Report progress while provisioning, and -v to stream the log - #14
Merged
Conversation
First boot spends minutes in `mise install` -- Ruby, Node, and since #13 the tools the mounted project pins -- while orbx sits in wait_ready polling /etc/sandbox-status and printing nothing. The machine is created and running by then, so from the outside a provision and a hang look identical. The only way to see anything was `orbx logs` from a second terminal, which you have to know about and cannot reach for once orbx has your prompt. The template already logs every step with a timestamp, so nothing new has to be produced -- orbx just never surfaced it. wait_ready now reports: - a spinner + elapsed time, redrawn in place, on a TTY; - one plain line naming the machine, plus `ready in Xm YYs`, when stderr is redirected (CI, an agent driving `orbx run`) -- escape codes there would be noise, and this is what closes the hang ambiguity for an agent; - with -v/--verbose, the guest's provisioning log inline instead. The stream is incremental polling, not a background `tail -f`: orbx::stream_log asks for everything past the line count it was handed and echoes the new count back, so there is no child process to supervise and kill, and a log that does not exist yet simply comes back empty. Latency is one poll interval. A machine that answers `ready` on the very first poll was already provisioned: nothing is being waited on, so wait_ready returns silently and does not replay a log of work that finished days ago. The everyday path -- `orbx` in a project whose sandbox exists -- prints exactly what it printed before, which is nothing. Every exit from the wait funnels through orbx::wait_done (clear the line, drop the trap), the INT trap included: Ctrl-C wipes the spinner's last frame and re-raises with `kill -INT "$BASHPID"`, so an interrupted orbx dies of the signal (130) rather than returning as if the wait had finished. $BASHPID rather than $$ because the script gets sourced into other shells, where the signal must stay with the subshell running the wait. No test can deliver a real Ctrl-C -- a backgrounded job inherits SIGINT ignored, and bash cannot trap what was ignored on entry -- so the tests assert the arm/disarm and the handler's re-raise separately, and AGENTS.md records why, so it is not "fixed" into a flaky signal test later. Verified with 97 bats tests (7 new) and shellcheck clean, plus an end-to-end run of `orbx up` against the suite's fake orb: the spinner redraws in place and clears, -v prints each log line exactly once across polls (tail -n +1 then +7), an already-ready machine prints nothing, and a redirected stderr carries no escape codes.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
First boot spends minutes in
mise installwith no output at all — the machine is already created and running, so a provision and a hang look identical from the outside.orbx logsin a second terminal was the only way to see anything.wait_readynow reports:orbx: provisioning <name> …/orbx: ready in Xm YYs) when stderr is redirected — CI, or an agent drivingorbx run, which is what closes the hang ambiguity for an agent;-v/--verbosestreams the guest's/var/log/sandbox-provision.loginline instead.No template change:
default.yamlalready logs every step: orbx just never surfaced it.Notes:
tail -f—orbx::stream_logasks for everything past the line count it was handed and echoes the new count back. No child process to supervise, and a log that does not exist yet comes back empty.readyon the first poll and returns silently, so the everyday path prints nothing, exactly as before.kill -INT "$BASHPID"), so an interruptedorbxexits 130 instead of returning as if the wait had finished. Tests cannot deliver a real Ctrl-C (a backgrounded job inherits SIGINT ignored, and bash cannot trap what was ignored on entry), so they assert the arm/disarm and the re-raise separately — AGENTS.md records why.97 bats tests (7 new), shellcheck clean.