Repository navigation
test(generators): HTTP/2 sample frames are all written on stream 0 #1156
Description
Describe the bug
examples/generators/options.py _httpv2_build writes every HTTP/2 frame on stream 0. That puts illegal stream-0 DATA, HEADERS and similar frames in options-transport.pcap. Once #1153 lands, those frames decode as Raw over TCP/80. test_option_coverage_runtime checks only the Ethernet:IPv4 prefix, so it does not notice.
Expected behavior
Each frame uses a stream ID that is legal for its type (RFC 9113 §6).
Blocked until
#1132 and #1153 merge. #1132 edits options.py.
Activity
JarryShaw commentedon Oct 6, 2026
Also from the #1153 review:
_httpv2_buildstill writes the HTTP/2 Length counting the 9-octet header, for example000009000000000000. Since #1138, the parser rejects that, sooptions-transport.pcapframes 35–43 decode as Raw onmainregardless of #1153. Fix the Length convention here as well as the stream IDs.JarryShaw commentedon Oct 6, 2026
Unblocked: #1132 and #1153 have both merged. A worker is on it now.
JarryShaw commentedon Oct 6, 2026
Correction to my earlier comment here: the generator's HTTP/2 Length field was already payload-only on
2c7a97c84, because #1138 fixedHTTP.make, so it was not the cause.The real cause is that
_httpv2_buildputs every frame on stream 0 (examples/generators/options.py:1129-1130). Since #1153,_test_http2_framerejects stream 0 for DATA, HEADERS, PRIORITY, RST_STREAM, PUSH_PROMISE and CONTINUATION frames. The fix is in #1161.