Two related v5-sweep fixes for the online/runtime tool family:
1. Auto-login helper for headless mode
In headless mode each MCP call spawns a fresh CODESYS --noUI process,
so the login state established by connect_to_device dies before the
next call. Pre-fix, only connect_to_device and download_to_device did
their own login(); the other four (start_stop_application,
read_variable, write_variable, read_running_version_online) silently
failed in headless with 'Application not logged in.' (start/stop) or
'Invalid expression' (read/write). They worked in persistent mode
only because the login carried across calls.
Added ensure_logged_in(online_app, login_wait_seconds=30) to
ensure_online_connection.py. Idempotent: short-circuits via
online_app.is_logged_in (persistent mode is a no-op, no extra login
roundtrip). When not logged in, runs the same enum-probe + call-shape
probe + STABLE_STATES settle-wait pattern as connect_to_device.py.
Added to start_stop_application.py, read_variable.py,
write_variable.py, read_running_version_online.py.
2. _MCP_PROJECT_VERSION GVL emitted as plain VAR_GLOBAL, not CONSTANT
CODESYS inlines VAR_GLOBAL CONSTANT scalars at compile time and
strips them from the online symbol table. The whole point of
_MCP_PROJECT_VERSION.sVersion is to be readable live from the
running PLC, so CONSTANT was the wrong storage class.
read_running_version_online failed against EVERY project bumped via
the old template -- 'Invalid expression' on the runtime read.
Dropped CONSTANT from VERSION_GVL_DECLARATION_TEMPLATE in
bump_project_version.py. Existing projects auto-migrate on the next
bump because maintain_version_gvl()'s existing-GVL branch overwrites
textual_declaration with the (now non-CONSTANT) template. The string
is still effectively read-only at runtime -- only the bump tool
updates it.
read_running_version_online.py also got a more precise error message
that explicitly fingerprints the 'Invalid expression' failure mode
and points at the CONSTANT root cause. Useful for any user landing
on a project that pre-dates this fix.
Verified end-to-end against local CODESYS Control Win V3 (PLATEA, port
11740) on MCPTest2 v1.3.4.0:
- connect_to_device, get_application_state, download_to_device,
start_stop_application (both directions), read_variable
(PLC_PRG.watchdog1 = 225 ticking), write_variable (200 -> 204 in 4s
proves write took), disconnect_from_device: all 7 PASS.
- read_running_version_online failure reproduced (CONSTANT inlined),
fix landed -- next bump on MCPTest2 will validate.
37/37 unit/integration tests green. TEST_OVERVIEW.md updated with the
v5 device sweep, with the headless-mode deep-dive, and with the
broken-by-design notes on read_running_version_online.
Prior behaviour: lm.add_library(LIBRARY_NAME) was called with a string,
which always hits the placeholder overload of ScriptLibManObject.add_library
(see helpme-codesys.com "ScriptLibManObject" / local SP22 stub
ScriptLib/Stubs/scriptengine/ScriptLibManObject.pyi). If the named
placeholder is not registered in the IDE, the resulting reference fails
to resolve at load time and the next project open throws
Library Manager: Error: Could not open library 'X'.
(Reason: The placeholder library 'X' could not be resolved.)
after which script_engine.projects.primary returns None and the project
is effectively bricked until the binary is reverted. add_library reported
SUCCESS in this scenario.
Fix:
1. Pre-resolve LIBRARY_NAME via the IDE-level
library_manager.find_library(name) global. If found, pass the
resulting ManagedLib to lm.add_library(...) -- the V3.5.5.0 ManagedLib
overload which produces a managed reference instead of a placeholder.
2. After the add, walk lm.references to locate the new entry and verify
it resolved (managed -> always; placeholder -> non-empty
effective_resolution per ScriptPlaceholderReference in the stub).
3. If the reference did not resolve, call lm.remove_library(name) to back
out the bad reference and refuse to save the project, returning an
actionable error instead.
Also tightened the libman lookup to use the documented
has_library_manager / get_library_manager() container API
(ScriptLibManObjectContainer in the stub) instead of name-searching for
"Library Manager" as a tree node, matching what list_project_libraries.py
already does. The legacy name-search fallback is preserved for older SPs.
Regression test added in tests/integration/e2e.test.ts asserting the
rendered script template carries the resolution gate, the managed-overload
preference, the back-out call, and that primary_project.save() in the
add_library body lives downstream of the _is_resolved gate.
Followup: lm.remove_library(name) is documented for SP22 and is the
clean back-out path. On SPs that lack remove_library (none observed in
the 3.5.21+ docs but possible on truly old branches) the script reports
the constraint and exits non-zero rather than silently saving a bad
reference.
Three changes:
1. Fix the bench harness's set_pou_code case. The previous version
passed POU_PATH / DECLARATION_CODE / IMPLEMENTATION_CODE, but the
script template expects POU_FULL_PATH / DECLARATION_CONTENT /
IMPLEMENTATION_CONTENT. Plus the recently-added SET_DECLARATION /
SET_IMPLEMENTATION boolean flags from 35abc8c. With these
corrections, set_pou_code passes for the first time in the bench.
2. Add bench-results-v5.json from a fresh persistent-mode run on
MCPTest2 v1.3.4.0. All 10 cases PASS.
3. Update TEST_OVERVIEW.md with the v5 numbers alongside the v1
historical numbers. v5 is ~2x faster across the board than v1
(likely the cumulative effect of the ScriptManager cache removal +
SP22 Patch 1 IPC improvements). first-cold open_project is slower
than v1 (10.6 s vs 7.7 s) -- one-shot, noise probably; warm
already-open call dropped from 740 ms to 314 ms.
The v5 set_pou_code pass is the third end-to-end signal that the
omitted-decl wipe fix is good (after the e2e regression test and the
live PLC_PRG restoration earlier today).
Calls primary_project.get_compilerversion() (ScriptEngine 4.2.0.0+) and
emits the result through the JSON payload. Renders as:
- library.md: a 'Project compiler version' row in the Versions table
- list_project_libraries chat output: a 'Compiler version: X.Y.Z.W'
line in the header section
Motivation: changing the project's compiler version (Project > Project
Settings > Compiler version, or set_compilerversion_to_newest()) only
touched the .project binary -- mirror_export couldn't see it, so the
release classifier had to fall back to SHA comparison and emitted the
generic 'device-tree / library refs / task config / visu / Save() touch'
classification. Compiler-version changes now leave a textual diff in
mcp-mirror/library.md, letting the classifier issue an honest revision
bump instead of the bare build-bump SHA fallback.
Defensive: get_compilerversion() is wrapped in try/except so older
ScriptEngines (< 4.2.0.0) that lack the method don't crash the tool;
they just emit compiler_version=null and the field is omitted from
output.
Bug: calling set_pou_code with implementationCode only (declarationCode
omitted) would wipe the POU's PROGRAM/VAR...END_VAR block in the binary.
After such a call mirror_export classified the POU as 'UNKNOWN' (no
PROGRAM/FUNCTION_BLOCK keyword in the empty declaration), and the var
block disappeared from the .st mirror file.
Root cause: the TS wrapper substituted '' (empty string) into the Python
template when declarationCode was undefined, giving DECLARATION_CONTENT
= "". The Python script then took the truthy-ish branch (empty string
is not None) and called decl_obj.replace('') -- wiping textual_decl.
Fix: pass explicit SET_DECLARATION / SET_IMPLEMENTATION boolean flags
from the TS wrapper, gate the replace() calls on those flags. Empty
string remains a valid intentional value (caller wants to wipe).
- Reproduced on MCPTest2: PLC_PRG declaration block was wiped between
v1.3.0.0 and v1.3.1.0 by exactly this code path.
- Regression test added in tests/integration/e2e.test.ts covering the
omitted-declarationCode path.
- Existing set_pou_code test extended to assert SET_DECLARATION /
SET_IMPLEMENTATION are emitted in the rendered script.
The dual-SHA tracking commit (146d950) wrote the tag annotation with
`git tag -a -m JSON.stringify(body)`. JSON.stringify escapes newlines
as the literal two-char sequence "\n", and the shell passes those
through unchanged -- so git stored the body as one big line with
literal "\n" chars instead of real LF bytes.
The reader (readTagShas) used a multiline regex anchored on `^` and
`$`, which doesn't match across literal "\n" -- so v1.3.2.0's tag was
written with SHAs in the body but they're invisible to the next
release's read-back.
Two fixes:
- Reader (readTagShas): normalise literal "\n" sequences to real
newlines before applying the regex. Backward-compatible -- handles
the v1.3.2.0 tag transparently and works on properly-formed tags
from v1.3.3.0 onward too.
- Writer (release_project_version step 8): write the body to a
temp file and use `git tag -F <tempfile> --cleanup=verbatim` so
real LF bytes go in. Also adds os import for os.tmpdir().
