fix dirtybit #4095
fix dirtybit #4095
Conversation
|
I'll give this a try locally. |
|
Can confirm that this does indeed seem to fix #4094 |
| @@ -248,7 +248,7 @@ export default class Renderer { | |||
| .reduce((lhs, rhs) => x`${lhs} | ${rhs}`); | |||
| } | |||
|
|
|||
| return x`${dirty} & /*${names.join(', ')}*/ ${bitmask[0].n}` as BinaryExpression; | |||
| return x`${dirty}[0] & /*${names.join(', ')}*/ ${bitmask[0].n}` as BinaryExpression; | |||
There was a problem hiding this comment.
If we're not doing the destructuring, do we even need a separate case for this now? (And do we need the concept of context_overflow?)
There was a problem hiding this comment.
oh a good point, i didnt know what else context_overflow is for, let me do a global search and understand them... 🤔
There was a problem hiding this comment.
If I remove the check for context_overflow here and always act like there was an overflow, the tests still pass. It looks like the other place that context_overflow is used is in conditionally constructing an array of -1s to pass to the @init helper. If nothing is passed, that argument defaults to [-1]. I think that makes sense to continue to be a default (since a lot of the time there will be at most 31 things to track), but I think it would make sense to just do the <= 31 check there where we're (conditionally) filling the array, and get rid of context_overflow.
There was a problem hiding this comment.
yup let me update them
Fix #4094
Found the problem is that we destructure the dirty bits array if it doesnt overflow.
but problem is, it is a number, instead of an object.
The
changemap can still be updated within thepfunction. and previously is ok, because it is an object.Now, although the
dirtybitmap array is updated, thedirtyis not updated, because we eagerly destructure it into a number.so after this change,
dirtyis an array, whether or not it is overflowed.