Headless EDT
Auto-start of a headless EDT, waiting for the project models in a CI job, and moving a ready workspace to another machine.
The wrapper starts a headless EDT when EDT is not running. This page describes that start, the wait for the project models in a CI job, and moving a workspace that EDT has already indexed.
Auto-start
During auto-start, a detached supervisor keeps the CLI input open and tracks the launched
1cedtcli process. If the CLI exits before the bridge is ready, the wrapper reports its exit
code and up to 16 KiB from the end of its combined stdout/stderr, decoded as UTF-8 with replacement
for invalid bytes. The supervisor exits with the CLI, and a later call can retry the launch.
--start-timeout limits the readiness wait while the CLI is still running. A timeout leaves
the live process running; EDT also continues running after a short-lived wrapper command exits.
The timeout message names the supervisor process and says whether the EDT console had already
answered. If it had, EDT itself started and the plugin did not, so waiting longer will not help.
Before the start, the wrapper checks that it can read the plugin jar in dropins: EDT runs as
the same user and could not read it either. While EDT starts, the wrapper reads the workspace log
.metadata/.log. When the plugin cannot load there, the wrapper stops the headless EDT it started
and reports the cause.
One headless EDT serves one workspace. The supervisor holds a lock for its workspace while it runs. A second wrapper started on the same workspace waits for that EDT, even if the process list does not show it yet or the PowerShell query behind that list on Windows times out.
When the bridge does not come up, headless or in the EDT window, run edt-bridge-mcp doctor --workspace <ws>. The command starts and changes nothing. It checks the jar in dropins and
whether EDT can read it, which bridge version each EDT configuration registers, whether EDT runs
as administrator, and whether the last EDT session logged a plugin load error. Each problem it
finds comes with the step that fixes it. Right after self-update the EDT configuration still
registers the replaced jar. That is not a problem: EDT rewrites the entry on its next start, and
the command shows it as a note with exit code 0.
Waiting for the models
The bridge answers before a large project is indexed, and a tool called in between is refused
with “no BM model”. In a CI job, wait for the models with edt-bridge-mcp wait-ready --project <name> --timeout <seconds>. The command starts a headless EDT when none is running and polls
/status until every named project reports modelReady; without --project it waits for every
open EDT project. It exits with code 0 when the projects are ready, 1 when EDT is not running or
has exited, and 2 when the time runs out while EDT is still loading. A project imported after the
start is awaited as well.
A ready workspace
Importing and indexing a large configuration takes a headless EDT tens of minutes, while a
workspace that EDT saved on a graceful stop opens in a few. edt-bridge-mcp workspace moves
such a workspace to the next machine or container:
- Start EDT on the workspace, import the projects and wait for the models:
edt-bridge-mcp wait-ready --workspace <ws> --project <name>. - Save the EDT checks of the projects:
edt-bridge-mcp workspace checks --workspace <ws> --object <fqn> --output before.json. The command runsedt_project_errorswith a refresh for every project, or for every object given with--object, and refuses a list that is incomplete, not settled or cut at its limit. On a large configuration check a few objects whose checks depend on the indexed model: the whole project can exceed the problem limit. - Stop EDT with
edt-bridge-mcp shutdown --workspace <ws>, so that it saves its state. - Pack the workspace:
edt-bridge-mcp workspace pack --workspace <ws> --archive ready.tar.gz. - On the next machine restore it:
edt-bridge-mcp workspace restore --archive ready.tar.gz. - Start EDT there, wait for the models and compare:
edt-bridge-mcp workspace checks --workspace <ws> --compare before.json.
The archive is a PAX tar, .tar, .tar.gz or, with Python 3.14 and newer, .tar.zst. It keeps
the modification times with their fraction of a second, because EDT compares them with the state
it saved. It leaves out the workspace lock, the logs of .metadata (EDT writes its command line
with the bridge token there) and the pid and log lock files. EDT keeps absolute paths in its
metadata, so restore unpacks the workspace only under the path it was packed from, into a
missing or empty directory, on the same kind of system. pack refuses while an EDT uses the
workspace, and both commands name the process or the lock that holds it. A project whose files
live outside the workspace directory is refused as well, because the archive would not carry it.
checks --compare exits with code 0 when the checks match and 2 when they differ, and lists
the problems found on one side only. Where the archives are kept and which inputs name one is
left to the job that builds them.