Verified locally: dist/server.js loads cleanly. End-to-end behaviour
confirmable on the next release_project_version call against any
project.
Implements bidirectional change detection for the release pipeline.
Two SHA-256 fingerprints are now stored in every release tag's annotated
body:
project-sha256: <hash of the .project binary>
mirror-sha256: <hash of the mcp-mirror/ tree>
On the next release_project_version call, these are read back via
git cat-file -p <prior-tag> and compared against the current values
to detect three classes of change that the mirror-only diff missed:
(a) binary changed AND mirror unchanged (working tree, before
mirror_export). Normal "user edited in IDE" path. Classifier
handles this as it always did.
(b) binary unchanged AND mirror changed (working tree, before
mirror_export). User edited .st files in mcp-mirror/ directly
with a text editor. mirror_export is about to overwrite those
edits, so we surface a WARNING in the release log. Future:
a mirror_import tool would push these back into the binary;
until then, mirror is one-way (binary -> mirror).
(c) binary changed AND mirror UNCHANGED after mirror_export. The
.project binary has a non-textual change that mirror_export
doesn't capture: device tree, library refs, task config,
visualizations, OPC UA / symbol config, application composer,
or just a Save() touch (CODESYS embeds timestamps). Classifier
sees no diff but project SHA flipped. Promote 'no-changes' to
a build-level bump so the version still ticks. The Changelog
entry calls out the SHA-fallback evidence so it's visible in
review.
Helper functions added at module scope:
- sha256OfFile(filePath): single-file SHA-256.
- sha256OfDirectory(dirPath): deterministic tree walk, sorts
entries by name, hashes (relative-path, content) pairs separated
by NULs.
- readTagShas(projectDir, tagName): parses project-sha256 /
mirror-sha256 lines out of an annotated tag body. Returns
undefined for either field if missing -- gracefully handles
older tags that don't carry the fingerprints.
The dual-SHA approach was suggested by the user after observing that
a manual edit to MCPTest2.project (made via the IDE) wasn't surfaced
by the mirror-only classifier when the change happened to be in a
non-mirrored region (likely device tree or library refs).
Verification pending: needs a vsc reboot to load the new server.js
into the running MCP process. Once reloaded, the next call against
v1.3.1.0 should:
- Read priorShas from the v1.3.1.0 tag (likely empty since this
is the first release with the new tag format).
- Treat empty priorShas as "no info, can't fall back" and behave
exactly like the pre-fix orchestrator. So nothing breaks.
- Write project-sha256 + mirror-sha256 into the v1.3.2.0+ tag bodies.
- From v1.3.2.0 onward, all three change-detection cases work.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Bumped status from "fixed (factory fallback)" to "fixed in c87f3a9
(4-iteration debug saga)" with the v1->v4 iteration table:
* v1 2607063: name= kwarg -- FAIL (SP22 wants foldername)
* v2 e07f281: positional + foldername= -- FAIL (silent None return)
* v3 32e6120: project-level + cache removal -- FAIL (still None=fail)
* v4 c87f3a9: walk children to detect side-effect success -- PASS
- Real root cause documented: SP22's create_folder returns void; the
fix verifies by walking parent.get_children(False).
- Side benefit from v3 (ScriptManager cache dropped) noted as the
hot-reload mechanism that made v4 reachable in one debug session.
- "Lesson for future fork work" callout: probably applies to other
create_* methods too, audit pending.
- Inventory table row for create_folder updated with the c87f3a9
link + a more accurate description of the strategy chain.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The v3 fix from 32e6120 still fell through every strategy because the
None-as-failure check was wrong. CODESYS scripting's create_folder
methods (both ScriptObject.create_folder and
ScriptProject.create_folder) RETURN VOID -- the folder is created via
side effect; the return value is Python None. v1/v2/v3 all treated
None as "this strategy didn't work" and tried the next, eventually
giving up.
v4 fix: after each create call, walk parent_object.get_children(False)
looking for a child whose get_name() == FOLDER_NAME. First strategy
that produces such a child wins; the rest are skipped (avoids
duplicates). The strategy order also got reshuffled: parent.create_folder
positional is now strategy 1 (per the SP22 ScriptObject stub
signature), with project-level fallbacks behind it.
Verified end-to-end on MCPTest2 + SP22 P1:
> create_folder(folderName='Test_Bench_Folder',
parentPath='PLCWinNT/Plc Logic/Application')
Folder 'Test_Bench_Folder' created [...]. Project saved.
> delete_object(objectPath='.../Test_Bench_Folder')
Object [...] deleted [...]. Project saved.
Per-version trace of the iteration:
v1 (2607063): name= kwarg -> "create_folder() got an unexpected
kwarg 'name'" -- SP22 stub uses
foldername.
v2 (e07f281): positional / foldername= -> silent None return on
Application; fell through.
v3 (32e6120): added primary_project.create_folder(name, SV_POU)
first; also dropped ScriptManager cache for hot-reload.
Still treated None as failure.
v4 (this): walk children after each call; succeeds.
Memory note for the kit: the SP22 scripting API has a class of
methods that mutate via side effect and return void. When porting
fork scripts, ALWAYS verify by walking children, never by checking
the return value of create_*. Same pattern probably applies to
create_pou / create_dut / create_gvl too -- worth a separate audit.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
create_folder v2 (positional foldername) returned None silently against
the SP22 Application object -- no exception raised, no folder created.
Investigation showed:
- ScriptObject.create_folder(foldername) is documented to "create a
folder in the structured view of the parent node", but on Application
specifically it's a silent no-op (the structured view isn't pinned to
POU view there).
- ScriptProject.create_folder(foldername, structured_view=None) on the
project itself with explicit SV_POU GUID
({21AF5390-2942-461a-BF89-951AAF6999F1}) is the documented and
reliable pathway -- the resulting folder appears under Application
in the IDE tree because that's where SV_POU lives.
v3 fix: try strategies in order until one returns non-None:
(1) primary_project.create_folder(name, SV_POU_GUID) -- new, primary
(2) parent.create_folder(name) positional -- pre-SP21 path
(3) parent.create_folder(foldername=name) -- alt keyword
(4) primary_project.create_folder(name) -- default view
(5) parent.create_object(typeUuid='85d1215e-...') -- alt factory
(6) parent.add(script_engine.types.IecFolder, name=name) -- legacy
Each strategy guards on hasattr + return-value-not-None, so a silent
no-op falls through instead of being mistaken for success.
ScriptManager: dropped the in-memory template cache. Each loadTemplate
call now reads the .py from disk fresh. Cost: ~1ms per call vs ~1.5s
of CODESYS execution time -- invisible. Win: edits to dist/scripts/
take effect without an MCP restart, which makes iterating on script-
side fixes (like this very create_folder loop) much faster. Existing
"cache hit" unit test rewritten as "two loads return equal content".
tests/test-fixes.mjs: standalone harness that drives a single persistent
CODESYS through HeadlessExecutor + CodesysLauncher to verify the four
broken-tool fixes end-to-end. Useful for regression-testing without
needing a vsc reboot loop. Currently only smoke-tests
create_folder + compile + cross-project; expand as more fixes need
verification.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The v1 fix from 2607063 used `parent.create_folder(name=FOLDER_NAME)`
which raised "create_folder() got an unexpected keyword argument 'name'"
on SP22 -- the actual stub signature in
C:\Program Files\CODESYS 3.5.22.10\CODESYS\ScriptLib\Stubs\scriptengine\ScriptObject.pyi
is
def create_folder(self, foldername): ...
The fallback chain (create_object + add(IecFolder)) caught the failure
and reported correctly, so no damage -- but the primary path was wrong.
Verified via the SP22 stubs that the keyword is `foldername`. Use
positional form first (agnostic to the keyword name across SP releases),
fall through to foldername= if positional fails for some reason, then
keep the existing alternate-factory chain as deeper fallbacks.
Tested partially in this session: create_folder DEBUG output confirmed
the v1 fix's fallback chain was running as designed; the keyword fix
will be runtime-verified on the next vsc reboot since script-manager
caches dist/scripts at MCP startup.
Verification trace (v1 against SP22):
DEBUG: Calling parent.create_folder(name='Test_Bench_Folder')
WARN: parent.create_folder() raised: create_folder() got an
unexpected keyword argument 'name'
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Headline result table at the top: persistent is 15-24x faster than
headless across all CODESYS-roundtripping tools. Numbers are from a
fresh bench-results.json run on MCPTest2.
- Updated mode primer with measured first-call vs warm-call costs:
persistent first launch ~14.6s, subsequent calls ~1.5s; headless
first cold call ~58s, warm calls ~22s.
- Inventory table: replaced "typical" estimates with measured numbers
for the 9 tools the bench covers.
- Deep-dive section: each broken tool now marked "fixed in 2607063"
with concise summary of the fix that landed (instead of "proposed
fix"). Removed the long pre-fix code blocks since they're in the
commit history now.
