Add host functions to support EVM #316
Conversation
|
A few build related things I've noticed through a quick look over so far: Need to add the LICENSE files for the new libraries as installed files; down in this area https://github.com/eosnetworkfoundation/mandel/blob/main/CMakeLists.txt#L235 These libraries need to be added to the libtester.cmakes. libff as configured adds a requirement of SSE4.1. I guess this isn't a problem 🤷 but we may want to consider adjusting compile flags elsewhere if we're going to require it in one localized spot already. libff adds back gmp as a dependency. I just got done removing that slippery dep (#128) 😞 Is there no other way? Did OpenSSL not have the needed functionality? Will need to fork or PR a change to libff to ensure it always dynamically links to GMP otherwise it'll be a GPL violation. You can see how that was done historically over here |
|
Also may need to detect and USE_ASM=Off for compatibility with #317 |
| std::memcpy( output.data(), res.data(), copy_size ); | ||
| } | ||
|
|
||
| int32_t interface::ecrecover( span<const char> signature, span<const char> digest, span<char> pub) const { |
There was a problem hiding this comment.
Feels like a missed opportunity to add this without support for EOS key types. Such functionality has long been requested.
There was a problem hiding this comment.
If any of these could be useful for non-EVM specific use cases, like you're suggesting here and I suspect others would also be useful as well, then might it make sense to give this feature a more generic/feature specific name and not something that suggests it's only useful for EVM implementations?
There was a problem hiding this comment.
That's a good point. These arguably should have a more generic protocol feature name because some of the new host functions are certainly not specific to EVM: anyone might want to make a sha3 hash for all sorts of purposes.
There was a problem hiding this comment.
Any suggestion for the protocol feature name?
Alternatively we can split EVM_PRECOMPILES into many protocol features.
- ALT_BN128
- MODEXP
- BLAKE2F
- SHA3
- SAFE_ECRECOVER
- Add libff to libtester cmakes - Update libfc
|
We are going to need to resolve the libff installs as I think they're going to pollute the package. Worst case I can deal with it once we merge. |
spoonincode
dismissed
their stale review
I can deal with build fixes after merge if need be
Added an updated configuration `deep_mind_handler` to support special adjusted behavior for the deep mind logger to be used specifically in the context of tests. In particular a new config flag called `zero_elapsed` is added to this config which if set to true will first zero out the `elapsed` fields within a copy of the transaction trace before writing it out to the logs. The `zero_elapsed` flag is set to true in the deep_mind_tests in order to guarantee a deterministically generated log to compare against the committed log. This appears to be a more reliable solution to avoiding non-determinism in the tests than relying on the auto-generated `[[:xdigit:]]` masks in the committed logs. It may now be reasonable to greatly simply that log merge and comparison logic, but that is not addressed in this commit.
Updated to reflect the changes to logs propagated from the changes to the initial chain startup sequence in tests (due to new protocol features that are now also being activated at test startup by default).
This does not change anything, but it is a nice pattern to follow so that we can easily recall by looking at the source in which order we introduced new protocol features into newer releases of the code.
Add the two new protocol features to the list of expected protocol feature within the script.
|
All jobs in https://github.com/eosnetworkfoundation/mandel/runs/6891133585?check_suite_focus=true passed except for "Ubuntu 20.04 | Contract Builder Docker". That job fails because the CICD is running for a fork and so it does not have access to the appropriate GitHub secrets needed for that job. But since all relevant tests pass, I will be merging this PR in anyway. |
No description provided.