bindings/go: make it work with upcoming go1.24. #247
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
In an unexpected turn of events go1.24 reclassified import "C" non-local, which rendered C.blst_* type aliases defunct. The only working solution appears to be to wrap them into structs. Since all the C.blst_* types were opaque, the wrapping shouldn't affect applications. In the sense that the shell struct won't obstruct anything, because nothing was visible.
|
Thanks for the quick turnaround @dot-asm. I can confirm that with this change, I am able to successfully build my project using the blst Go bindings, whether I am still using go1.23, go1.24rc1, or with Go tip. |
mark-rushakoff
added a commit
to gordian-engine/gordian
that referenced
this pull request
Jan 8, 2025
This uses the supranational/blst package for an implementation of BLS signatures. We use the minimized signature form, hence the package name. The package includes a BLS-backed gcrypto.PubKey implementation and a CommonMessageSignatureProof implementation. There are likely some missing details in the BLS setup, at least in adding configurability for the signatures' salt. I think the domain separation tag is correct. I cannot find the whitepaper or technical document that introduced me to the concept, but the model of arranging the validators in a tree for key and signature aggregation is based on the concept that the leftmost validators are trustworthy and likely to be online, while the rightmost validators are the least expected to vote. See further documentation on the SignatureProof type. It has not been fully integrated yet, but we have tests around the usage of both of those types. Next up are some slight interface changes that will affect the ed25519 implementation too. And I think we may need some special handling such that the signature committed to chain is a "final aggregation" of keys that would not normally be aggregated together. For instance, if there are only four validators, then during consensus gossip, we would pair 0-1 and 2-3, but never 0-2 or 1-3. If the final votes were only from 0, 1, and 3 -- that is to say validator 2 was absent or perhaps voted nil -- then we would commit with the aggregated signaure 0-1-3, to minimize space used for signatures. This also removes the gmerkle package, which was left in for specifically this use case, but which turned out to be an incorrect fit. Note that with the introduction of this package, we have a dependency on supranational/blst, which is a C dependency; and there is currently the outstanding issue supranational/blst#245 which prevents building blst with Go 1.24. However, supranational/blst#247 is an open PR that fixes the build for Go 1.24 and Go tip.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In an unexpected turn of events go1.24 reclassified import "C" non-local, which rendered C.blst_* type aliases defunct. The only working solution appears to be to wrap them into structs. Since all the C.blst_* types were opaque, the wrapping shouldn't affect applications. In the sense that the shell struct won't obstruct anything, because nothing was visible.