client transport : ignore http status for gRPC mode #8548

mbissa · 2025-09-01T06:27:16Z

When a gRPC response is received with content type application/grpc, we then do not expect any information in the http status and the status information needs to be conveyed by gRPC status only.

In case of missing gRPC status, we will throw an Internal error instead of Unknown in accordance with https://grpc.io/docs/guides/status-codes/

Changes :
- Ignore http status in case of content type application/grpc
- Change the default rawStatusCode to return Internal for missing grpc status

RELEASE NOTES:

client : Ignore http status for gRPC mode and return Internal error for missing grpc status

Fixes grpc#8486 When a gRPC response is received with content type application/grpc, we then do not expect any information in the http status and the status information needs to be conveyed by gRPC status only. In case of missing gRPC status, we will throw an Internal error instead of Unknown in accordance with https://grpc.io/docs/guides/status-codes/ Changes : - Ignore http status in case of content type application/grpc - Change the default rawStatusCode to return Internal for missing grpc status RELEASE NOTES: * client : Ignore http status for gRPC mode and return Internal error for missing grpc status

codecov · 2025-09-01T06:30:45Z

Codecov Report

❌ Patch coverage is 43.90244% with 23 lines in your changes missing coverage. Please review.
✅ Project coverage is 80.77%. Comparing base (e60a04b) to head (7087664).
⚠️ Report is 3 commits behind head on master.

Files with missing lines	Patch %	Lines
internal/transport/http2_client.go	43.90%	9 Missing and 14 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #8548      +/-   ##
==========================================
- Coverage   81.64%   80.77%   -0.88%     
==========================================
  Files         413      413              
  Lines       40621    40754     +133     
==========================================
- Hits        33167    32920     -247     
- Misses       5991     6202     +211     
- Partials     1463     1632     +169

Files with missing lines	Coverage Δ
internal/transport/http2_client.go	`63.21% <43.90%> (-20.15%)`	⬇️

... and 30 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

arjan-bal · 2025-09-01T06:33:42Z

In the grpc-go repo, we use the assignee to indicate who the review is pending on (author or reviewer). We ignore the reviewer status (the yellow pending re-review dot). Assigning to myself.

arjan-bal · 2025-09-01T10:55:13Z

For the PR title, we add the name of the Go package (e.g. transport), with the prefix if it's ambiguous (e.g. xds/bootstrap), followed by the change description as a complete sentence, written as though it were an order (an imperative sentence), see go/cl-descriptions#first-line.

internal/transport/http2_client.go

arjan-bal · 2025-09-01T11:04:43Z

internal/transport/transport_test.go

+			if test.wantStatusEndStream != nil {
+				want = test.wantStatusEndStream
+			}


I think we should explicitly wantStatusEndStream for all the tests which have non nil values. In table-driven tests, it's fine to duplicate parameters across test cases, even if most share the same value. This practice keeps the tests declarative and simplifies the logic, often to just an equality or diff comparison.

Done, added duplicate parameters

Thanks, we should be able to remove this if block now, right?

internal/transport/transport_test.go

…rpc-content-type-8486

arjan-bal · 2025-09-10T12:41:12Z

internal/transport/http2_client.go

+	//if not a gRPC response - evaluate entire http status and process close stream / response
+	//for 200 -> Unknown, else decode the error


style: Please use complete sentences for comments. Also leave a after //. See https://google.github.io/styleguide/go/decisions#comment-sentences

arjan-bal · 2025-09-11T08:50:01Z

internal/transport/http2_client.go

 		// headerError is set if an error is encountered while parsing the headers
-		headerError string
+		headerError        string
+		receivedHTTPStatus string


nit: We can omit the received prefix int the variable name, shortening it to httpStatus since it doesn't effect readability.

arjan-bal · 2025-09-11T08:57:10Z

internal/transport/http2_client.go

+			statusCode := int(c)
+			if statusCode >= 100 && statusCode < 200 {
+				//In case of informational headers return
+				//For trailers, since we are already in gRPC mode, we will ignore all http statuses and not enter this block


It looks like the check for the endStream flag was removed.

grpc-go/internal/transport/http2_client.go

Lines 1521 to 1528 in 8420f3f

if statusCode >= 100 && statusCode < 200 {

if endStream {

se := status.New(codes.Internal, fmt.Sprintf(

"protocol error: informational header with status code %d must not have END_STREAM set", statusCode))

t.closeStream(s, se.Err(), true, http2.ErrCodeProtocol, se, nil, endStream)

}

return

}

I was confused if endstream can come when it is not a gRPC response - @dfawley confirmed that sometimes only one header comes in the response which is set with end stream and it is possible that it may not be gRPC in which case we still need to throw an error if it is informational header. I have added this code back.

arjan-bal · 2025-09-11T10:20:40Z

internal/transport/http2_client.go

+		case "200":
+			grpcErrorCode = codes.Unknown


Is this explicit check required? Since 200 isn't present in the HTTPStatusConvTab map, I think it may be omitted.

The idea is when no error information can be derived from http status since it is 200, we can simply return Unknown code.

arjan-bal · 2025-09-11T10:23:38Z

internal/transport/http2_client.go

 		if httpStatusErr != "" {
 			errs = append(errs, httpStatusErr)
 		}
+
 		if contentTypeErr != "" {
 			errs = append(errs, contentTypeErr)
 		}
 		// Verify the HTTP response is a 200.


This comment doesn't seem to be accurate. Should we remove it now?

removed it. Not sure why it was present.

arjan-bal · 2025-09-11T10:28:46Z

internal/transport/transport_test.go

+			if test.wantStatusEndStream != nil {
+				want = test.wantStatusEndStream
+			}


Thanks, we should be able to remove this if block now, right?

arjan-bal · 2025-09-11T10:30:56Z

internal/transport/transport_test.go

 			metaHeaderFrame: &http2.MetaHeadersFrame{
 				Fields: []hpack.HeaderField{
 					{Name: ":status", Value: "100"},
 				},
 			},
-			httpFlags: http2.FlagHeadersEndStream,


I think the test was correct before these changes. This test was recently added to fix a bug: #8485

Agreed. Reverted it back.

arjan-bal · 2025-09-11T10:33:05Z

test/http_header_end2end_test.go

@@ -103,10 +103,9 @@ func (s) TestHTTPHeaderFrameErrorHandlingInitialHeader(t *testing.T) {
 		errCode codes.Code
 	}{
 		{
-			name: "missing gRPC status",
+			name: "missing gRPC content type",


In my opinion, a better fix for this test would be to change the assertion to Ok instead, keeping the test params the same. codes.Ok has the zero value for the integer type, so it can be omitted, see https://google.github.io/styleguide/go/decisions#zero-value-fields

there is a test for missing gRPC status already. The expected outcome will not be Ok, it would be an Internal error since gRPC content type is mandated to have the status.

arjan-bal · 2025-09-11T10:41:44Z

test/http_header_end2end_test.go

-			name: "malformed grpc-status-details-bin field with status 200",
+			name: "malformed grpc-status-details-bin field with status 404 and no content type",


I think we should keep this test for malformed grpc-status-details-bin and add a new test for the 404 and no content type case if necessary. This would ensure we have coverage for the existing scenario.

The status 200 or 404 is immaterial in case of content type being application/grpc, the previous test already caters to that and we are finally checking the malformed grpc-status-details-bin field which we were not doing earlier. Hence, expected outcome code in the previous test is no longer Unimplemented (due to 404), the malformed grpc-status-details-bin now results in an Internal error code.

mbissa added this to the 1.76 Release milestone Sep 1, 2025

mbissa added the Type: Bug label Sep 1, 2025

mbissa requested a review from arjan-bal September 1, 2025 06:27

mbissa self-assigned this Sep 1, 2025

arjan-bal assigned arjan-bal and unassigned mbissa Sep 1, 2025

mbissa requested a review from dfawley September 1, 2025 06:34

arjan-bal reviewed Sep 1, 2025

View reviewed changes

internal/transport/http2_client.go Outdated Show resolved Hide resolved

arjan-bal assigned mbissa Sep 1, 2025

arjan-bal reviewed Sep 1, 2025

View reviewed changes

arjan-bal assigned dfawley and unassigned arjan-bal Sep 1, 2025

dfawley removed their assignment Sep 3, 2025

addressing review comments for grpc#8486

b1a7265

mbissa force-pushed the ignore-status-for-grpc-content-type-8486 branch 2 times, most recently from e1f6dbc to f123124 Compare September 9, 2025 04:31

Merge remote-tracking branch 'origin/master' into ignore-status-for-g…

b4cce55

…rpc-content-type-8486

mbissa force-pushed the ignore-status-for-grpc-content-type-8486 branch from f123124 to b4cce55 Compare September 9, 2025 04:34

arjan-bal changed the title ~~ignoring http status when content type is application/grpc grpc#8486~~ client : Ignore http status for gRPC mode and return Internal error for missing grpc status Sep 10, 2025

arjan-bal assigned arjan-bal and dfawley and unassigned mbissa Sep 10, 2025

mbissa changed the title ~~client : Ignore http status for gRPC mode and return Internal error for missing grpc status~~ client : Ignore http status for gRPC mode Sep 10, 2025

mbissa changed the title ~~client : Ignore http status for gRPC mode~~ client transport : ignore http status for gRPC mode Sep 10, 2025

mbissa removed the request for review from dfawley September 10, 2025 17:59

arjan-bal reviewed Sep 11, 2025

View reviewed changes

arjan-bal assigned mbissa and unassigned arjan-bal Sep 11, 2025

addressing review comments for grpc#8486

7087664

		//if not a gRPC response - evaluate entire http status and process close stream / response
		//for 200 -> Unknown, else decode the error

	if statusCode >= 100 && statusCode < 200 {
	if endStream {
	se := status.New(codes.Internal, fmt.Sprintf(
	"protocol error: informational header with status code %d must not have END_STREAM set", statusCode))
	t.closeStream(s, se.Err(), true, http2.ErrCodeProtocol, se, nil, endStream)
	}
	return
	}

		name: "malformed grpc-status-details-bin field with status 200",
		name: "malformed grpc-status-details-bin field with status 404 and no content type",

client transport : ignore http status for gRPC mode #8548

Are you sure you want to change the base?

client transport : ignore http status for gRPC mode #8548

Uh oh!

Conversation

mbissa commented Sep 1, 2025

Uh oh!

codecov bot commented Sep 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

arjan-bal commented Sep 1, 2025

Uh oh!

arjan-bal commented Sep 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mbissa Sep 12, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mbissa Sep 12, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

codecov bot commented Sep 1, 2025 •

edited

Loading

arjan-bal commented Sep 1, 2025 •

edited

Loading

mbissa Sep 12, 2025 •

edited

Loading

mbissa Sep 12, 2025 •

edited

Loading