Files
LEDMatrix/test/test_web_error_detail.py
T
44f59ede07 fix(web): say what actually went wrong instead of "unknown" (#448)
* fix(web): say what actually went wrong instead of "unknown"

Every failing endpoint returned "An error occurred; see logs for
details" and nothing else. That is survivable until the logs are the
thing you cannot reach: a device whose SD card was failing answered the
restart action, /system/status and /logs with that same sentence -- the
log viewer included, because journalctl could not be executed -- while
the exception underneath said

    [Errno 5] Input/output error: 'systemctl'

which names the fault outright. The only endpoint that helped was
/health, and only because it happens to pass a subprocess's stderr
through. Diagnosis came down to guessing which endpoint leaked something.

Add describe_exception(), returning "TypeName: message" on one line, and
populate the `details` field that the response schema has always had and
nothing ever filled. The type alone carries information -- a bare
PermissionError says more than any generic sentence.

Exception text is not automatically safe to echo: a requests error
quotes the URL it failed on, and plugins that authenticate by query
string put their key there. Credential values are redacted while the
parameter name is kept, since knowing which credential was involved is
part of the diagnosis. Length is capped and newlines collapsed so a
parser's context cannot flood a JSON field.

Nine handlers in api_v3 bound the exception and never used it, so the
promised log entry was never written either -- "see logs for details"
was false, not merely unhelpful. Those now log with a traceback and
carry the detail. The other 60 already logged and are unchanged; they
can adopt the helper as they are touched.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Udr6MfaFLUPhX5Fgo67Jf5

* fix(web): redact auth headers and URL userinfo, and cover every handler

Three review findings.

The sanitizer missed two credential shapes that requests puts in its
exception text verbatim: `Authorization: Bearer <token>` and
`https://user:password@host`. Both would have gone straight into a
response. The auth-scheme name and the username are kept -- they say
which credential and whose without being the secret.

The AST test only asked whether *something* had been logged, so a
`logger.info("failed")` satisfied it while discarding the exception just
as completely. It now requires an error-level record carrying exc_info
and `describe_exception()` called on the handler's own bound exception.

Enforcing that revealed the first cut had scoped itself wrongly. I had
converted the nine handlers that logged nothing and left the sixty that
logged, reasoning their detail was at least in the journal. But
/system/status is one of the sixty, and on the failing device it told me
nothing -- the journal was exactly what could not be read. Splitting
them left most of the diagnostic surface unhelpful for the case this
change exists for, so all sixty-nine now carry the detail.

Two handlers had no bound exception name, and three passed the message
through a variable rather than a literal; both shapes needed doing by
hand. Full suite: 2383 passed, one pre-existing unrelated failure.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Udr6MfaFLUPhX5Fgo67Jf5

* fix(web): stop reporting client errors as server faults

Werkzeug's HTTPExceptions subclass Exception, so the catch-all handler
saw them too and turned every 405, 400, 413 and 415 into a 500
UNKNOWN_ERROR. A GET on a POST-only route answered "an error occurred;
see logs for details", which tells the caller nothing and blames the
wrong side -- found while probing a device whose POST-only config
endpoints did exactly that.

Hand HTTPExceptions back as themselves, with their own status and
description. A genuine server fault still reports as one, with the
detail this branch adds.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Udr6MfaFLUPhX5Fgo67Jf5

* fix(web): redact any auth scheme, and require the detail in the response

Two review findings.

The auth-header pattern listed Bearer, Basic, Digest and Token, so
`Authorization: ApiKey SECRET` or `Negotiate SECRET` went to the client
intact. A fixed list silently leaks whatever it does not name, and
plugin APIs invent their own schemes, so match any scheme name and keep
it while redacting the credential.

The AST test accepted a describe_exception(e) call anywhere in the
handler, which a handler could satisfy by computing the detail and
dropping it before returning the generic message. It now requires the
call inside every return expression, which is where it has to be to
reach the caller.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Udr6MfaFLUPhX5Fgo67Jf5