- Status legend: create_folder, compile_project, get_compile_messages,
connect_to_device, open_project all flipped from broken to fixed
(with caveats: some need PLC or runtime verification).
- Footnote about set_pou_code bench-harness failure (multi-line code
through triple-quoted-string interpolation -- not a tool bug).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
create_folder SP21+ fallback, ensure_project_open cross-project switch)
+ bench results
compile_project.py + get_compile_messages.py:
- IronPython 2.7's json.dumps cannot serialize System.Int64-backed `long`
values, which is what CODESYS's compile-message objects expose as
line_number / position. Added _coerce_int + _coerce_str helpers and a
shared _build_message_entry function. Three duplicated message-building
blocks collapsed into single helper calls.
- Defensive `try: json.dumps(...) except TypeError: json.dumps(default=str)`
so a stray field that slips past the helpers doesn't kill the emit.
connect_to_device.py:
- SP21+ may expose the login enum as LoginMode rather than
OnlineChangeOption. Extended the candidate sweep to probe both
script_engine.LoginMode and script_engine.OnlineChangeOption AND
online_app.LoginMode/OnlineChangeOption (some builds attach it to the
app object). Added "OnlineChange" + "Login" + "Download" to the
preferred-priority list. Added a 3-arg call shape variant for SPs
that take (mode, secondary-mode, force-bool).
create_folder.py:
- parent_object.create_folder() is no longer exposed on every parent type
in SP21+. Added two fallback factories tried in order:
1. parent.create_object(typeUuid='85d1215e-6520-4983-9a55-2d39d1f24cb4', name=...)
2. parent.add(script_engine.types.IecFolder, name=...)
with detailed warnings when each path fails. Final TypeError now lists
every factory tried so a future SP rotation surfaces clearly.
ensure_project_open.py:
- Uncommented the close-prior-project branch (was a TODO since the
initial fork). Cross-project switches in a persistent CODESYS now do
save() -> close() -> 500ms pump -> open(target). Without this,
projects.open against a different already-primary project fails
intermittently on file lock contention or pops a "project in use"
modal that freezes the IDE thread.
- save() is best-effort: if it raises (transient lock, save-as required)
we still proceed with close + open rather than getting stuck in a
half-switched state forever.
tests/bench-results.json:
- Captured timings from a clean run on MCPTest2 (PLCWinNT, 5 lib refs,
~12 POUs). 9 tools x 2 modes x iterations.
Headers (mean ms): persistent vs headless --
open_project 7700 vs 40041 (~5x; first call cold)
mirror_export 1547 vs 23723 (~15x)
list_project_libraries 1565 vs 23322 (~15x)
get_all_pou_code 1607 vs 23376 (~15x)
save_project 2095 vs 23321 (~11x)
create_pou (FB) 1540 vs 23903 (~16x)
delete_object 1544 vs 27420 (~18x)
bump_project_version 1540 vs 30678 (~20x)
bump_project_version #2 1556 vs 37769 (~24x)
- set_pou_code FAILED in both modes -- bench harness param-shape issue
(multi-line code passed verbatim to triple-quoted-string interpolation
doesn't survive the round-trip). Tool itself works fine through the
MCP tool call path; bench needs to escape newlines / use the same
prepareScriptWithHelpers shape the server uses. Filed for follow-up.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
TEST_OVERVIEW.md: complete tool inventory (37 tools), each tagged
working / broken with status notes and per-mode timing characteristics.
Mode primer up top explains the per-call vs first-call cost asymmetry.
Deep-dive on every broken tool with proposed fixes:
- create_folder: parent_object.create_folder() not exposed in SP21+;
fall back to create_object(typeUuid=...) or types.IecFolder.
- compile_project / get_compile_messages: IronPython 2.7 json.dumps
can't serialize System.Int64 (line_number / position fields).
Fix is a _coerce_int helper applied uniformly.
- connect_to_device: SP21+ may expose the login enum as LoginMode
instead of OnlineChangeOption. Extend the candidate sweep over
multiple enum sources, with priority on TryOnlineChange-equivalents.
- open_project (cross-project switch): ensure_project_open has the
"close prior project" branch commented out; uncomment with a
save+close+delay sequence and silent-mode guard.
list_project_libraries is flagged as ✅ working in current SP22
(historical entries in the project memory should be cleared).
bench.mjs: standalone benchmark harness driving HeadlessExecutor and
CodesysLauncher directly (no MCP server in the loop). Copies the source
.project to a temp dir so write tools don't mutate the original. Covers
9 tools (read-only + write-revertible) with configurable iterations,
emits markdown to stdout + JSON to --out.
Run with:
node tests/bench.mjs --modes headless,persistent --iterations 2
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Defensive guard against silent version regressions that the in-script
pi-vs-GVL cross-check can miss when CODESYS's in-memory project tree
is stale (both pi.version and the GVL read come from the same in-memory
source -- a stale tree gives consistent-but-wrong values that defeat
the cross-check).
Two observed regressions on MCPTest2 went undetected by the in-script
check and only failed at git-tag time:
- 2026-04-26 v1.0.4.0 (script saw 1.0.3.0, on-disk was 1.2.0.0)
- 2026-04-26 v1.1.0.0 (script saw 1.0.0.0, on-disk was 1.2.1.0)
This patch adds an orchestrator-side check after bump_project_version
parses its result: re-read the latest v* tag with `git describe`,
compare lexicographically (4-tuple int compare) against the new
version, and abort with a detailed recovery message if the new value
is not strictly greater. The abort happens BEFORE the post-bump
mirror_export, library.md/pou-dump.md regen, README rewrite, Changelog
append, and any git ops -- so no bad state gets published.
The .project binary on disk has still been mutated with the regressed
value at this point (bump_project_version saves at the end), but the
recovery path is well-understood: shutdown + relaunch + reopen to
clear the stale in-memory tree, then retry. Recovery instructions are
included in the error message.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Recurring "GVL misread" bug in release_project_version: when
Project Information.Version drifts BEHIND the runtime-anchor GVL
(_MCP_PROJECT_VERSION.sVersion), the bump used the stale pi.version
as the resume point and silently regressed the version, often
colliding with an existing v* tag.
Observed twice on the MCPTest2 sandbox:
1. v1.0.4.0 (2026-04-26): bump from on-disk 1.2.0.0 read pi.version
as 1.0.3.0 -> revision -> 1.0.4.0. Tag deleted; recovered as
v1.2.1.0 via manual finish script.
2. v1.1.0.0 collision (2026-04-26): bump from on-disk 1.2.1.0 read
pi.version as 1.0.0.0 -> minor -> 1.1.0.0. Tag already existed,
git tag step failed, release pipeline aborted. Recovered by
two manual minor bumps (1.1.0.0 -> 1.2.0.0 -> 1.3.0.0) and an
amended commit, released as v1.3.0.0.
Root cause: the MCPTest2 v1.2.0.0 and v1.2.1.0 releases were
finished by external (non-MCP) Node scripts that updated the GVL
via inject-once but never wrote pi.version back through the
bump_project_version pathway. So pi.version stayed pinned at
whatever value the LAST true bump_project_version run left it at
(in MCPTest2's case, ~1.0.0.0), while the GVL kept moving forward.
Fix: in the pi-present branch, read both pi.version and the GVL
sVersion, parse both as 4-tuples, and take the max as the resume
point. The max is always safe: both sides only ever move forward
in the normal case, so the higher of the two is by construction
the true latest version. When a drift is detected (pi behind GVL),
emit a WARNING and self-heal pi.version forward to the GVL value
before the bump so the warning doesn't recur on the next call.
The pi-missing branch is unchanged (still falls back to GVL).
Documented inline in the function with the regression scenario for
future maintainers. No new test (the affected logic runs inside
CODESYS's IronPython and doesn't have a unit-test scaffold here);
the inline comment + this commit message are the regression record.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two real bugs surfaced from the MCPTest2 v1.1.0.0 -> v1.2.0.0 round.
Fix 1: classifier didn't see untracked files (CLASSIFICATION GAP)
Adding FB_Position + FB_Random5s via create_pou + mirror_export
produced new untracked .st files in mcp-mirror/. The classifier
ran `git diff --name-status` which ONLY reports tracked changes;
the new files were invisible until git-added. Result: classifier
counted only 1 'modified' (PLC_PRG, the wiring update) instead
of 1 modified + 2 added, and resolved 'revision' instead of
'minor'. Added a `git ls-files --others --exclude-standard --
mcp-mirror/` pass that pulls untracked files and tags them as
adds (`added (untracked): <path>`). Now create_pou + release
correctly classifies as minor.
