BitSquare/Codeman
2
0
Fork 0
mirror of https://github.com/Ark0N/Codeman.git synced 2026-09-30 12:39:42 +02:00
Code Issues Packages Projects Releases Wiki Activity

docs(sessions): correct what a failed rebuild is actually likely to be

Browse Source 
Ran the feature against a real server for the first time, on an isolated
instance, and two claims in the code turned out to be wrong.

A rebuild that fails after the session is registered was documented as
commonly caused by a CLI binary missing from a freshly booted machine's
PATH. It is not: the resolver finds its binary by absolute path, so PATH
never enters into it, and a server started without claude on PATH restored
every session normally. Nor does an un-enterable workspace fail — tmux falls
back to another directory and the pane comes up there. Neither obvious cause
throws, so the discard path is defended rather than expected, and the
comments now say that instead of naming a cause that cannot happen.

The four review rounds that shaped this path all reasoned about a trigger
none of them could test. The path itself is still worth having, since a mux
failure would reach it, but its comments should not claim a likelihood the
machine disagrees with.

Refs #411

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Michael Grundberg
2026-09-17 08:26:29 +02:00
co-authored by Claude Opus 5
parent 39976041e0
commit 18ab2ab595
2 changed files with 17 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/routes/reboot-restore-rebuild-failure.test.ts
+9 -4
View File
@@ -4,10 +4,15 @@
* The other route test file deliberately uses workspaces that do not exist, so it
* never reaches `new Session()`. This one mocks the `Session` module so the route
* runs its whole construction path — `addSession`, `setupSessionListeners`,
* `reapplyPersistedSessionState`, `startInteractive` — and then throws where a
* real one would when the CLI binary is missing from a freshly booted machine's
* PATH. Without the mock there is no way to exercise that path, which is how the
* original version of this route shipped a session leak the tests could not see.
* `reapplyPersistedSessionState`, `startInteractive` — and then throws.
*
* The mock is the only way in. Driven against a real server, `startInteractive()`
* does not throw for either obvious cause: the CLI resolver finds its binary by
* absolute path rather than through PATH, and tmux falls back to another
* directory rather than failing when it cannot enter the workspace. A mux-layer
* failure is what is left, and it cannot be provoked from a test. Without the
* mock this path would go unexercised, which is how the original version of this
* route shipped a session leak the tests could not see.
*
* It also covers the session caps, because those too are only reachable once the
* route is actually willing to build something.