Closed Stebalien closed 3 years ago
The alternative is to just over-estimate and pick a high number. Unfortunately, it would have to be pretty high for 2k sectors...
This needs to land in V3. At the moment, we have some issues on calibration net with 2k sectors because window post proofs are much larger.
Larger = 192*3000 = 562KiB because up to 3K 2-sector partitions (for 2k sectors) can be proved at once.
Merging this now to get it into Lotus 1.5.0 code-freeze.
Specifically, this fixes an issue where we could submit a window post for many (i.e., more than 10) partitions in a single window post for 2k sectors. This would lead to proof sizes over 1920 bytes (the old max).
Now, we accurately compute the proof size limits from actual known constants.
Depends on https://github.com/filecoin-project/go-state-types/pull/23.