---------

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-11 13:56:16 -04:00

249 lines
11 KiB
Python

"""Tests for surfacing the underlying error in web responses.
Regression under test: every failing endpoint returned "An error occurred; see
logs for details" and nothing else. On a device whose storage was failing that
sentence came back from the restart action, from /system/status, and from
/logs -- the log viewer itself -- because journalctl could not be executed. The
exception underneath said `[Errno 5] Input/output error: 'systemctl'`, which
names the fault outright, and nine handlers were discarding it entirely rather
than even logging it.
"""
import pytest
from src.web_interface.error_handler import describe_exception
class TestDescribeException:
def test_names_the_type_and_message(self):
detail = describe_exception(OSError(5, "Input/output error", "systemctl"))
assert detail == "OSError: [Errno 5] Input/output error: 'systemctl'"
def test_the_reported_failure_is_legible(self):
# The whole point: this string is the diagnosis.
assert "Input/output error" in describe_exception(
OSError(5, "Input/output error", "systemctl"))
def test_a_bare_exception_still_names_its_type(self):
# A PermissionError with no message still says more than "unknown".
assert describe_exception(PermissionError()) == "PermissionError"
assert describe_exception(Exception()) == "Exception"
def test_message_is_kept_when_present(self):
assert describe_exception(ValueError("bad port")) == "ValueError: bad port"
class TestCredentialRedaction:
"""Exception text quotes URLs, and plugins authenticate by query string."""
@pytest.mark.parametrize("secret_text,leaked", [
("failed: https://api.x.com/v1?api_key=SEC123&city=Tampa", "SEC123"),
("token=abcdef123456 was rejected", "abcdef123456"),
("connect failed password=hunter2", "hunter2"),
("GET /?access_token=zzz999", "zzz999"),
('{"secret": "topsecret"}', "topsecret"),
# requests quotes the URL it failed on, and both of these forms turn
# up in real client exceptions.
("401 for https://user:hunter2@example.com/api", "hunter2"),
("headers: {'Authorization': 'Bearer eyJ.SECRET.sig'}", "eyJ.SECRET.sig"),
("Authorization: Basic dXNlcjpwYXNzd29yZA==", "dXNlcjpwYXNzd29yZA=="),
("Proxy-Authorization: Bearer ptok999", "ptok999"),
# Any scheme, not a fixed list -- a list silently leaks whatever it
# does not name, and plugin APIs invent their own.
("Authorization: ApiKey SECRET123", "SECRET123"),
("Authorization: Negotiate YIIZnegotiateblob", "YIIZnegotiateblob"),
("Authorization: NTLM TlRMTVNTUAAB", "TlRMTVNTUAAB"),
("authorization: barecredential", "barecredential"),
])
def test_credentials_never_reach_the_response(self, secret_text, leaked):
detail = describe_exception(RuntimeError(secret_text))
assert leaked not in detail
assert "<redacted>" in detail
def test_the_parameter_name_survives_redaction(self):
# Knowing *which* credential was involved is part of the diagnosis.
detail = describe_exception(RuntimeError("https://x/y?api_key=SEC123"))
assert "api_key" in detail
def test_unknown_schemes_keep_their_name(self):
for scheme in ("ApiKey", "Negotiate", "NTLM", "AWS4-HMAC-SHA256"):
detail = describe_exception(
RuntimeError("Authorization: %s SECRETVALUE" % scheme))
assert scheme in detail, detail
assert "SECRETVALUE" not in detail, detail
def test_auth_scheme_and_username_survive(self):
# Which kind of credential, and whose, without the credential itself.
assert "Bearer" in describe_exception(
RuntimeError("Authorization: Bearer eyJ.SECRET.sig"))
assert "user" in describe_exception(
RuntimeError("https://user:hunter2@example.com"))
def test_non_secret_context_is_preserved(self):
detail = describe_exception(RuntimeError("https://api.x.com/v1?city=Tampa"))
assert "city=Tampa" in detail
assert "<redacted>" not in detail
class TestBounds:
def test_long_messages_are_truncated(self):
detail = describe_exception(ValueError("x" * 5000))
assert len(detail) <= 400
def test_newlines_are_collapsed_to_one_line(self):
detail = describe_exception(ValueError("line one\nline two\tthree"))
assert "\n" not in detail and "\t" not in detail
assert detail == "ValueError: line one line two three"
def test_custom_length_is_honoured(self):
assert len(describe_exception(ValueError("y" * 500), max_length=50)) <= 50
class TestHandlersCarryDetail:
"""The response shape callers actually see."""
def test_no_api_v3_handler_discards_its_exception(self):
"""Every generic-message handler must log a traceback and return detail.