Fix 2: orchestrator didn't re-mirror after bump (DESYNC)
Pipeline was: mirror_export -> classify -> bump -> regen md ->
git commit. mirror_export ran BEFORE the bump, so the captured
mcp-mirror/_MCP_PROJECT_VERSION.st reflected the pre-bump GVL
value. Then bump updated the in-memory GVL + saved the .project
binary. Then commit went out with mcp-mirror at the OLD value
while the binary already had the NEW value. Two consequences:
- Next release call sees _MCP_PROJECT_VERSION.st as 'modified'
vs the just-pushed v* tag (because the next mirror_export
pulls the post-bump value, which now differs from the
still-pre-bump mirror in the tag), triggering another bump.
- The runtime-anchor on disk and the runtime-anchor inside the
running PLC binary were silently desynced from the docs.
Surfaced as MCPTest2 v1.1.0.0 (8d79193) shipping with binary
GVL = 1.0.2.0 while docs said 1.1.0.0; resolved by the v1.2.0.0
re-sync (1513e9c).
Added a SECOND mirror_export call right after the bump, before
regenerating library.md / pou-dump.md / Changelog. Soft-fails
with WARNING -- the bump itself already succeeded, post-bump
mirror is the documentation step.
Together these two fixes make release_project_version end-to-end
deterministic: 1 release call -> 1 release commit, no manual finish,
no re-bump on the next call. Verified offline: the path of the new
untracked-detection through ls-files --others, plus the second
mirror_export, give the orchestrator the post-bump state it
previously lacked.
The Python bump script emits 'Project Information.Version: <before>
-> <after>' usually, but when there's no Project Information node
(Standard-template projects -- created via create_project), the
output line becomes 'Project Information.Version: (skipped -- node
missing) -> <after>'. The previous regex used \S+ for the before-
group, which choked on the '(skipped' token (whitespace inside the
parenthetical broke the boundary).
Surfaced on MCPTest2 today during the FB_Position + FB_Random5s
release. release_project_version's bump succeeded (1.0.1.0 -> 1.1.0.0
via the GVL-resume path), but the orchestrator returned 'bump
succeeded but new version could not be parsed' because parseBumpedVersion
returned null and the post-bump pipeline (Changelog + library.md +
README + git ops) never ran.
Fix: non-greedy (.+?) capture for the from-group, treat
parenthesised values or 'none' as null. Also added a Runtime-anchor
fallback regex in case future Python changes alter the metadata
line shape -- the runtime-anchor line carries the same to-version
and is more stable.
REAL root cause of the phantom-release loop. The classifier fixes
in a570132 (--ignore-cr-at-eol -w) and the .gitattributes pinning
on the project repos addressed CRLF noise -- but the actual
trigger turned out to be deeper.
Every mirror_export run was emitting a line like:
(* Generated: 2026-04-26 01:25:14 *)
at the top of every .st file. Wall-clock time. So every run produced
byte-different output even when the project hadn't changed at all.
git diff (with or without whitespace flags) saw real, content-level
diffs. Classifier saw 92 modifications. Orchestrator bumped + tagged
+ pushed. Three phantom v1.0.1.0 releases on X33 in a row, each
reverted but each provoked by the next release call.
Fix: drop the Generated: timestamp line entirely. The git commit
history records when each file was committed; the in-file timestamp
was both redundant AND actively harmful (broke idempotency, defeated
the no-changes short-circuit in release_project_version).
Side effect: the next mirror_export against an existing project
will produce an N-file 'remove that timestamp line' commit -- a
one-time schema migration. Subsequent runs are then idempotent.
This bug took three rounds of recursive fork fixes to track down:
9fb569b orphan-shutdown -- unrelated, real fix
a570132 classifier --ignore-cr-at-eol -w -- defensive but
not the root cause
THIS drop the timestamp -- root cause
Lesson: when a no-op call has side effects, look for content
non-determinism in the no-op itself before blaming the classifier.
Real bug surfaced after a VSC restart left a CODESYS.exe alive that
the new MCP server didn't spawn (state=stopped/error, this.pid=null).
The previous shutdown() early-returned at the top -- the launcher
was 'stopped' so it considered itself done -- which:
1. Left the orphan CODESYS.exe alive (couldn't run the project).
2. Made the refuse-on-duplicate guard (95a884b) block every
subsequent launch_codesys with 'CODESYS already running'.
The launcher was effectively bricked: shutdown said success but
did nothing, launch refused. Hit during today's release-pipeline
test loop.
Fix: when shutdown() is called with state=stopped/error AND this.pid
is null AND findRunningCodesysPids() returns non-empty, taskkill the
orphans before the early-return. Graceful WM_CLOSE first, then 2s
grace, then -F force-kill anything still alive.
Doesn't change the happy-path (launcher tracks its own PID, state
ready -> stopping -> stopped) -- that flow is untouched. The new
code only runs when the launcher would otherwise have been a no-op
on something the OS still has active.
Real bug surfaced on X33 (commit 6c23e38 on karstein.kvistad/x33,
reverted in 3e6f12f): the classifier called git diff --name-status
without any whitespace flags, so a fresh checkout that re-normalised
.st files from LF to CRLF (Windows working copy via Samba share)
showed every file as M. The orchestrator obediently bumped the
project to v1.0.1.0 with no actual code change, committed,
tagged, pushed -- a phantom release.
Fix: add --ignore-cr-at-eol AND -w to the git diff invocation so
the classifier only reports diffs with real content changes.
--ignore-cr-at-eol ignore the carriage-return at the end of line
when comparing lines (handles CRLF<->LF flips)
-w ignore whitespace differences entirely
(defensive; protects against stray blank
lines and indent normalisation that aren't
real changes)
The companion fix is to also add a .gitattributes to each project
that pins the .st files to a stable line-ending in storage so the
phantom diffs don't appear in the first place. That's a per-project
artefact, shipped alongside the project repos (X33 + MCPTest2)
rather than this fork.
Two changes bundled:
1. Adds release_project_version, the one-shot orchestrator that runs
the whole sync from a CODESYS code change to a tagged + pushed
git commit. Sequence:
mirror_export refresh mcp-mirror/
classifier diff vs latest v* tag
-- if no changes short-circuit, no commit
bump_project_version resolved-level bump
Changelog.md append entry with classification
list_project_libraries regen library.md as markdown
get_all_pou_code regen pou-dump.md as markdown
README.md regex-replace v<old> -> v<new>
git add mcp-mirror, .md files, .gitignore,
.project binary
git commit "release v<new> (label)"
git tag v<new> -a + message
git push --follow-tags (configurable via push arg)
This is the standard the project README points at: "ask Claude to
run release_project_version after every confirmed change in
CODESYS." All four sources of truth (Project Information.Version,
_MCP_PROJECT_VERSION.sVersion, Changelog.md, v* git tag) move
together in one call.
Markdown rendering helpers (renderLibraryMd, renderPouDumpMd,
gfmSlug) are extracted to module level so the orchestrator can
call them directly. The list_project_libraries tool's response
formatting still uses inline rendering since it returns plain
text; markdown is for the on-disk artefact only.
2. Changelog entry timestamp now includes HH:MM in local time
(YYYY-MM-DD HH:MM, no seconds, no TZ suffix). User feedback:
date alone was too coarse to distinguish multiple bumps in the
same day. Format chosen for compact heading + sort-friendly +
no timezone-conversion friction. appendChangelogEntry handles
both auto-mode (called from bump_project_version --auto) and
manual-mode (called from release_project_version directly).
Previously, when a project had no Project Information node (e.g. one
created from the Standard template via create_project), the bump
flow read pi.version as None, the seed-check fired, and every call
re-seeded to 1.0.0.0 -- subsequent revision/minor/major bumps were
no-ops because the script never saw the actual current version.
Surfaced on MCPTest2 today: bumping revision after editing PLC_PRG
returned '1.0.0.0' instead of '1.0.1.0' because the seed kept firing.
Fix: when Project Information is missing, fall back to reading the
existing _MCP_PROJECT_VERSION.sVersion via the textual_declaration
of the GVL we ourselves maintain. So the source-of-truth chain is:
pi.version (when Project Information exists)
-> falls back to GVL.sVersion (when Project Information missing
but the GVL has been written
by a prior bump)
-> falls back to seed at 1.0.0.0 (true first-run, no GVL yet)
Implementation: read_version_from_gvl(primary_project) walks the
active Application's children for the named GVL and parses the
sVersion := '...' literal out of its textual_declaration with a
4-part regex. Returns None if the GVL doesn't exist OR its decl
doesn't match the expected shape; caller treats None as "no prior
version, seed". Soft-fails on any access exception (the bump is
the primary outcome, this is just resume-from-state).
This is the kind of "every arising problem fixed at the fork, not
worked around in one-offs" hygiene the user called out.
Standard-template projects (those created via create_project) often
don't have a Project Information node at all -- it's added lazily by
the IDE when the user opens Project menu -> Project Information for
the first time. Surfaced when running auto-bump against MCPTest2
(copied from MCPTest, which was created via create_project).
