better unit tests for mixins

[Antoine Martin]

Follow up from #1761. See also #1778

We already have basic unit tests for client and server mixins:

• r18558 server: mmap and remote logging
• r18534 client: mmap and network state
• r18535 client: remote logging

Only getting good test coverage will ensure that each mixin is independent from the others.

[Antoine Martin]

• r23106 adds an audio client mixin test, new issues to deal with:
  • duplication of self._remote_machine_id
  • python2 segfaults (meh)
  • first audio packets may come out of order! ('sound-data' before 'sound-control')
• r23107 split test
• r23108 adds a clipboard test, new issues:
  • run under X11 - needs to start a new display
  • verify clipboard state externally? (using xclip?)

[Antoine Martin]

Added more client mixins tests - now mostly complete (just without rpc test):

• r23112 display
• r23113 encodings
• r23114 notifications + refactoring
• r23115 server info
• r23116 tray
• r23117 + r23118: webcam, has dependencies on "server_ping_latency" and "server_encodings"
• r23119 window manager (very limited)

Still TODO:

• test multiple times with different options combinations / different server capabilities
• test coverage
• server mixins tests

[Antoine Martin]

Server mixins tests:

• r23120 refactoring + audio test
• r23121 refactoring
• r23122 child commands - needs _upgrading flag and session_name..
• r23123 mixin test improvements for:
• r23125 clipboard
• r23126 display manager needs _server_sources
• r23127 encoding
• r23128 file-print
• r23129 input needs _server_sources
• r23131 network state
• r23132 notification forwarder
• r23133 webcam
• r23134 window server

New tasks:

• source mixin tests
• parse some network packets
• control commands: split them up?

[Antoine Martin]

• r23139 refactoring so we can test packets without an actual connection
• r23140 re-order mixins init to make them more resilient
• r23141 test sending audio packets. (audio mixin cannot assume that machine_id or uuid exist)
• r23142 test starting a new command and then sending it a signal
• r23143 encoding mixin - cannot assume that _idle_refresh_all_windows, send_windows and ui_client exist (heavy dependency on window mixin)
• r23144 fixup file-print test
• r23147 remote-logging packet test
• r23149 fixes a huge bug in the ping / client-latency calculations
• r23151 all network state packets - cannot assume we have statistics object
• r23152 test notification packets
• r23155 + r23157 move file-print and network-state packet handlers out of the UI thread
• r23156 logging locking and cleanup

New tasks:

• display, input and clipboard tests need to start their own display
• file-print: needs to send chunks and verify file gets created
• input test: could be used to test character sets and foreign keyboard layouts (hard!)
• notification code needs to be moved to a source mixin

[Antoine Martin]

Updates:

• r23166 fixes a test error in the command-server test now that we pass a fake source mixin to the server mixin
• r23167 + r23168: fix stop audio from client using sequence number, allowing:
• r23169 audio client test made deterministic thanks to fixes above
• r23170 audio server prevents spurious audio packets after stop request

Updated TODO:

• audio needs test mp3 data to test receiving
• verify all server source mixins have a server test
• control commands
• file-print: needs to send chunks and verify file gets created
• many tests need to start an X11 display
• test coverage: #2356
• test multiple times with different options combinations / different server capabilities
• combine client and server mixins in one test: #2357
• input test: could be used to test character sets and foreign keyboard layouts (hard!)
• notification code needs to be moved to a source mixin

[Antoine Martin]

Test updates:

• audio:
  • r23171 don't run for too long
  • r23172 + r23173 refactoring
  • r23174 test audio with opus data
  • r23175 more complete audio tests
  • r23213 avoid pulseaudio warnings
• wayland vs x11 issues: r23176, r23250 (fix for win32..)
• cleanups / cosmetic: r23177, r23178, r23186, r23192, r23194, r23195, r23196, r23197, r23208, r23209, r23210, r23216, r23217, r23222, r23234, r23248, r23249, r23251, r23258, r23259
• refactoring: r23181, r23184, r23185, r23187, r23190, r23191, r23199, r23254
• clipboard: r23179, r23219 (python3 support), r23232
• display fix: r23182
• network state fixes: r23183, r23207 (warning)
• tests that require a vfb: r23180, r23211, r23212, r23231
• authentication test fixes (python3): r23188
• socket tests (python3 fixes etc): r23189
• x11 bindings usage bug fix: r23193
• webcam tests: r23198
• file-print needs fake sockets: r23201
• new test for util modules: r23220 AtomicInteger, r23221, r23238 (bugs fixed in r23237 + r23240)
• build / packaging updates: r23235
• notifications test gets its own dbus instance: r23253
• bandwidth-limit: r23257

Bugs found:

• cython rectangle tests no longer fired after refactoring: r23224, r23225
• fd leak bug: r23202
• invalid command bug: r23204
• clipboard: r23215
• server startup failure handling: r23223
• ssl test failures: r23226
• python3 server socket upgrade failures: r23228 (+r23229 logging error)
• win32 new monotonic_time code was wrong: r23233
• don't allow invalid bandwidth limits: r23236

Will follow up in #2356, #2357 and #2362.

[migration script]

this ticket has been moved to: ​https://github.com/Xpra-org/xpra/issues/1773