See go123@515a6d14

- without registering M to tracer it was failing as

  BenchmarkGetObjectNetPipe
      neo_test.go:491: main: expect: *neo.eventMStartReady:
          want: &{m true}
          have: &{ true}
          diff:  {
          - Where: "m",
          + Where: "",
            Ready: true,
           }

- Client is now required to be explicitly Run, but I forgot to update
  thi function in eae3b427.

* master:
  go/*: Adapt to xnet.Networker changes wrt Listen and cancellation
  go/internal/xtesting: DrvTestWatch: Verify that watching works with empty transactions
  go/zeo: tests: Terminate spawned processses with SIGTERM instead of SIGKILL
  go/zodb/zeo: Factor-out verifying empty database into xtesting.DrvTestEmptyDB
  go/neo/t: .nxdtest: nxdtest changes summaryf signature to include stderr

See

	go123@3354b401
and

	go123@b03d65ff

The wrapping logic in LinkListener goes away because Accept from xnet
now supports cancellation.

previously

- we were explicitly calling it on errors.Cause(err)
- and not in every place.

See

	go123@3354b401
and

	go123@b03d65ff

The wrapping logic in LinkListener and requireIdentifyHello goes away
because Accept from xnet now supports cancellation.

NEO/py no longer writes config.partition at all. This way sqlite backend
fails to open any database:

    xtesting.go:453: start neo/go /tmp/neo375346404/1: /tmp/neo375346404/1.sqlite: NEO/go POC: not ready to handle: sqlite:///tmp/neo375346404/1.sqlite: config: partitions: not found

- Move ca/cert/key out of query part into credentials part of URL
- As a consequence move cluster name out of credentials part into path part
- Allow both neo:// and neos:// schemes to be used.
  neo:// means alwasy without SSL and neos:// means always with SSL.
  Even if neos:// URL comes without credentials embedded into it - they
  can be obtained from elsewhere.

I need this change in wendelin.core 2 which needs to take a zurl and
compute a mountpoint patch that corresponds to it, so that multiple
several zopes that use the same NEO storage, could be associated with
the same single WCFS instance.

See next patch.

Tighten up security. NEO/py does something similar:

https://lab.nexedi.com/nexedi/neoppod/blob/v1.12-61-gc1c26894/neo/lib/app.py#L76-78

On CPython 2.7.18:

    In [7]: list((k for k in dir(ssl) if k.startswith("PROTOCOL_TLSv")))
    Out[7]: ['PROTOCOL_TLSv1', 'PROTOCOL_TLSv1_1', 'PROTOCOL_TLSv1_2']

It was not the case with ZEO4 server because we did not had
ZEO@bf80d23d yet.

For ZEO it is not strictly required, but for upcoming NEO for example
NEOCluster needs to shutdown gracefully, else there are processes left for e.g.
storage nodes and they dump somthing as below on the terminal after tests
completion:

    === RUN   TestLoad
    2020/10/21 14:33:00 zodb: FIXME: open ../zodb/storage/fs1/testdata/1.fs: raw cache is not ready for invalidations -> NoCache forced
    === RUN   TestLoad/py
    I: runneo.py: /tmp/neo445013868/1: Started master(s): 127.0.0.1:24661
    WARNING: This is not the recommended way to import data to NEO: you should use the Importer backend instead.
    NEO also does not implement IStorageRestoreable interface, which means that undo information is not preserved when using this tool: conflict resolution could happen when undoing an old transaction.
    Migrating from ../zodb/storage/fs1/testdata/1.fs to 127.0.0.1:24661
    Migration done in 0.19877
    --- PASS: TestLoad (0.75s)
        --- PASS: TestLoad/py (0.74s)
    PASS
    ok      lab.nexedi.com/kirr/neo/go/neo  0.749s
    (neo) (z-dev) (g.env) kirr@deco:~/src/neo/src/lab.nexedi.com/kirr/neo/go/neo$ Traceback (most recent call last):
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/tests/functional/__init__.py", line 182, in start
        getattr(neo.scripts,  command).main()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/scripts/neostorage.py", line 66, in main
        app.run()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/storage/app.py", line 147, in run
        self._run()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/storage/app.py", line 178, in _run
        self.doOperation()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/storage/app.py", line 266, in doOperation
        poll()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/storage/app.py", line 87, in _poll
        self.em.poll(1)
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/lib/event.py", line 155, in poll
        self._poll(blocking)
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/lib/event.py", line 253, in _poll
        timeout_object.onTimeout()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/lib/event.py", line 259, in onTimeout
        on_timeout()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/storage/database/manager.py", line 207, in _deferredCommit
        self.commit()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/storage/database/manager.py", line 193, in commit
        self._commit()
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/storage/database/sqlite.py", line 90, in _commit
        retry_if_locked(self.conn.commit)
      File "/home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/neo/storage/database/sqlite.py", line 45, in retry_if_locked
        return f(*args)
    OperationalError: disk I/O error

And use it to verify other storages.

See nexedi/nxdtest@53064e71

nexedi/nxdtest@53064e71

It was not the case with ZEO4 server because we did not had
ZEO@bf80d23d yet.

* master:
  go/zodb/zeo: Correctly prefix zLink-related errors with operational context
  go/{zeo,neo}: handshake: Report "unexpected EOF" if peer sends only partial data
  go/{zeo,neo}: Don't tolerate if a packet reads only partially
  go/zodb/fs1: Factor-out noEOF and okEOF into common place (-> internal/xio)

Go through zrpc and add error context to every function that can create
error. This is similar to what NEO does in neonet. The reason to do this
was the following obscure error from wcfs:

    E1105 11:32:33.295497   24639 wcfs.go:2505] zwatch zeo://:28359: EOF

which, after this patch, is now

    E1105 12:53:15.922052   30731 wcfs.go:2510] zwatch zeo://:27024: zlink 127.0.0.1:47768 - 127.0.0.1:27024: recvPkt: EOF

NEO was already doing this, but let's use xio.NoEOF everywhere for
uniformity.

If remote peer closes the link half-way of packet data - turn EOF into
ErrUnexpectedEOF.

We will need to use those utilities for ZEO and NEO.

Go through zrpc and add error context to every function that can create
error. This is similar to what NEO does in neonet. The reason to do this
was the following obscure error from wcfs:

    E1105 11:32:33.295497   24639 wcfs.go:2505] zwatch zeo://:28359: EOF

which, after this patch, is now

    E1105 12:53:15.922052   30731 wcfs.go:2510] zwatch zeo://:27024: zlink 127.0.0.1:47768 - 127.0.0.1:27024: recvPkt: EOF

NEO was already doing this, but let's use xio.NoEOF everywhere for
uniformity.

If remote peer closes the link half-way of packet data - turn EOF into
ErrUnexpectedEOF.

We will need to use those utilities for ZEO and NEO.

* master:
  go/neo/neonet: Increase signal/noise in dump of packets with decode problems
  fixup! go/neo/proto: Switch enum encoding from int32 to int8
  go/zodb/zeo: Load: loadBefore RPC can return None to indicate non-existing object
  go/internal/xtesting: DrvTestLoad: Verify load of non-existing and not-yet-created objects

We recently hit "decode: buffer overflow" errors due to mismatch in
between NEO/go and NEO/py (see previous patch). With dumpio=true that
problem was showing itself as

    127.0.0.1:50624 > 127.0.0.1:41863: .5 GetObject &{0000000000000000 ffffffffffffffff 0285cbac258bf266}
    127.0.0.1:50624 < 127.0.0.1:41863: .5 (proto.Error) decode: buffer overflow; #24 [24]: 00 00 00 03 00 00 00 10 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 Error ACK ;  [24]tail: 00 00 00 03 00 00 00 10 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30
        xtesting.go:306: load 0285cbac258bf265:0000000000000000: returned err unexpected:
            have: neo://1@127.0.0.1:24078: load 0285cbac258bf265:0000000000000000: 127.0.0.1:50624 - 127.0.0.1:41863 .5: decode: decode: buffer overflow
            want: neo://1@127.0.0.1:24078: load 0285cbac258bf265:0000000000000000: 0000000000000000: object was not yet created

Here after printing error and dumping all bytes from packet payload, it
was printing again message type, message value from to-be-decoded place
(which is zero-initialized) and data bytes again.
-> Fix it not to print anything after dump of payload data:

    127.0.0.1:60518 > 127.0.0.1:46719: .5 GetObject &{0000000000000000 ffffffffffffffff 0285cbac258bf266}
    127.0.0.1:60518 < 127.0.0.1:46719: .5 (Error) decode: buffer overflow; #24 [24]: 00 00 00 03 00 00 00 10 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30
        xtesting.go:306: load 0285cbac258bf265:0000000000000000: returned err unexpected:
            have: neo://1@127.0.0.1:27853: load 0285cbac258bf265:0000000000000000: 127.0.0.1:60518 - 127.0.0.1:46719 .5: decode: decode: buffer overflow
            want: neo://1@127.0.0.1:27853: load 0285cbac258bf265:0000000000000000: 0000000000000000: object was not yet created