mcptask-rails-runner 0.3.8 → 0.3.10
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/CHANGELOG.md +69 -0
- data/lib/mcptask_rails_runner/version.rb +1 -1
- metadata +1 -1
checksums.yaml
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
---
|
|
2
2
|
SHA256:
|
|
3
|
-
metadata.gz:
|
|
4
|
-
data.tar.gz:
|
|
3
|
+
metadata.gz: 49073b5f681e3f91650a3a80ed73393195b074f901a9753cbed0d4ce40df9f4f
|
|
4
|
+
data.tar.gz: 11fa1c39be7ea7fa8b6ad658ce3271d2e57226f17c63fd14f1b1defb45faa7a9
|
|
5
5
|
SHA512:
|
|
6
|
-
metadata.gz:
|
|
7
|
-
data.tar.gz:
|
|
6
|
+
metadata.gz: e78f09b82fc40c94a1fc0467603908e1acacc0de9e3f0795c6416b8587e86cae982f83b6cda6b4e788fb6010366a6840c0e7265df07e606dfd5745221e95e675
|
|
7
|
+
data.tar.gz: dc1b48b2dc9734633d05a8b710f0a156e48d7ec6e3e41483a6b8e98f8b929ca981f249fe79d0a75da5b80a9772b35fe11d7f0af166ba59c2c8a93b5abe0650ed
|
data/CHANGELOG.md
CHANGED
|
@@ -5,6 +5,75 @@ entries below describe wrapper changes only. Binary changes are listed in the
|
|
|
5
5
|
[mcptask-releases](https://github.com/jchsoft/mcptask-releases/releases)
|
|
6
6
|
release notes for the same tag.
|
|
7
7
|
|
|
8
|
+
## 0.3.10
|
|
9
|
+
|
|
10
|
+
No wrapper changes. Full notes in
|
|
11
|
+
[mcptask-releases v0.3.10](https://github.com/jchsoft/mcptask-releases/releases/tag/v0.3.10).
|
|
12
|
+
|
|
13
|
+
- **The bundled helpers find Ruby on a machine that does not use rvm.** A Mac
|
|
14
|
+
mini managing Ruby with rbenv got the system 2.6 for every test and CI run, and
|
|
15
|
+
the binstubs died before printing anything worth reading. The helper was written
|
|
16
|
+
where rvm was: it looked for `.ruby-version` in the current directory only,
|
|
17
|
+
probed three fixed paths under the home directory, and matched on what those
|
|
18
|
+
directories were named. None of that covers the case rbenv actually fails on —
|
|
19
|
+
the shell these helpers run in is neither interactive nor a login shell, and
|
|
20
|
+
rbenv, asdf and mise all install themselves from a line in `~/.zshrc` that such
|
|
21
|
+
a shell never reads. They now ask each manager where it keeps the pinned
|
|
22
|
+
version, fall back to its documented layout when the manager itself is
|
|
23
|
+
unreachable, cover rvm, rbenv, asdf, mise, chruby, frum and Homebrew alike, and
|
|
24
|
+
confirm each candidate by asking that interpreter its version rather than
|
|
25
|
+
trusting a directory name. Which interpreter a run got is recorded in its log
|
|
26
|
+
header, and a pinned version the machine does not have says so instead of
|
|
27
|
+
failing later somewhere else.
|
|
28
|
+
|
|
29
|
+
- **The shipped permission baseline no longer names one developer's home
|
|
30
|
+
directory.** It granted read access to `/Users/josefchmel/.rvm` literally,
|
|
31
|
+
which on anybody else's machine grants nothing. The paths are relative to the
|
|
32
|
+
home directory now, and the other version managers are in the list.
|
|
33
|
+
|
|
34
|
+
## 0.3.9
|
|
35
|
+
|
|
36
|
+
No wrapper changes. Full notes in
|
|
37
|
+
[mcptask-releases v0.3.9](https://github.com/jchsoft/mcptask-releases/releases/tag/v0.3.9).
|
|
38
|
+
|
|
39
|
+
- **A ticket the runner declines as too large no longer ends the working day.**
|
|
40
|
+
Reading a bug report and concluding that it is a multi-day feature is a fair
|
|
41
|
+
judgement to reach, but the only word the runner had for it was "failure" — and
|
|
42
|
+
"failure" stops the loop. One such call, made 96 seconds into a run, exited the
|
|
43
|
+
process at 10:17 with seven hours of the day left: exit code zero, a clean run
|
|
44
|
+
in the scheduler's record, and nothing anywhere to say a day had been given up
|
|
45
|
+
over one oversized ticket. The judgement now has an outcome of its own. The
|
|
46
|
+
agent writes its scope assessment on the piece, logs the analysis as the work it
|
|
47
|
+
was, blocks the ticket so triage stops offering it, and the loop moves to the
|
|
48
|
+
next task. A second refusal of the same ticket does still stop the loop —
|
|
49
|
+
that means the block did not take — and it says so rather than exiting quietly.
|
|
50
|
+
|
|
51
|
+
- **A task belonging to a different project is discarded instead of worked.** The
|
|
52
|
+
runner never checked that the piece triage picked came from the project named in
|
|
53
|
+
CLAUDE.md. Upstream, one runner's answer was reaching every connected client, so
|
|
54
|
+
a runner installed for one project picked up another project's task and set
|
|
55
|
+
about it. Triage now reports which project the piece it fetched belongs to, and
|
|
56
|
+
a mismatch is thrown away.
|
|
57
|
+
|
|
58
|
+
- **One launcher log file per run, instead of one that grows forever.** Nothing
|
|
59
|
+
ever truncated the per-project log: 401 MB across projects on one machine, 277 MB
|
|
60
|
+
of it in a single file, and diagnosing a run meant locating its start marker
|
|
61
|
+
inside all of that. Each run gets its own file now, pruned at 30 days or 500 MB
|
|
62
|
+
per project, whichever bites first. Existing installs keep appending to their old
|
|
63
|
+
file until install is re-run, and that file is left alone rather than deleted —
|
|
64
|
+
the disk space is the operator's to reclaim.
|
|
65
|
+
|
|
66
|
+
- **The installer says which account a scheduled job will authenticate as.** On a
|
|
67
|
+
host carrying more than one mcptask account, regenerating a job could silently
|
|
68
|
+
swap its identity: nothing failed, nothing was logged as wrong, and the work
|
|
69
|
+
landed on the wrong dashboard until somebody decoded the token by hand.
|
|
70
|
+
|
|
71
|
+
- **The bundled skills follow the mcptask parameter renames.** The MCP tools are
|
|
72
|
+
moving to names that state which id form each parameter carries. The skills ship
|
|
73
|
+
the new names now with the old ones kept as a transitional note, so a host that
|
|
74
|
+
updates before the server change still matches whichever names its own server
|
|
75
|
+
currently expects.
|
|
76
|
+
|
|
8
77
|
## 0.3.8
|
|
9
78
|
|
|
10
79
|
No wrapper changes. Full notes in
|
|
@@ -10,5 +10,5 @@ module McptaskRailsRunner
|
|
|
10
10
|
# Binary::INSTALL_DIR. So the pin is per project and per foreground rake
|
|
11
11
|
# invocation; the installed copy is per machine, and the newest binary any of
|
|
12
12
|
# those projects offers is the one that stays.
|
|
13
|
-
VERSION = "0.3.
|
|
13
|
+
VERSION = "0.3.10"
|
|
14
14
|
end
|