Stroke/type mismatch for 丱

[ospalh]

Looks like there is a consistent inconsistency between the stroke order (numbers and stroke vector data) and the stroke types inside the SVGs for all occurrences of the 丱 element:
For the right side, what is shown is vertical (the rightmost stroke) horizontal vertical (left of the group).
But the kvg:type attribute is "㇑", "㇑", "㇐", that is vertical, vertical, horizontal.
I have no idea what the correct order is. Maybe both are allowed. (And Google is not your friend, as many sites use the kanjivg data.)
It looks like the Chinese write this from left to right, vertical, vertical horizontal.

grep gave me this list of files:

04e31.svg
05618-Jinmei.svg
0589f-HzFstLeRi.svg
0589f-HzFstRiLe.svg
0589f.svg
0589f-VtFstRiLe.svg
06b54-HzFstLeRi.svg
06b54-HzFstRiLe.svg
06b54.svg
06b54-VtFstRiLe.svg
0806f-Kaisho.svg
0806f-KaishoVt3.svg
0806f.svg
095dc-Kaisho.svg
095dc.svg

[eichhirn]

Kakijun explicitly repeats the stroke order verbally in the notes for that kanji. The stroke order prescribed by kakijun matches the number/stroke vector order of kanjivg, so the type attributes are swapped.
http://kakijun.jp/page/kan05200.html
Stroke 2 of 丱 (component "丿") is also incorrect: it must be 31D3 (vertical + throw away), not 31D2 (throw away). This is not picked up by my stroke type/direction checker.
[Update:
Looking at my component group analysis I see the "丿" component is (almost) consistently classified as 31D2 in 2725 files. (Component 丿, 4E3F is Radical 4 = piě, "slash".) The 31D3 stroke classification is never used in any file. Showcase examples for the 31D3 stroke type would include 人, 用,月 (first stroke each).
Conversely, the 31D2 stroke type is used in 9609 files. Why only some of them have been boxed into a "丿" component is mysterious.
The omission of the 31D3 stroke type is a systematic error. Fixing that in one place only would just increase the chaos.
]
As far as i know the 丱 component is never used in any Jouyou-Kanji, so this is a bug of lowest priority.

[ospalh]

I mostly wanted to make sure that the visible numbers were right.
I admit that i didn’t fix some stroke type attributes i had found in other files, so maybe i won't fix these any time soon, either.

[ghost]

The commit above fixes the stroke types for the three erroneous files returned by git grep 丱.

[ospalh]

Sorry, i may be a bit confused.
But i am not sure that all instances of this are fixed now.

[benkasminbullock]

I've visually inspected all of the 05618, all of the 06b54, all of the 0806f, and all of the 095dc and I've spot checked the source SVG of one file from each of these sets, as well as checking with the multicoloured stroke approach I used in my kanjivg viewer. I don't see any evidence that there is anything wrong with the 丱 element in any of these files.