Fix HTTP connect_timeout enforcement

[Coldwings]

Description

Enforce the shared HTTP client connect_timeout during TCP connection setup and TLS handshake. This fixes clients that could hang in connection setup even though base_client_config::connect_timeout advertised a timeout.

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Performance improvement (optimization that improves speed/memory usage)
• Documentation (changes to documentation, comments, or examples)
• Refactoring (code changes that neither fix bugs nor add features)
• Tests (adding or modifying tests)
• Build/CI (changes to build system, CI configuration, or dependencies)

Related Issues

Closes #412

Changes Made

Core Changes

• Add cancellable net::tcp_connect overloads backed by cancel_token.
• Add a cancellable TLS handshake overload that uses cancellable TCP readiness polling.
• Enforce base_client_config::connect_timeout in the shared client_connect path used by HTTP/1, WebSocket, and SSE clients.
• Preserve specific connection setup errors such as ETIMEDOUT instead of always mapping acquire failure to ECONNREFUSED.
• Update API and networking wiki docs for the enforced timeout semantics.

API Changes (if applicable)

Before:

// connect_timeout existed in base_client_config, but HTTP/1 connection setup
// did not pass a cancellation path into TCP connect or TLS handshake.

After:

// connect_timeout now bounds TCP connect plus TLS handshake for the shared
// HTTP/1, WebSocket, and SSE connection path. Values <= 0 disable it.

Migration Guide (if breaking change)

N/A. This enforces existing configuration rather than changing the public configuration surface.

Testing

Unit Tests

• Added new tests for the changes
• Updated existing tests if needed
• All tests pass locally

Integration Tests

• Tested with existing examples
• Tested in real-world scenarios (if applicable)

Sanitizer Testing

• Tested with ASAN (AddressSanitizer)
• Tested with TSAN (ThreadSanitizer)
• No new warnings or errors in the local debug build

Test Results

cmake --build /tmp/elio-fix-412-build --target elio_tests --parallel 2
/tmp/elio-fix-412-build/tests/elio_tests "[http][client]"
timeout 120s /tmp/elio-fix-412-build/tests/elio_tests "[http2][timeout]"
timeout 120s /tmp/elio-fix-412-build/tests/elio_tests "[io][cancel]"
timeout 120s /tmp/elio-fix-412-build/tests/elio_tests "[websocket]"
timeout 120s /tmp/elio-fix-412-build/tests/elio_tests "[sse]"
git diff --check

Checklist

Code Quality

• My code follows the project's coding standards
• I have added/updated comments for complex logic
• I have removed any debug code, TODOs, or commented-out code
• My changes generate no new warnings

Documentation

• I have updated documentation (wiki, README, code comments)
• I have added examples for new features (if applicable)
• I have updated API documentation (if applicable)

Testing

• I have added tests that prove my fix is effective or my feature works
• New and existing unit tests pass locally with my changes
• I have tested with ASAN and TSAN

Compatibility

• My changes are backward compatible (or I've documented breaking changes)
• I have considered the impact on existing users
• I have updated CHANGELOG.md (if applicable)

Performance (if applicable)

• I have considered the performance impact
• I have added benchmarks for performance-critical changes

Screenshots / Diagrams

N/A.

Additional Notes

Full ctest was not run because this configuration includes NOT_BUILT sanitizer/RDMA placeholder tests; the targeted executable-level tests above were run instead.

Reviewer Guidance

Areas requiring special attention:

• Cancellation and fd ownership in the new cancellable tcp_connect path.
• Timeout error preservation from client_connect through http::client::send_request.
• Documentation distinction between HTTP/1/WebSocket/SSE timeout behavior and the existing HTTP/2 reserved connect_timeout note.

Questions for reviewers:

• None.

[Copilot]

Pull request overview

Enforces HTTP connection timeouts across TCP setup and TLS handshakes for HTTP, WebSocket, and SSE clients.

Changes:

• Adds cancellable TCP connect and TLS handshake APIs.
• Applies shared connection deadlines while preserving setup errors.
• Adds TLS timeout coverage and updates documentation.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
wiki/Networking.md	Documents client timeout configuration.
wiki/API-Reference.md	Clarifies timeout semantics.
tests/unit/test_http_client.cpp	Tests stalled TLS handshake timeout.
include/elio/tls/tls_stream.hpp	Adds cancellable TLS handshaking.
include/elio/net/tcp.hpp	Adds cancellable TCP connection overloads.
include/elio/http/websocket_client.hpp	Passes WebSocket connection timeout.
include/elio/http/sse_client.hpp	Passes SSE connection timeout.
include/elio/http/http_client.hpp	Enforces timeout through connection pooling.
include/elio/http/client_base.hpp	Implements shared connection watchdog.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Coldwings]

@copilot Please re-review the latest changes only. Do not push commits or make direct modifications.

[Copilot]

@copilot Please re-review the latest changes only. Do not push commits or make direct modifications.

Reviewed commit 97947cf. All four concerns from the prior review are addressed:

errno preservation (client_base.hpp):
stop_watchdog_preserving_errno saves errno into saved_errno (a coroutine-frame local) before awaiting, then restores it after — correct across scheduler-worker migrations. The all-addresses-failed paths use the same pattern (connect_errno saved before co_await stop_watchdog(), restored after). One minor observation: errno ? errno : ECONNREFUSED is defensive but reasonable since errno should be set by the last failed tcp_connect attempt.

API ambiguity (tcp.hpp):
The constrained template requires std::is_same_v<std::remove_cvref_t<Token>, coro::cancel_token> is the right fix. Braced-init-list {} cannot be used for template argument deduction, so it never selects the cancellable overload; the non-template tcp_options overload wins unambiguously. The added compilation test confirms this.

Constructor reorder (tcp_connect_awaitable):
The (addr, opts, token) order matches the new factory-function signature and removes the previous mismatch.

Cancellation tests:
Both tests are sound. The already-cancelled-token test cleanly exercises the await_suspend early-return path. The pending-connect test uses 192.0.2.1 (RFC 5737 TEST-NET) sensibly, handles the "platform refused before cancellation" race with a graceful early SUCCEED, and the fd check (after_fds < pending_fds) is correct given both counts use the same helper (the DIR fd is consistently included in both and closed before returning).