hyperledger-labs / fabric-smart-client / 36714024032 / 23
91%
main: 91%

DEFAULT BRANCH: main
Ran
Files 479
Run time 35s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 28.403% (+0.01%) from 28.389%
36714024032.23

push

github

web-flow 
fix(view): stop late websocket frames from killing the connection (#1971)

When the multiplexed server closed a sub-connection, whether because the
stream was closed or because its receiver fell behind, any data frame
the client had already sent on that ID looked like the opening frame of
a new sub-connection. The payload is not a StreamMeta, so the peer ID
check failed and the server killed the physical connection with every
other sub-connection on it. The same happened after a cap rejection once
a slot was free again.

Each side now remembers the IDs it closed until the peer acknowledges
with its own Err frame, which peers already send, and drops frames for
them in the meantime. A close the peer started is acknowledged without
waiting, and Kill discards the pending IDs.

Also simplify the transport:
- subConn.ReadMessage is a single channel receive; the done channel is
  gone, since Close closes receiverChan at the same time
- the TLS builders share withExtraCAs, which reads ExtraCAs once
- newServerTLSConfig drops VerifyConnection, whose check duplicated
  RequireAndVerifyClientCert
- MultiplexedProvider.NewClientStream no longer derives ServerName,
  which client.OpenStream already does
- SimpleProvider.NewClientStream wraps the dial error and logs success
  only after the dial succeeds

Fixes #1969

Signed-off-by: Marcus Brandenburger <bur@zurich.ibm.com>

8854 of 31173 relevant lines covered (28.4%)

0.28 hits per line

Source Files on job itest-fabricx-configupdate - 36714024032.23