b6f81b51

By: Tom Sydney Kerckhove <syd@cs-syd.eu>

One exec of one binary for a deployment

deploy-for.nix built three wrappers, each execing the tool with a
different subcommand and its flags, because a deployment needed apply and
the two things a stopped deployment needs. It builds one now: nix writes
a file saying which router this is and where the halves are, and the
wrapper is one exec of one binary with one flag.

  nix run .#deploy-ccr2004                 applies
  nix run .#deploy-ccr2004 -- download     reads the router into the record
  nix run .#deploy-ccr2004 -- import-rest  finishes a stopped deployment

Which router is being talked to moved from in front of every subcommand
into the ones that need it, since router is told by its file and the rest
are told on the command line.

--download-with and --finish-with are gone with the wrappers they named:
the messages take argv[0] and add the subcommand, so they name whatever
the reader actually typed.

Two other duplications while looking: run_command_remotely was
capture_command_remotely with the answer thrown away and its own copy of
the channel handling, including the last unwrap on a read in that file;
and download_export was fetch_export with a write on the end.

The vm test's environment is runtimeEnv for everything known when it is
built, which is all of it but the disk it makes and the two a caller
passes in.
GitHub

Total

20m25s

Created Started Finished

Time to Start Worker time Duration Time to finish Idle
Config 1h41m17s 0s 0s 1h41m18s 1h41m17s
Eval 1h41m21s 9s 9s 1h41m30s 2s
Build 1h41m30s 17m22s 20m13s 2h01m43s 11m51s
Suite 1h41m17s 17m32s 20m25s 2h01m43s 1h53m11s

Timeline

Config Eval Build Failed Retried Compressed gap
0s1h45m1h46m1h47m1h48m1h49m1h50m1h51m1h52m1h53m1h54m1h55m1h56m1h57m1h58m1h59m2h0m2h1m