Two changes:
1. Use the documented is_project_info marker (ScriptProjectInfoMarker
per the SP22 stub Stubs/scriptengine/ScriptObject.pyi) to find the
node, instead of name-matching 'Project Information'. Also robust
against localised IDE display names ('Projektinformation' in DE).
Walks up to depth 4 from the root.
2. If still no node found, log a WARNING and SKIP the metadata write
(Project Information.Version), but continue with the GVL
maintenance. The GVL is the runtime source-of-truth anyway -- the
running PLC reads _MCP_PROJECT_VERSION.sVersion, not the .project
metadata. Subsequent bumps after the user adds Project Information
manually (Project menu -> Project Information in the IDE) will
pick up both sides.
Output line is also adjusted -- 'Project Information.Version: (skipped
-- node missing) -> 1.0.0.0' instead of pretending to have updated
something that doesn't exist.
After every successful bump (auto or manual), append an entry to
<projectDir>/Changelog.md describing the change. Newest entries at
the top under a one-time intro header. Each entry carries:
## v<X.Y.Z.W> -- YYYY-MM-DD (<label>) [(from `<prev>`)]
- <evidence bullet 1>
- <evidence bullet 2>
...
Where:
<label> is one of:
seed -- first-run, no prior version
auto: <level> -- classifier resolved to <level>
manual: <level> -- explicit level passed
<evidence> is the same classification list shown in the tool
response (D/R/A/M file paths from the mcp-mirror/ diff against
the latest v* tag); for manual bumps it's empty + a "(no
classification evidence -- manual bump)" placeholder line.
Soft-fail: any I/O error during Changelog write logs a warning but
does not fail the bump itself -- the Project Information.Version
update has already succeeded by the time this runs, and the
Changelog is documentation, not state-of-truth.
Format choices:
- File path: <projectDir>/Changelog.md (alongside README.md /
library.md / pou-dump.md, mirroring the existing convention).
- Heading style: H2 ## per version, H1 # for the file title only.
Matches GitLab GFM auto-anchor expectations.
- Date: ISO YYYY-MM-DD in UTC, so collaborators in different
timezones see the same date for the same bump.
- Insertion: before the first existing ## v<...> heading, after
the intro. So a chronological reader sees newest first.
- Versions explicitly cross-referenced to the runtime anchor
(_MCP_PROJECT_VERSION.sVersion) in the intro so the reader knows
a Changelog entry == a value the running PLC will report back
via read_running_version_online.
Now rounds out the version-tracking convention end-to-end:
Project Information.Version (offline metadata)
_MCP_PROJECT_VERSION.sVersion (runtime anchor in IEC code)
Changelog.md (human-readable history)
v* git tag (machine-readable history /
classifier baseline for
subsequent auto-bumps)
All four move together on every bump.
New MCP tool that reads the running project's version from a
connected PLC over the CODESYS online protocol (port 11740 / gateway).
Returns the value of `_MCP_PROJECT_VERSION.sVersion` -- the runtime
anchor that bump_project_version maintains automatically (commit
00d2dd8). Closes the loop on the version-tracking convention:
bump_project_version writes _MCP_PROJECT_VERSION.sVersion
into the project (compiled into the
boot application on next download)
read_running_version_online reads the same symbol back from the
running PLC over the online protocol
Pairs with the existing connect_to_device + read_variable pattern
but with sharpened error messages tailored to the version-read use
case:
- missing GVL -> 'has bump_project_version run on
this project? Or has the boot
application not been downloaded
since the bump?'
- read returns None -> 'variable exists in project but
not in boot app -- download_to_device
after the last bump.'
- missing online API -> typed clearly, suggests SP-version
drift.
Implementation: ensure_project_open + ensure_online_connection +
online_app.read_value('_MCP_PROJECT_VERSION.sVersion'); strips quote
characters from the returned STRING; sanity-checks the shape against
\b\d+\.\d+\.\d+\.\d+\b and warns if it doesn't match the
4-part convention. TS handler extracts the matched RUNNING_VERSION
line from the script output and surfaces it as the headline of the
tool response.
SSH transport variant (read_running_version_ssh) lands as a separate
later commit once a real PFC is reachable to test against.
Establishes the runtime-readable version anchor convention. Every
bump (manual or auto) now ALSO ensures the Application has a GVL
named '_MCP_PROJECT_VERSION' with:
{attribute 'qualified_only'}
VAR_GLOBAL CONSTANT
sVersion : STRING := '<X.Y.Z.W>';
END_VAR
Created on first bump; updated in place thereafter. Soft-fails if
the Application object can't be found or create_gvl() raises -- the
primary outcome (Project Information.Version updated and saved) has
already happened by the time GVL maintenance runs, so a GVL hiccup
is logged as a WARNING but doesn't fail the whole tool.
Why this matters:
- Project Information.Version is metadata. The running PLC binary
embeds it but exposing it at runtime requires the auto-generated
Project_Info library helpers (GetVersion etc.), which not every
project has wired up.
- A plain VAR_GLOBAL CONSTANT in a known-name GVL is the simplest,
most portable runtime anchor. Any IEC code can read it as
`_MCP_PROJECT_VERSION.sVersion`. The future read_running_version_online
tool will pull it via online connect + read_variable. The future
SSH transport variant can pull it via libcmd-symbol-export or by
grepping a debug log line that the project author can wire to
write at startup.
qualified_only is set so the symbol can't accidentally shadow a
same-named local in user code.
The GVL convention will be exercised end-to-end on MCPTest running
on the local soft PLC (port 11740) once the read_running_version_online
tool ships -- that's the next ship in this sequence.
Previously, level=auto with no diff against the latest v* tag still
called the Python bump with level='build', incrementing 1.0.0.0 to
1.0.0.1 even though nothing in mcp-mirror/ had changed since the
baseline. That's wrong -- 'no changes' should mean 'no bump'.
Refactored the classifier to return a tagged ClassifyResult:
kind: 'no-changes' -> short-circuit, no Python call, return a
'no version change' message with evidence.
kind: 'first-run' -> no v* tag yet (or not a git repo); call
Python with level='build' which triggers
the seed-at-1.0.0.0 path when Version is
unset.
kind: 'bump' -> resolved level + evidence; call Python.
Test against X33 right now: latest tag is v1.0.0.0, mcp-mirror/ has
no changes against that tag, so auto would correctly return 'no
version change' without bumping.
Replaces the manual 'pick the right level' workflow with automatic
classification driven by git-diff over the project's mcp-mirror/
folder. Today's standard is the only one we care about (no project
has version-tracking hooked up before this fork shipped it), so the
classifier looks at exactly the artefacts the MCP itself writes.
Classifier rules (file-granularity in v1):
any D (delete) or R (rename) -> major (public symbol gone)
any A (add) -> minor (new public symbol)
any M (modify) -> revision (internal change)
no changes / no v* tag -> build (also triggers the
Python-side seed-at-
1.0.0.0 first-run path
when Version is unset)
When level=auto:
1. Resolve the project's parent directory.
2. Verify it's a git repo (fall back to 'build' if not).
3. Find the latest v* tag via `git describe --tags --abbrev=0
--match "v*"`.
4. `git diff --name-status -M50% <tag> -- mcp-mirror/` and tally
D/R/A/M counts.
5. Resolve to one of major/minor/revision/build per the rules above.
6. Pass the resolved level to the existing Python bump script.
The classification evidence (each D/R/A/M file path) is included in
the tool response so the user can audit the decision -- 'why did this
bump revision and not minor?' has a one-line answer.
Future iterations: split each modified .st file at its
`(* === IMPLEMENTATION === *)` separator and distinguish decl-only
changes (minor / major) from impl-only changes (revision); also wrap
this into a release_project_version orchestrator that re-runs
mirror_export, regenerates library.md, updates the README header,
and tags + pushes the resulting commit. Out of scope for this commit.
Enriches the tool's output with project-level metadata that previously
lived only in hand-edited library.md headers:
Project info:
Version: 1.0.0.0 (Project Information.version)
Title: ... (Project Information.title)
Company: ... (Project Information.company)
Author: ... (Project Information.author)
IDE: CODESYS V3.5 SP22 Patch 1, ScriptEngine.plugin 4.2.0.0
Devices (N):
MainPLC [4096 / 1006 120D / 6.2.0.1]
MainPLC/Kbus [32778 / Wago 750-Series Local Bus Interface / 2.1.0.1]
...
Implementation:
- collect_project_info() reads .version / .title / .company / .author
on the Project Information node (first child of project root). Each
field is read defensively (try/except) since some installs leave
them unset; missing fields are dropped from the output.
- collect_devices() walks the tree depth-first for nodes where
is_device is True, captures get_device_identification() into a
type/id/version triple. The triple is the offline target id the
IDE uses to pick a compiler + runtime when building -- not the
live firmware reported by a connected PLC over a runtime
connection (the latter would require an online connect).
- sys.version inside IronPython under CODESYS reports the IDE
version directly (same string we see in ready.signal).
server.ts renders these as a Header block above the existing
library-by-container tables. Hand-edited X33/library.md "Versions"
section is now redundant -- next regeneration will produce the
header automatically.
Verified via the local SP22 install + the live X33 watcher: pi.version
read-back works after bump_project_version sets it, devices walk
returns 13 entries on X33 (MainPLC + 11 Kbus modules + the network
adapter).
Per user feedback. Previous behaviour treated 'no version yet' as
0.0.0.0 and bumped from there, so the very first call with level=build
produced 0.0.0.1 -- awkward as a canonical starting point. Most
projects start tracking at 1.0.0.0 the moment they turn on versioning.
New behaviour: if Project Information.version is None / empty / '0.0.0.0',
the tool seeds the value at 1.0.0.0 directly and ignores the level
argument for that one call. Subsequent calls bump per the level as
before.
Test cases:
None + level=build -> 1.0.0.0 (seed)
None + level=major -> 1.0.0.0 (seed)
'' + level=minor -> 1.0.0.0 (seed)
0.0.0.0 + level=revision -> 1.0.0.0 (seed)
1.0.0.0 + level=build -> 1.0.0.1
1.0.0.0 + level=minor -> 1.1.0.0
1.2.3.4 + level=major -> 2.0.0.0
New MCP tool that increments one part of the 4-part
Project Information.Version field of the primary project, saves the
project, and reports the before/after.
Behaviour:
- level=major -> bump major, reset minor/revision/build to 0
- level=minor -> bump minor, reset revision/build to 0
- level=revision-> bump revision, reset build to 0
- level=build -> bump build only
Convention follows the rest of CODESYS / 3S / WAGO library practice
(visible in any X33 library reference like 'WagoAppCanLayer2,
1.6.1.4 (WAGO)'):
Major -- incompatible API break (rename FB, change public
signature, remove method).
Minor -- backward-compatible feature add.
Revision -- bug fix only, no API change.
Build -- internal / CI counter, often 0 for hand-released.
Implementation notes:
- Project Information lives as the first child node of the project
root. Its .version property is read/written directly; IronPython
coerces strings like '1.2.3.4' to a System.Version on assignment,
str(System.Version) gives the dotted form back. None / empty /
unset is treated as '0.0.0.0'.
- Verified live against X33 (MRCodesysX33_0021): set version to
'1.0.0.0' from None, project.save() persisted it; reload via
primary_project.get_children() found the same value. Probe done
via the inject-once.mjs bridge against the live watcher in PID
23056 before this commit landed.
- project.save() is called after the bump so the new value sticks
in the .project file. Soft-fails on save error (visible WARNING
in DEBUG output but the bump itself is still reported as
successful) so a save permission glitch doesn't mask the actual
version change.
Used by the X33 GitLab project's "version in README header" workflow:
the version surfaces at the top of README.md (and at the top of the
library list once list_project_libraries gets enriched in a follow-up
commit).
Two parallel smoke-test files had drifted out of sync:
docs/SMOKE-TEST-2026-04-25.md (morning v1, 28 invocations,
pre device-side fix)
docs/SMOKE-TEST-2026-04-25-v2.md (afternoon v2, 29 invocations,
post device-side fix)
Neither covered the late-afternoon work (six new git_* tools,
mirror_export, launcher refuse-on-duplicate, list_project_libraries
rewrite). Consolidating to one doc named SMOKE-TEST-2026-04-25.md
that:
- Covers the full 36 distinct tool invocations (28 OK / 4 FAIL /
1 PARTIAL).
- Reduces the open-bugs list from 5 to 4 (list_project_libraries
is fixed; the read/write inconsistency now narrows to
add_library's missing dedupe + placeholder).
- Adds rows for the 7 new tools verified end-to-end on X33 (8
library refs, 91 .st mirror files) and on GitSmokeTest (full
init -> commit -> remote_add -> branch_set_upstream_to -> push
round-trip against gitlab.usv.no via cached HTTPS creds).
- Documents the new infrastructure (PDE-license rewrite, launcher
refuse-on-duplicate) as separate sections rather than bugs.
- Cites the late-afternoon commits in the "What this proves" table
so the smoke test traces every fix back to its source.
The morning v1 is removed; the afternoon v2 file is renamed onto the
canonical filename to preserve git's rename history.
Drops the redundant MCP/ folder layer. Previous default put the
mirror at <projectDir>/MCP/mirror -- two levels deep with a folder
called "MCP" containing exactly one item called "mirror". After
real-world usage (X33 layout reorg this session) the cleaner default
is <projectDir>/mcp-mirror -- one level, descriptive name.
Existing X33 layout was already migrated:
X33/MCP/mirror/ -> X33/mcp-mirror/ (renamed)
X33/MCP/library.md -> X33/library.md (moved up)
X33/MCP/pou-dump.md -> X33/pou-dump.md (moved up)
Tool description and the mirrorRoot arg description updated to match.
Match the de-facto community convention. .st (IEC 61131-3 Structured
Text) is what the existing PLC tooling around forge.codesys.com,
ArthurkaX/cds-text-sync, and the VS Code "IEC Structured Text"
extension already syntax-highlight by default. The previous .iecst
was specific but unhelpful: nothing else in the toolchain knew what
to do with it.
User asked for the switch right after Phase 1 landed (commit 7a6e725).
Re-ran the export against MRCodesysX33_0021 (X33) on disk: 91 files
re-emitted as .st, identical content + tree shape, total 254 KB.
Phase 1 of the "CODESYS project as a filesystem you can edit + diff +
ai-tooling against" idea. Today the only way to read code in a project
is via a one-shot get_all_pou_code dump (one giant JSON) or by clicking
through the IDE; neither is friendly to AI-assisted edits, code review,
git diffs or external tooling. mirror_export walks the live project
tree and emits one .iecst file per code-bearing object, preserving the
project's folder structure as nested directories on disk.
What the tool does:
- Walks every node from script_engine.projects.primary.get_children()
recursively (depth-first).
- Structural nodes (Device, Application, Folder, ...) become
directories under MIRROR_ROOT.
- Code-bearing nodes (Program, FB, Function, Method, Property, DUT,
GVL, Interface, ...) become <name>.iecst files in their parent
directory; if a code-bearing node has child code objects (e.g. an
FB with methods), those children land in a sibling subdirectory
with the parent's name.
- File header: `(* === CODESYS export -- KIND === *)` + project path
+ generated timestamp, so a future write-back tool can map each
file back to set_pou_code's pouPath.
- Body: declaration block, then `(* === IMPLEMENTATION === *)`
separator (when both are present), then implementation block.
Defaults the mirror root to `<projectDir>/MCP/mirror` so it lands next
to the existing library.md / pou-dump.md if the user has been
following the same folder convention.
Implementation details that would have bitten without testing:
- UTF-8 output via codecs.open. CODESYS POU text occasionally
contains non-ASCII (smart quotes from copy-paste, degree signs,
etc.). IronPython 2.7's builtin open() defaults to ASCII and
would raise; saw it on X33's ST_HetronicIn2 (smart quote) until
fixed.
- Filesystem-illegal characters in CODESYS object names (`/`, `\`,
`<>:"|?*`) replaced with `_`. CODESYS lets you name a folder
"Remote / Hetronic"; on Windows that splits as two folders with
naive os.path.join.
- Kind classifier strips leading `//` and `(* *)` comments and
`{attribute := '...'}` pragmas before matching the IEC keyword,
otherwise GVLs decorated with attribute pragmas got bucketed as
OTHER (X33 had 2 of these; first try misclassified them).
Verified against MRCodesysX33_0021 (X33): 91 files, 254 KB, 0 errors,
preserves the full tree (MainPLC/Plc Logic/Application/MRLib/...).
Phase 2 (write-back via a `sync_pou_from_file` tool that reads a
mirror file, splits decl/impl on the IMPLEMENTATION separator, and
calls set_pou_code) is out of scope for this commit.
RTFM. Per the helpme-codesys.com Library Manager scripting page and
the local SP22 stub Stubs/scriptengine/ScriptLibManObject.pyi:
- ScriptLibManObjectContainer is added to BOTH the project AND every
Application object. It exposes:
has_library_manager -- @property (NOT a method) returning bool
get_library_manager() -- method returning the LibMan ScriptObject
- ScriptLibManObject (the libman itself) exposes:
.references -- @property, ScriptLibraryReferences (list-
like) of ScriptLibraryReference objects
with structured fields (name, namespace,
is_placeholder, is_managed, system_library,
effective_resolution, ...)
get_libraries(recursive=False) -- list[str], names only
- ScriptLibManObjectMarker.is_libman is the universal marker, useful
as a fallback when walking the tree.
The previous implementation searched the project tree by NAME for an
object literally called "Library Manager" via primary_project.find()
and a children name probe. That never matches because the libman's
actual name is generated, not "Library Manager." On the X33 project
(MRCodesysX33_0021) it returned empty -- false negative on a project
that obviously has dozens of system + application libraries. This
also took add_library down with it (same wrong axis), and the smoke
test from earlier in the fork's history flagged the inconsistency
without identifying the cause.
Fixed:
- Walk the tree depth-first, accept any node where has_library_manager
is True (depth-limited at 8 to be safe).
- For each, call get_library_manager() and iterate .references for
structured data; fall back to get_libraries() name-only enumeration
if .references is unavailable on the SP.
- Capture every documented ScriptLibraryReference field defensively
(each access wrapped in try/except since some fields raise on
placeholders / unmanaged / SP-version skew).
- Return a structured JSON shape grouped by container so the TS side
can show which Application owns which libraries.
server.ts:
- Updated the result-parsing block to handle the new structured shape.
- Distinguishes "no library managers found" (suspicious, libman
discovery probably broken) from "found managers, all empty" (just
empty applications).
- Renders flags ([system, placeholder, managed, optional, redirected])
+ namespace + effective_resolution per reference.
- Tool description rewritten to advertise the actual mechanism and
cite the doc + local stub source.
add_library is NOT fixed in this commit -- it has the same wrong-axis
bug but landing the lookup-only fix first to verify the API contract.
add_library will land separately once we know list_project_libraries
sees the right libman in production (X33 smoke test).
The launcher only knew about its own state machine -- so any time a
prior CODESYS.exe was still alive (orphan from a crashed MCP session,
the user's own interactive IDE, or a CODESYS still mid-shutdown),
calling launch() happily spawned a SECOND CODESYS. The two instances
then raced on the project file lock and the loser surfaced the
"project is currently in use by <user> on <host>" modal, which blocks
the IDE thread and freezes every subsequent script call -- the agent
keeps timing out at 60s with no useful diagnostic.
Hit twice in a single session on 2026-04-25 during git_* smoke tests:
the user pointed at the running taskbar twice ("you are opening
shitloads of codesys sessions" / "you keep opening TWO codesyses --
then this message comes up and halts all your progress").
Adds findRunningCodesysPids() -- a tasklist-based scan -- and a
pre-spawn guard at the top of launch() that returns a clear,
action-oriented error listing the offending PIDs and the two
remediations (close the IDE manually, or call shutdown_codesys if the
launcher owns the existing process). No-op on non-Windows.
Doesn't try to adopt the existing process: that would require sharing
its IPC dir + watcher state, which the launcher cannot recover after
the fact. Refusing the spawn is the only safe response.
Second RTFM remediation. The helpme-codesys.com Git scripting page
(https://content.helpme-codesys.com/en/CODESYS%20Git/_git_using_scripting.html)
says explicitly:
"Project Save: Call project.save() after operations to persist
changes. Appears after operations like push() and merge() to
persist changes."
The example flows on that page show project.save() interleaved with
init / commit / push / merge. None of the git_* wrappers shipped so
far called save(), which means binding info, configured remotes,
upstream tracking, and post-commit state could fail to persist when
the IDE closes -- silently degrading every flow that spans more than
one CODESYS session.
This commit adds a soft-fail primary_project.save() after every
mutating op:
git_init after git.init(...)
git_commit after git.commit_complete(...)
git_remote_add after git.remote_add(...)
git_branch_set_upstream_to after git.branch_set_upstream_to(...)
git_push after git.push(...)
git_status is unchanged -- it's read-only.
Soft-fail rationale: a save() failure does NOT undo a successful git
op. We log a WARNING and continue, so the visible result still
reflects what actually happened in the repo. Bubbling save errors
would risk telling the user "init failed" when in fact the .git/ is
on disk and the only loss is the in-memory binding. The honest
failure mode is "git op succeeded, persistence may not have."
RTFM. The helpme-codesys.com Git scripting page
(https://content.helpme-codesys.com/en/CODESYS%20Git/_git_using_scripting.html)
spells out the canonical "init -> commit -> remote_add -> push"
sequence and explicitly inserts a mandatory step between remote_add
and push:
"Set Upstream Before Push: After adding a remote, use
project.git.branch_set_upstream_to(origin_remote) before pushing."
The previous commit (8a6059b) shipped git_remote_add + git_push but
left this step labelled "out of scope," which guaranteed the very
first end-to-end smoke test would fail with:
sLocalBranchName: The branch 'master' ('refs/heads/master')
does not track an upstream branch.
Confirmed against the GitSmokeTest project (C:\Temp\GitSmokeTest)
on 2026-04-25 immediately after the commit landed.
This commit adds git_branch_set_upstream_to as a thin wrapper over
project.git.branch_set_upstream_to(remoteName, branchName?) using
the simplest of the four overloads in
Stubs/scriptengine/GitScriptProject.pyi. branchName defaults to the
current branch (per the stub default arg). Defensive checks +
HasGitLicense rewrite mirror the rest of the git_* tool family.
Tool description carries the doc citation up front and the exact
error message you get without it, so the next user sees the missing
step before having to discover it experimentally.
Two new MCP tools wrapping the remaining pieces needed for an
end-to-end "init -> commit -> push to GitLab" flow:
git_remote_add Wraps project.git.remote_add(name, url). Required:
remoteName + remoteUrl. Conventionally name='origin'
for the primary upstream.
git_push Wraps project.git.push(...). Three overloads handled:
- push() when no branch + no creds (relies on tracked
upstream + git config / Windows Credential Manager)
- push(branch) when only branch is provided
- push(branch, user, SecureString(token)) when both
credentials are provided; derives current branch via
branch_show_current() if branchName is omitted.
Credentials handling: when a token is supplied it is converted to
System.Security.SecureString before being handed to push(), per the
CODESYS Git scripting docs guidance ("Use SecureString passwords
whenever possible"). Tool description carries an honest security note
that the token is briefly templated into the IronPython source that the
watcher executes -- prefer cached credentials when feasible. server.ts
neutralises backslashes/double-quotes and rejects newlines in the
templated values up front.
Both tools share the same defensive checks as the existing git_* trio:
- hasattr(script_engine, 'git') / project.git is None handling so the
user gets a "run git_init first" hint when the project is unbound.
- attribute probe + dir() dump if the API surface is missing the
expected method.
- HasGitLicense detection in the catch-all that rewrites SCRIPT_ERROR
into the clear "PDE subscription required" message.
API contract verified against:
- Stubs/scriptengine/GitScriptProject.pyi (local SP22 install) -- the
remote_add and push overload set.
- helpme-codesys.com Git scripting page
(https://content.helpme-codesys.com/en/CODESYS%20Git/_git_using_scripting.html)
-- SecureString recommendation for password params.
Out of scope: pull, fetch, branch_set_upstream_to, push_delete,
clone (top-level script_engine.git, not project-bound). Can land
separately.
CODESYS Git uses a dual-storage model: the .project file stays where
it is, the git working tree lives in a SEPARATE empty directory. The
prior default ("init in the project's own folder") tripped over the
project file itself and CODESYS would raise the unhelpful
'gitProjectStoragePath: ... is not a valid Git repository location:
DirectoryNotEmpty'. Hit this end-to-end during the first PDE-Demo
smoke test on 2026-04-25 -- the user had to dance around it manually.
git_init now:
- Defaults LocalRepoPath to '<project_basename>_git' as a sibling of
the project's own dir when not supplied (or when supplied equal to
the project dir, which has the same trap).
- Auto-creates the target dir if missing.
- Pre-validates emptiness and raises a clear, action-oriented error
if non-empty, instead of letting CODESYS surface DirectoryNotEmpty.
server.ts tool description updated to call out the dual-storage rule
and the auto-default behaviour up front.
Verified API contract via Stubs/scriptengine/GitScriptProject.pyi
(local SP22 install) and the helpme-codesys.com Git scripting page
(https://content.helpme-codesys.com/en/CODESYS%20Git/_git_using_scripting.html).
The 3 git_* tools previously surfaced the raw CODESYS exception
"Permission denied: Rule 'HasGitLicense' failed with state 'False'."
which is opaque to anyone who does not already know that CODESYS Git
scripting is gated behind the Professional Developer Edition
subscription. Verified 2026-04-25 on this workstation that the
plug-in is installed (PlugIns/GitIntegration.plugin.dll v1.7.0.0,
ScriptDriverGit.plugin.dll, full ScriptLib/Stubs/scriptengine/Git*.pyi
set, and the Git menu visible in the IDE) but the runtime rule still
returns False because no PDE subscription is active. Per the CODESYS
Git store page the subscription is the actual gate (Additional
Requirements / Licensing).
Each script now detects 'HasGitLicense' in the exception or traceback
and rewrites SCRIPT_ERROR to a clean, actionable message pointing at
https://store.codesys.com/en/codesys-git.html with a "what to activate"
hint, falling back to the original generic error formatting on any
non-license failure.
git_status: also adds an early license probe via project.git.has_working_tree()
(license-gated per the SP22 stubs) so the rewrite triggers reliably
instead of being swallowed by the per-method probe loop further down.
