I see the errors:

api.HTMLInputElement.webkitdirectory - Error → The data for (chrome_android) says no support, but contains additional properties that suggest support.
api.HTMLInputElement.webkitdirectory - Error → chrome_android cannot have a version_added: false in an array of statements.

How can I convey the meaning I intend while passing these checks?

Does this make sense?

            "chrome_android": {
              "version_added": "132",
              "notes": [
                "In Chrome 131, the property could be set, but choosing a directory crashed the browser.",
                "Before Chrome 131, the property could be set, but had no effect."
              ]
            },

Does this make sense?

It makes sense, but based on the following existing case, it might be more consistent to move those notes to a [131, 132) statement with partial implementation.

The reasoning probably is that these notes are not relevant for Chrome 132, where the feature is fully supported.

If feels weird to mark the implementation partial in Chrome 131. In a way, this was more functional before Chrome 131 than in Chrome 131. In Chrome 131, this crashed the browser. Before Chrome 131, it was merely ignored.

If possible, I'd mark the implementation unsupported until Chrome 131 and differently unsupported from Chrome 131 until Chrome 132, but I don't think yari supports that.

If feels weird to mark the implementation partial in Chrome 131.

There is no perfect solution, but it's definitely also weird to have the note on the current support statement, as if the caveat in Chrome 131 was relevant for Chrome 132+. @ddbeck has proposed to allow notes with version ranges, which could help here.

I really think it's fine to add the partial implementation entry for Chrome 131. In MDN's BCD table, you wouldn't notice this version, unless you actively expand the history.

Added to tomorrow's BCD call agenda.

I think the issue here is that the notes are vague and we're missing data for the webkitdirectory HTML attribute, which is where some of this information should live.

The primary problem that the webkitdirectory HTML attribute is untracked. We're missing a feature, html.elements.input.webkitdirectory, that captures support for the actual attribute. It's really hard to tell what support for attribute reflection should be without knowing the support status of the underlying HTML attribute.

Before Chrome 131, the property could be set, but had no effect.

The failure mode here is not explained. I think there are three plausible readings:

1. "webkitdirectory" in HTMLInputElment.prototype is true, but someInputElement.webkitdirectory = "" doesn't set the attribute (i.e., after someInputElement.webkitdirectory = "", someInputElement.webkitdirectory is still undefined).
2. "webkitdirectory" in HTMLInputElment.prototype is true, setting the attribute succeeds, but the attribute has no behavior (i.e., it's equivalent to <input type="file" fake-bool-attr>).
3. "webkitdirectory" in HTMLInputElment.prototype is false. That is, reflection of this attribute is not supported.

In the first two cases, you'd have a partial_implementation, but you'd have to write rather different notes. In the last case, it'd be a noteless "version_added": false entry.

In Chrome 131, the property could be set, but choosing a directory crashed the browser.

This sounds like the attribute reflection works just fine. I'd much rather have a note (and partial implementation) on the hypothetical html.elements.input.webkitdirectory feature that says, "In Chrome 131, choosing a directory crashes the browser." api.HTMLInputElement.webkitdirectory could mention this fact too, but it would have no partial_implementation.

One last note: I assumed that 131 still crashes, not that 131 shipped and was hotfixed to fix this crash—otherwise some very different notes and data apply. Writing the notes in the present tense (i.e., "In Chrome for Android 131, choosing a directory crashes the browser.") helps avoid this kind of ambiguity.

The primary problem that the webkitdirectory HTML attribute is untracked. We're missing a feature, html.elements.input.webkitdirectory, that captures support for the actual attribute.

@danielhjacobs Would you be up for adding that feature under html.elements.input here?

"webkitdirectory" in HTMLInputElment.prototype is true, setting the attribute succeeds, but the attribute has no behavior (i.e., it's equivalent to <input type="file" fake-bool-attr>).

This is the one that applies here.

Additionally, 131 does still crash

In #24543 (comment), @caugner suggested "version_added": false, so I guess we're going back on that.

The primary problem that the webkitdirectory HTML attribute is untracked. We're missing a feature, html.elements.input.webkitdirectory, that captures support for the actual attribute. It's really hard to tell what support for attribute reflection should be without knowing the support status of the underlying HTML attribute.

Isn't it more accurate to track that with https://developer.mozilla.org/en-US/docs/Web/HTML/Element/input/file#webkitdirectory?

Something like this for input type file - webkitdirectory?

Unless I misunderstood something, the bottom line seems to be that we need three support statements here (on api.HTMLInputElement.webkitdirectory):

1. Supported in 132+
2. Partially supported in 131 (mention crash)
3. Partially supported in 18-130 (setting the attribute succeeds, but the attribute has no behavior).

Isn't it more accurate to track that with https://developer.mozilla.org/en-US/docs/Web/HTML/Element/input/file#webkitdirectory?

I think it makes more sense to keep it next to the other input attributes (like capture, which also only applies to the file type), i.e. html.elements.input.webkitdirectory.

Unless I misunderstood something, the bottom line seems to be that we need three support statements here (on api.HTMLInputElement.webkitdirectory)

But didn't @ddbeck say:

I'd much rather have a note (and partial implementation) on the hypothetical html.elements.input.webkitdirectory feature that says, "In Chrome 131, choosing a directory crashes the browser." api.HTMLInputElement.webkitdirectory could mention this fact too, but it would have no partial_implementation.