Latest — this was the bug and it is now fixed on curio feat/pdp

This is a real tricky one that’s been taking me a few days

Proving fails when I add a bunch of 100 MiB chunks, very surprising since Magik has tested out chunked proving stuff and its worked before

I must be exercising a weird code path

I chunk 100 MB and then padding pushes those to 134.. MB (100 MiB)

All of the pieces refs look good in the table, padding numbers check out

The subtree proof appears to be working fine. It looks like the issue with verification is probably in the recomputed tree tops. However this code looks fine from inspection

One unexpected phenomena so far has come from adding additional logs

sub roots are at level 23. At level 25 the “computed” root no longer matches the output of the hash of the level 24 nodes on the merkle path. It might just be a logging thing or it might mean we are creating the wrong proof going up the wrong path

We get a root == 6b225db051b6fb6511d4a4b9f51e5328f7b44ca640952fe3d7aaca1a35e60105 from the partial tree computing with the weird anomoly

For the actually verification proof we get 9bcf4395becf711d8534a0679c1b9994682ba673cb856929d8665799639e1a96

Another run we compute out dbf8cf8ec90cde4b3af0b9aad2f54827926824b45dee0356ad0b66a5807b89c3 for the root different

Partial tree root looks like 6b225db051b6fb6511d4a4b9f51e5328f7b44ca640952fe3d7aaca1a35e60105 the same

and again on level 25 we see the anomaly