This is a long one.
So, `instanceof` in Typescript is incredibly flimsy. There's a
documented bug with what happens when you extend built-in classes, such
as Error or Array. You can read more about it here: https://github.com/Microsoft/TypeScript/wiki/FAQ#why-doesnt-extending-built-ins-like-error-array-and-map-work
However, this normally works fine. A previous commit moved some schemas
from server/ to common/, our gameSpecific UGPT settings schemas. This
caused them to break in an interesting way. For any schema recursive
with a depth greater than 1, all errors would be ignored. Why on earth
was this happening?
Well Prudence passes errors up the stack with a "return Err" -> "if
ret instanceof Err: return Err" kind of approach. This normally works
fine.
The real kicker here is that because server/ and common/ were on 0.9.7
and 0.9.8 respectively, the "Err" instance in question would actually be
erased when ran at runtime. The "if ret instance of Err" would fail,
because the Err would be from 0.9.7, not 0.9.8. This subtle difference
in file results in catastrophic damage.
In short. InstanceOf is a footgun. I hate it. Lets never do it again.
Thanks.
between 8th CS and 16th CS, jelly kiss was changed such that its
DPN became a DPH, and its DPH became a DPA.
this is a problem, as we haven't accounted for this change properly.
we're just going to remove those versions from the db, and change it to
16-cs.
This patch
- Updates broken notecounts on a series of charts
these broken notecounts are *extremely* sinister, only off by one
or two in some cases. Furthermore, this has happened on relatively underplayed charts.
As such, this has gone unnoticed for a while!
- theme of denjin J is now rated 7 instead of 0
no idea why it was 0, but it was
- jelly kiss's another charts now appear
no idea why they didn't
- back into the light (recut) and the hope of tomorrow (recut) now appear in 28-omni
they didn't before, no idea why.
- サナ・モレッテ・ネ・エンテ(B.L.T.STYLE) now appears in 28-omni
same as above, no idea why.
this works by cartesian producting our IDs, so that they're all added
to the set of unique IDs. If a chart has a unique ID On
["A", "B"], C, D
then A C D and B C D will be correctly be added to the set of uniq ids
now.
This is a regression test for the behaviour that let slightly broken
khamen break charts get into prod.
This chart has 48 alternative versions, and is only playable
in certain prefectures. Since this is antithetical to the point of a
score tracker (not everyone can play the chart), we're removing it
effective immediately.
What a stupid gimmick.
Christopher Cooper