Tool descriptions in server.ts updated to advertise the PDE
subscription requirement up front, so MCP clients see it without
having to fail first.
Cross-references the official Git scripting docs at
https://content.helpme-codesys.com/en/CODESYS%20Git/_git_using_scripting.html
which exposes a project-bound git API as 'primary_project.git' (when the
CODESYS Git plug-in is loaded). Confirmed earlier today via dir(scriptengine)
that 'git' is one of the top-level scriptengine modules on this install.
Three tools in this commit, ordered shallow-to-deep:
git_status Read-only. Returns current branch via
project.git.branch_show_current() AND a defensive probe
of any status/changes/diff/changed_files methods on
project.git (the docs page lists the call patterns by
example but does not enumerate the full API surface, so
the probe + diagnostic dump is how we'll discover the
rest in the next iteration).
git_init Wraps project.git.init(local_repo_path). Defaults the
repo path to the project file's parent directory so a
plain 'init this project's folder' call needs no extra
args. One-shot setup; pair with git_status afterwards.
git_commit Wraps project.git.commit_complete(message, user, mail).
Required: message + authorName + authorEmail (latter
validated as email by zod). Stages all working-tree
changes and commits in one shot per the docs. Multi-line
messages handled via triple-quoted Python injection with
standard backslash + triple-quote escaping.
Defensive checks across all three:
- hasattr(script_engine, 'git') so we can distinguish "Git plug-in not
installed" from "project not in a repo".
- project.git is None handled with a clear 'run git_init first' message.
- On missing methods the script dumps sorted(dir(project.git)) so the
next debug session sees the exact surface.
Out of scope for this commit (deliberate -- one feature per commit per
the project rule, more git ops can land separately):
- branch ops (branch_copy, checkout)
- remote ops (remote_add, push, pull, fetch, branch_set_upstream_to)
- merge
- clone (script_engine.git.clone -- top-level, not project-bound)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per the project rule (check helpme-codesys.com first / cross-reference
before any scripting fix), each of the 5 open bugs from
SMOKE-TEST-2026-04-25-v2.md gets:
- Empirical failure as observed today on SP22 Patch 1
- Relevant docs URL on content.helpme-codesys.com
- What the docs confirm (or fail to surface -- many pages are index-
only with method bodies rendered by JS, not WebFetch'able)
- Proposed fix path with rationale and approximate effort
Findings summarised:
1. create_folder kwarg mismatch -> try positional 'create_folder("name")'
(~10 LOC, high-probability fix)
2. compile_project / get_compile_messages JSON long ->
pre-process the dict to coerce IronPython long to int/str before
json.dumps; belt-and-suspenders default=str (~15 LOC, single helper)
3. list_project_libraries returns empty ->
use ScriptLibManObject.get_libraries() (canonical), reuse the
same Library Manager discovery pattern that add_library already
succeeds with (~30 LOC)
4. add_library duplicates / not placeholder ->
pre-check via get_libraries() to dedupe; default to add_placeholder()
to match Standard template convention; opt-in 'direct=true' for
the old behaviour (~40 LOC, mildly breaking)
5. rename_object partial refactor ->
no documented refactor variant in scriptengine; brute-force walk
of every POU + word-boundary regex replace, gated behind
updateReferences=true flag (~80 LOC, biggest fix)
Order-of-attack ranking included. Each fix is its own commit per the
project rule, with the relevant docs URL cited inline in the commit
message.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Refresh of docs/SMOKE-TEST-2026-04-25.md after this afternoon's commits:
e862846 + eee8ce2 connect_to_device login probe + loginWaitSeconds
010811b + 64906c4 write_variable: SP22 prepare-then-write API
b3bf4a8 download_to_device: same login probe + loginWait
All five previously-failing device-side ops (connect/read/write/start_stop/
download/disconnect) now PASS verified end-to-end against:
- SP22 Patch 1
- Control Win V3 runtime up on port 11740
- Test project MCPTest with the user's manual cleanup applied (one
dup library + one rename-without-callers reference fixed by hand
after the morning sweep)
Five upstream bugs remain, each tracked as a separate item with the
proposed cross-reference path to the official scripting docs at
content.helpme-codesys.com:
- create_folder kwarg mismatch
- compile_project / get_compile_messages JSON long serialisation
- list_project_libraries returns empty after successful add_library
- add_library duplicates instead of dedupe / placeholder
- rename_object partial refactor (own decl yes, callers no)
Diff vs morning baseline: 17 -> 22 PASS, 8 -> 5 FAIL.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Same SP21+/SP22 login() drift as connect_to_device:
- 'TryOnlineChange' is gone from OnlineChangeOption
- login() now requires (OnlineChangeOption, bool) -- two positional
The fork hard-coded the old call shape; observed today on SP22 Patch 1
with the runtime up and the project compiled, every download_to_device
attempt failed with "login() takes exactly 2 arguments (0 given)".
Fix mirrors e862846 / eee8ce2:
- Probe OnlineChangeOption members at runtime (priority order tuned
for download: WithDownload / ForceDownload come first, since
'download' implies a write).
- Try (enum, False) / (enum, True) / (enum,) / bool / no-arg shapes.
- Add LOGIN_WAIT_SECONDS post-login state-stabilisation poll for the
credential dialog (default 60s, configurable 0-600).
- Tool-side IPC timeout = waitSec + 120s headroom for the actual
download work (download is heavier than connect, hence 120 vs 30).
The download itself is still done via the existing fall-through:
online_app.download() if exposed, else create_boot_application().
On SP22 the dir() shows source_download + create_boot_application (no
'download'), so the create_boot_application path is what runs.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
After 010811b's diagnostic dump revealed the actual online_app surface on
SP22 Patch 1:
['Dispose', 'application', 'application_state', 'create_boot_application',
'force_prepared_values', 'get_forced_expressions', 'get_online_device',
'get_prepared_expressions', 'get_prepared_value', 'is_logged_in', 'login',
'logout', 'operation_state', 'read_value', 'read_values', 'reset',
'set_prepared_value', 'set_unforce_value', 'source_download', 'start',
'stop', 'timeout', 'unforce_all_values', 'write_prepared_values']
There is no direct write_value / write / set_value method. The supported
pattern is two-step:
online_app.set_prepared_value(path, value) # stage
online_app.write_prepared_values() # commit
Asymmetric to read_value() (which is direct), but it's what the SP21+/SP22
scriptengine surface exposes.
This commit:
- Makes the prepare-then-write path the primary code path.
- Keeps the direct write_value / set_value / write / set fallbacks for
older SPs that still expose them (probe order: prepare-first, then
direct).
- Falls back to dumping dir(online_app) on total failure, same diagnostic
pattern that revealed this API in the first place.
Note for future reference: 'force_prepared_values()' is the alternate
commit method when you want to FORCE a value (override what the program
will write next cycle), versus 'write_prepared_values()' which is a normal
one-shot write.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mirror of the connect_to_device probe pattern (e862846). The fork's
prior write_variable.py hard-coded write_value() then write() and
errored if neither existed; on SP22 Patch 1 neither is exposed on
online_app, even though the read counterpart (read_value()) works.
Now tries six method names in priority order:
- Single-write: write_value, set_value, write, set
- Batch-write: write_values, set_values (passes [(name, value)])
On total miss, dumps sorted dir(online_app) so the next debug session
sees exactly what the live online application object exposes -- the
same diagnostic technique that found 'librarymanager' in the earlier
install_library_file probe.
read_variable already works (uses read_value()) so the asymmetry is
specifically on the write side. Verifying by re-running write_variable
after the MCP restart will reveal which method actually exists, and
we can pin it explicitly in a follow-up if useful.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When connecting to a password-protected runtime for the first time,
CODESYS pops a modal credential dialog. login() in some SP versions
returns immediately (without waiting for the dialog), leaving the
application in an undefined state until the user fills it in.
Behaviour:
- New optional 'loginWaitSeconds' parameter on the tool (default 60,
range 0-600). After login() returns, the script polls
online_app.application_state once per second up to that many
seconds, exiting early when the state lands on a recognisable
value: run / stop / connected / halt / breakpoint.
- During the poll, system.delay(1000) pumps the UI message loop so
the dialog renders and stays interactive while we wait. Once the
user enters the password and clicks OK, login completes, state
transitions, and the loop exits.
- Tool-side IPC timeout extends accordingly (waitSec + 30s headroom)
so the IPC layer doesn't kill the script mid-dialog.
Concrete case observed today on SP22 Patch 1: prior connect appeared
to "succeed" only because the user happened to be at the keyboard and
manually entered the password while the dialog was up. Without the
poll, an unattended run would race past the dialog into a half-state.
Future companion: download_to_device hits the same login path and
should get the same parameter; will follow as a separate commit.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