Nine of them bound `e` and never used it, so the promised log entry was
never written either. Checking merely that *something* was logged is
too weak -- a `logger.info("failed")` would satisfy it while throwing
the exception away just as completely, so this asserts the two things
that actually make the failure diagnosable: an error-level record with
the traceback, and the sanitized detail in the response.
"""
import ast
src = open("web_interface/blueprints/api_v3.py").read()
tree = ast.parse(src)
generic = "An error occurred; see logs for details"
def logs_a_traceback(handler):
"""An error/exception-level log call carrying exc_info."""
for call in [n for n in ast.walk(handler) if isinstance(n, ast.Call)]:
func = call.func
if not isinstance(func, ast.Attribute):
continue
if func.attr == "exception": # implies exc_info
return True
if func.attr not in ("error", "critical"):
continue
if any(kw.arg == "exc_info" and getattr(kw.value, "value", False) is True
for kw in call.keywords):
return True
return False
def describes_this_exception(node, bound):
"""A describe_exception(<bound>) call anywhere under `node`."""
for call in [n for n in ast.walk(node) if isinstance(n, ast.Call)]:
if not (isinstance(call.func, ast.Name)
and call.func.id == "describe_exception"):
continue
if bound is None:
return True # bare `except:` cannot name it; accept
if any(isinstance(a, ast.Name) and a.id == bound
for a in call.args):
return True
return False
def returns_the_detail(handler):
"""The detail must be inside what the handler actually returns.
Looking anywhere in the handler is too weak: a handler could
compute describe_exception(e), drop it on the floor, and return the
generic message with no details field, while still passing. So the
call has to appear within a `return` expression.
"""
returns = [n for n in ast.walk(handler) if isinstance(n, ast.Return)]
if not returns:
return False
return all(describes_this_exception(r, handler.name) for r in returns)
offenders = []
for h in [n for n in ast.walk(tree) if isinstance(n, ast.ExceptHandler)]:
seg = ast.get_source_segment(src, h) or ""
if generic not in seg:
continue
missing = []
if not logs_a_traceback(h):
missing.append("error-level log with exc_info")
if not returns_the_detail(h):
missing.append("describe_exception(e) in the response")
if missing:
offenders.append((h.lineno, missing))
assert not offenders, (
"handlers returning the generic message without %s: %r"
% ("both a traceback log and the detail", offenders))
def test_client_errors_keep_their_own_status(self):
"""A 405 must not be reported as a server-side UNKNOWN_ERROR.
Werkzeug's HTTPExceptions subclass Exception, so the catch-all saw them
too: a GET on a POST-only route came back 500 "an error occurred",
which tells the caller nothing and blames the wrong side. Found while
probing a device whose POST-only config endpoints answered every GET
with UNKNOWN_ERROR.
"""
from flask import Flask, jsonify
from werkzeug.exceptions import HTTPException
app = Flask(__name__)
@app.errorhandler(Exception)
def handle(error):
if isinstance(error, HTTPException):
return jsonify({
"status": "error",
"error_code": (error.name or "HTTP_ERROR").upper().replace(" ", "_"),
"message": error.description,
}), error.code or 500
return jsonify({
"status": "error",
"error_code": "UNKNOWN_ERROR",
"message": "An error occurred; see logs for details",
"details": describe_exception(error),
}), 500
@app.route("/only-post", methods=["POST"])
def only_post():
return jsonify({"ok": True})
@app.route("/boom")
def boom():
raise OSError(5, "Input/output error", "systemctl")
client = app.test_client()
resp = client.get("/only-post")
assert resp.status_code == 405, "a wrong method must stay a 405"
assert resp.get_json()["error_code"] == "METHOD_NOT_ALLOWED"
# A genuine server fault still reports as one, with its detail.
resp = client.get("/boom")
assert resp.status_code == 500
assert "Input/output error" in resp.get_json()["details"]
def test_global_handler_reports_the_underlying_error(self):
from flask import Flask, jsonify
app = Flask(__name__)
@app.errorhandler(Exception)
def handle(error):
return jsonify({
"status": "error",
"error_code": "UNKNOWN_ERROR",
"message": "An error occurred; see logs for details",
"details": describe_exception(error),
}), 500
@app.route("/boom")
def boom():
raise OSError(5, "Input/output error", "systemctl")
client = app.test_client()
body = client.get("/boom").get_json()
assert body["error_code"] == "UNKNOWN_ERROR"
assert "Input/output error" in body["details"]