Output operation on closed SSL stream

[MattSkinner]

(handler-bind ((dex:http-request-failed #'dex:retry-request))
    (dexador:post "https://api001.backblazeb2.com/b2api/v1/b2_get_upload_url" 
        :content "{\"bucketId\": \"b78178f0df3db6975dd20544\"}" 
        :headers (list (cons "Authorization" "Something"))))

returns

output operation on closed SSL stream
   [Condition of type SIMPLE-ERROR]

Restarts:
 0: [RETRY] Retry SLIME REPL evaluation request.
 1: [*ABORT] Return to SLIME's top level.
 2: [ABORT] abort thread (#<THREAD "repl-thread" RUNNING {1002EB7FA3}>)

Backtrace:
  0: ((:METHOD SB-GRAY:STREAM-FORCE-OUTPUT (CL+SSL::SSL-STREAM)) #<unavailable argument>) [fast-method]
  1: (FORCE-OUTPUT #<CL+SSL::SSL-STREAM for #<FD-STREAM for "socket xxx.xxx.xxx.xxx:35752, peer: xxx.xxx.xxx.xxx:443" {10082339E3}>>)
  2: (DEXADOR.BACKEND.USOCKET:REQUEST #<unavailable argument> :METHOD :POST :CONTENT "{\"bucketId\": \"b78178f0df3db6975dd20513\"}" :HEADERS (("Authorization" . "Password")))
  3: ((LAMBDA ()))

It looks like in usocket.lisp an assumption is made that the stream is still open when force-output is called on it. A quick look at cl-plus-ssl would seem to indicate that using ssl-stream-handle can be used to see if a stream is still open. If the stream is not open then maybe the retry code in usocket.lisp should start a new connection?

[fukamachi]

Perhaps, 8a01105 fixes this problem.
Please check if it still happens with the latest Dexador.