0.0.1 was broken, and we pulled it¶
IntentumDiff 0.0.1 was published on 4 August 2026 and withdrawn shortly afterwards.
If you were one of the people who installed it in that window: it did not work properly, you were right to think something was wrong, and we're sorry. You spent time on something that should not have shipped.
0.0.1 has been withdrawn
The VS Code extension has been removed from the Marketplace and Open VSX, and the Python
package has been yanked from PyPI. pip will not install it, and you should not pin to
it. There is no working release available right now; the next one will be 0.0.2b1.
What was wrong¶
| What you saw | What was actually happening |
|---|---|
A wall of Failed to catalog parser plugin errors on every single run |
The package left its own name off its internal trust list, so all 78 bundled parsers were rejected as untrusted third-party code |
python -m intentumdiff failed outright |
There was no __main__.py. Only the intentumdiff command worked |
| An error message pointed at documentation that did not exist | It linked to a domain that had never been registered |
The README's headline example raised NameError |
It was a code fragment, not a runnable program. Nobody had ever executed it |
Every one of these was visible within a minute of installing the package.
Why we did not catch it¶
Every check was green when we shipped. That was the problem, not an excuse.
The tests proved the code compiled and behaved correctly in a source checkout. Nothing in our CI ever installed the published wheel and used it the way the README says to — and three of the four faults above exist only in an installed package. The fourth was in the documentation, which nothing executed at all.
We were measuring the wrong thing and mistook a green board for a working product.
What has changed¶
Not intentions — checks that fail the build:
- The published artefact is installed into a clean environment and used before any
release. Install, import, console script,
python -m, a real diff, and — the check that would have caught the worst of it — stderr must be empty. 0.0.1 returned correct results while printing 69 errors, and exit code 0 called that a success. - Every code example in this documentation is executed in CI against a real installed wheel, and its output compared with what the page claims. An example that cannot run now fails the build.
- Every link is checked, so documentation cannot point at a page that does not exist.
- Releases start as betas. A version number without a beta marker is a claim of stability, and that claim should follow use, not precede it.
Where things stand¶
The next release will be 0.0.2b1. It will not go out until the checks above pass against the
real artefact rather than a source checkout.
If you hit something that does not behave the way these docs describe — including the docs themselves — please tell us. That is exactly the kind of report that would have caught this one.