You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am making progress with ooni/probe#2531 and
I want to reactor model/netx.go such that the TLSHandshaker returns a
model.TLSConn rather than a net.Conn.
Returning a net.Conn and documenting it is a model.TLSConn is bad
compared to returning a model.TLSConn directly.
Note that we cannot apply the same transformation to netxlite's
TLSDialer.DialTLSContext because such a method must be assignable to
net/http and github.com/ooni/oohttp's Transport function also called
DialTLSContext.
The fact that we need code to be assignable to the Transport function is
what historically led the TLSHandshaker to return a net.Conn as well.
But it was quite clear from the get go that this choice led to some
quirks (and, in fact, this behavior was explicitly documented as such).
While there, slightly refactor `internal/experiment/echcheck/utls.go` to
avoid storing the conn inside the handshaker and make sure the test
coverage does not drop for this experiment.
While there, note that ooni/probe#2538 exists
and commit a mitigation.
I am making progress with ooni/probe#2531 and
I want to reactor model/netx.go such that the TLSHandshaker returns a
model.TLSConn rather than a net.Conn.
Returning a net.Conn and documenting it is a model.TLSConn is bad
compared to returning a model.TLSConn directly.
Note that we cannot apply the same transformation to netxlite's
TLSDialer.DialTLSContext because such a method must be assignable to
net/http and github.com/ooni/oohttp's Transport function also called
DialTLSContext.
The fact that we need code to be assignable to the Transport function is
what historically led the TLSHandshaker to return a net.Conn as well.
But it was quite clear from the get go that this choice led to some
quirks (and, in fact, this behavior was explicitly documented as such).
While there, slightly refactor `internal/experiment/echcheck/utls.go` to
avoid storing the conn inside the handshaker and make sure the test
coverage does not drop for this experiment.
While there, note that ooni/probe#2538 exists
and commit a mitigation.
I have seen this test of the
./internal/testingsocks5
package fail twice in ooni/probe-cli#1281.The full log of a failed tests run is here:
I am going to commit a mitigation for this test being flaky. We should investigate and fix it when possible.
The text was updated successfully, but these errors were encountered: