mirror of https://github.com/open-goal/jak-project synced 2026-09-11 20:41:41 -04:00

Files

T

Alexander J. Semenuk 5d5e35fb9b fix: Windows toolchain compatibility (curl 8.21 re-vendor, endless reconfigure loop) (#4355 )

## 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)

2026-07-27 19:19:18 -04:00

4.9 KiB

Vendored

Generated

Raw Permalink Blame History

WebSocket in curl

URL

WebSocket communication with libcurl is done by setting up a transfer to a URL using the ws:// or wss:// URL schemes. The latter one being the secure version done over HTTPS.

When using wss:// to do WebSocket over HTTPS, the standard TLS and HTTPS options are acknowledged for the CA, verification of server certificate etc.

WebSocket communication is done by upgrading a connection from either HTTP or HTTPS. When given a WebSocket URL to work with, libcurl considers it a transfer failure if the upgrade procedure fails. This means that a plain HTTP 200 response code is considered an error for this work.

API

The WebSocket API is described in the individual man pages for the new API.

WebSocket with libcurl can be done two ways.

Get the WebSocket frames from the server sent to the write callback. You can then respond with curl_ws_send() from within the callback (or outside of it).
Set CURLOPT_CONNECT_ONLY to 2L (new for WebSocket), which makes libcurl do an HTTP GET + Upgrade: request plus response in the curl_easy_perform() call before it returns and then you can use curl_ws_recv() and curl_ws_send() to receive and send WebSocket frames from and to the server.

The new options to curl_easy_setopt():

CURLOPT_WS_OPTIONS - to control specific behavior. CURLWS_RAW_MODE makes libcurl provide all WebSocket traffic raw in the callback. CURLWS_NOAUTOPONG disables automatic PONG replies.

The new function calls:

curl_ws_recv() - receive a WebSocket frame

curl_ws_send() - send a WebSocket frame

curl_ws_meta() - return WebSocket metadata within a write callback

Max frame size

The current implementation only supports frame sizes up to a max (64K right now). This is because the API delivers full frames and it then cannot manage the full 2^63 bytes size.

If we decide we need to support (much) larger frames than 64K, we need to adjust the API accordingly to be able to deliver partial frames in both directions.

Errors

If the given WebSocket URL (using ws:// or wss://) fails to get upgraded via a 101 response code and instead gets another response code back from the HTTP server - the transfer returns CURLE_HTTP_RETURNED_ERROR for that transfer. Note then that even 2xx response codes are then considered error since it failed to provide a WebSocket transfer.

Test suite

I looked for an existing small WebSocket server implementation with maximum flexibility to dissect and cram into the test suite but I ended up deciding that extending the existing test suite server sws to deal with WebSocket might be the better way.

This server is already integrated and working in the test suite
We want maximum control and ability to generate broken protocol and negative tests as well. A dumber and simpler TCP server could then be easier to massage into this than a "proper" WebSocket server.

Command line tool WebSocket

The plan is to make curl do WebSocket similar to telnet/nc. That part of the work has not been started.

Ideas:

Read stdin and send off as messages. Consider newline as end of fragment. (default to text? offer option to set binary)
Respond to PINGs automatically
Issue PINGs at some default interval (option to switch off/change interval?)
Allow -d to specify (initial) data to send (should the format allow for multiple separate frames?)
Exit after N messages received, where N can be zero.

Future work

Verify the Sec-WebSocket-Accept response. It requires a sha-1 function.
Verify Sec-WebSocket-Extensions and Sec-WebSocket-Protocol in the response
Consider a curl_ws_poll()
Make sure WebSocket code paths are fuzzed
Add client-side PING interval
Provide option to disable PING-PONG automation
Support compression (CURLWS_COMPRESS)

Why not libWebSocket

libWebSocket is said to be a solid, fast and efficient WebSocket library with a vast amount of users. My plan was originally to build upon it to skip having to implement the low level parts of WebSocket myself.

Here are the reasons why I have decided to move forward with WebSocket in curl without using libWebSocket:

doxygen generated docs only makes them hard to navigate. No tutorial, no clearly written explanatory pages for specific functions.
seems (too) tightly integrated with a specific TLS library, while we want to support WebSocket with whatever TLS library libcurl was already made to work with.
seems (too) tightly integrated with event libraries
the references to threads and thread-pools in code and APIs indicate too much logic for our purposes
"bloated" - it is a huge library that is actually more lines of code than libcurl itself
WebSocket is a fairly simple protocol on the network/framing layer so making a homegrown handling of it should be fine

4.9 KiB Vendored Generated Raw Permalink Blame History