl's deliberate guard: `#error "no non-blocking method was found/used/set"` in `third-party/curl/lib/nonblock.c`. 2. From the second configure onward, `cmake --build` re-runs CMake in an endless loop (observed 42 consecutive reconfigure cycles in a single build). Likely the same mechanism behind the "endlessly building" VS 2026 note in `docs/setup/dev/vs.md`. ## Root cause 1. `third-party/curl/CMake/CurlTests.c` passes `int *` to `ioctlsocket()`, whose third parameter is `u_long *`. Clang 22 promotes `-Wincompatible-pointer-types` to a hard error in C, so the `HAVE_IOCTLSOCKET_FIONBIO` try_compile silently fails and `curl_config.h` never defines it. Upstream CI does not see this because the windows-2022 runner image ships an older LLVM. GCC 14 promotes the same warning to a hard error, which is very likely the `CurlTests.c.obj` failure reported from MSYS2 in open-goal/jak-project#3551. Upstream curl hit the identical problem with GCC 14 and fixed the probe in curl 8.8.0 (curl/curl#13578). 2. The root CMakeLists copies the build tree's `compile_commands.json` into `<src>/build/` for clangd using `configure_file()`, which registers its input as a configure dependency. CMake rewrites `compile_commands.json` late in every generation, after `CTestTestfile.cmake` and `cmake_install.cmake` (outputs of the same Ninja regen rule), so once the dependency is registered the rule is deterministically dirty and every `ninja` invocation re-runs CMake. A pristine first configure is safe (the file does not exist yet, so the `if(EXISTS ...)` guard skips the copy), which is why the loop looks machine- or IDE-specific. ## Fix 1. Per review, re-vendor `third-party/curl` at the `curl-8_21_0` tag (previously `curl-8_3_0`), which carries the upstream probe fix plus two years of upstream development; `vendor.yaml` updated to match. Adjustments the version jump forced: - curl 8.15 removed the native macOS Secure Transport backend (`CURL_USE_SECTRANSP`), so macOS now builds curl against OpenSSL like Linux. The two macOS workflows install Homebrew `openssl@3` and export `OPENSSL_ROOT_DIR` (keg-only), and the macOS setup docs gained the same two lines. - `CURL_BROTLI` / `CURL_ZSTD` switched to AUTO-detection in curl 8.10; pinned OFF to keep the previous no-compression behavior and avoid silently linking whatever the CI images happen to have. - curl's new top-level `BUILD_EXAMPLES` cache option (default ON) leaked into discord-rpc's identically named option and broke configure at a nonexistent `examples/send-presence` directory; pinned OFF ahead of the third-party subdirectories. The diff is dominated by the mechanical tag-tree swap under `third-party/curl` (linguist-vendored, collapsed in review). The hand-written changes are `CMakeLists.txt`, the two macOS workflows, `docs/setup/system/macos.md`, and `vendor.yaml`. 2. Swap `configure_file()` for `file(COPY ...)`: the same clangd copy with no configure dependency registered. (`file(COPY_FILE ... ONLY_IF_DIFFERENT)` would be cleaner still but requires CMake 3.21, above the declared `cmake_minimum_required(VERSION 3.10)`.) ## Test plan - [x] Fresh `cmake --preset Release-windows-clang` (LLVM 22, no cache seeding) completes and logs `Enabled SSL backends: Schannel`; the FIONBIO probe passes without the previous `#error` - [x] Full Windows Release build from scratch in the branch worktree (all 1422 targets) - [x] goalc-test suite: 1509 passed, 0 failed - [x] Second consecutive configure with `compile_commands.json` present: the regen rule in `build.ninja` has no `compile_commands.json` input; `<src>/build/compile_commands.json` is still refreshed for clangd - [x] Repeated `ninja` invocations after a full build no longer re-run CMake - [x] macOS Intel and ARM CI green (first exercise of the OpenSSL backend switch) --- I work off a self-hosted forge, so this GitHub account is quiet; the configure logs and ninja dirty-node traces from the investigation are available if anyone wants the raw data. (AI-assisted)
4.1 KiB
Vendored
Generated
Thread Pool and Queue
The thread pool and queue manage asynchronous processing of "work items"
to a user. The "work items" are opaque to the pool and queue, represented
by a void *, handled via callback functions.
Thread pool and queue are available with pthreads or native Win32
builds.
Curl_thrdpool
This data structure manages a pool of threads for asynchronous operations.
Properties
A pool's properties are:
- minimum number of threads running, default 0
- maximum number of threads running
- timeout for idle threads before they shut down
The minimum number of threads is started at creation of the pool and kept always running. On demand, when more work is available but all existing threads are busy, it starts new threads, up to maximum.
When work ceases, the threads "above" the minimum number exit again after the given idle time.
Operation
The pool is created with providing three callback functions:
take: the pool calls this to take a new "work item" for processing. From the pool's point of view, a work item is avoid *. "take" is called from the pool's threads. When getting anything besidesNULL, the thread is "busy". On gettingNULL, the thread becomes idle.process: called by a pool thread to process a work item. This can not return any error. Any error handling must be done via properties in the work item itself, opaque to the pool.return: after processing, the work item is returned and the pool has no longer have any memory of it.
The pool only tries to "take" new work items when told to. Calling
Curl_thrdpool_signal(pool, n) awakens up to nthreads which then
take new work items. This may cause new threads being started. The other
time a pool thread "take"s work it when it has finished
processing and returned another item.
A thread pool can be destroyed via Curl_thrdpool_destroy(pool, join) where
join determines if active threads shall be joined or detached.
Safety
The thread pool operates use a mutex and condition variables to manage concurrency. All interactions and callback invocation are done under the pool's mutex lock, except the "process" callback which is invoked unlocked.
To avoid deadlocks, no callback must invoked other pool functions. Also, any call of pool functions may result in callback invocations.
The "work items", once "taken" by the pool, should not be referenced from any other place. Thread pools always invoke the "return" callback on a work item, even after the pool has been destroyed by detaching the threads.
There is a user_data in the pool's creation that is passed to "take"
and "return" callbacks. Once a pool is destroyed, this user_data is
cleared and "return" callbacks always see a NULL. This way,
the "return" callback may act on that fact.
Curl_thrdq
A thrdq is a two-way queue with a thread pool. Users of a thread queue may
"send" work items into the queue and "receive" processed items back.
Properties
A queue's properties are:
- The properties of the thread pool to create
- the maximum length of the "send" queue, 0 for unlimited
Operation
The queue is created with providing three callback functions:
free: called to free a work item that is in the queue but is no longer returned (or processed). This happens when the queue is destroyed or when work items are removed for other reasons.process: process the item. Can not fail.event: called when work items have been added to the "receive" list.
Users of a thread queue call Curl_thrdq_send() to add a work item to
the queue. Calling Curl_thrdq_recv() delivers processed items back.
Safety
The thread queue operates use a mutex and condition variables to manage concurrency. All interactions and callback invocation are done under the queue's mutex lock, except the "process" callback which is invoked unlocked.
Users of a thread queue should not hold any reference to work items sent into the queue. The provided "free" callback has to take care of any resources allocated by work items.