[AryanRG]

Multipart uploads with an explicit boundary such as test-boundary currently fail match_files because the matcher only extracts lowercase hexadecimal boundaries. This also makes get_requests(match_files=...) return no matching requests.

The matcher now reuses a nonempty ASCII boundary from the first body line and retains the full-body equality check. Non-ASCII prefixes remain nonmatches, so arbitrary binary uploads can reach later response matchers without an ASCII-decoding error. Includes sync/async selection and retrieval regressions and an Unreleased changelog entry.

Reproduction (HTTPX 0.28.1):

import httpx

def test_custom_boundary(httpx_mock):
    files = {"name": ("file_name", b"File content")}
    httpx_mock.add_response(match_files=files, status_code=201)
    with httpx.Client() as client:
        response = client.put(
            "https://test_url",
            headers={"Content-Type": "multipart/form-data; boundary=test-boundary"},
            files=files,
        )
    assert response.status_code == 201

Before: httpx.TimeoutException and an unused response at teardown. After: HTTP 201. For retrieval, register an unrestricted response, send the same request, then compare httpx_mock.get_requests(match_files=files) with [response.request]: before [], after the captured request.

Validation on macOS arm64: the full suite passed on Python 3.11.16 and 3.13.12 (390 tests each; 100% statement/branch coverage), 170 additional probes passed, and all eight pre-commit hooks passed. Python 3.12/3.14/3.15, Linux/Windows, and package builds were not run locally; upstream CI is separate. This retains the existing body-based comparison and does not add strict multipart syntax or header/body consistency validation.

Overlap: #239 is currently open and does not include this fix. If that migration lands first, these two new test modules need adaptation to its backend fixture and fresh HTTPX-only, HTTPX2-only, and combined validation; their direct import httpx currently fails collection with only HTTPX2 installed.

AI assistance was used to prepare implementation and regression tests.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud