* Add a quick run script
* Add more context to errors
* Fix the issue with corpus directory canonicalization
* Update the quick run script
* Edit the runner script
* Support specifying the path of the polkadot sdk
* Configure kitchensink to use devnode by default
* Update the kitchensink tests
* Fix the logic for finding the ABI in resolc
* Edit how CLI reporter prints
* Remove the old reporting infra
* Use the Test struct more in the code
* Implement the initial set of reporter events
* Add more runner events to the reporter and refine the structure
* Add reporting infra for reporting ignored tests
* Update report to use better map data structures
* Add case status information to the report
* Integrate the reporting infrastructure with the
CLI reporter used by the program.
* Include contract compilation information in report
* Cleanup report model
* Add information on the deployed contracts
* Fix the OS FD error
* Cache the compiler versions
* Allow for auto display impl in declare wrapper type macro
* Better logging and fix concurrency issues
* Fix tests
* Format
* Make the code even more concurrent
* WIP redo how we parse and use modes
* test expanding, too
* WIP integrate new Mode/ParsedMode into rest of code
* First pass integrated new mode bits
* fmt
* clippy
* Remove mode we no longer support from test metadata
* Address nits
* Add ability for compiler to opt out if it can't work with some Mode/version
* Elide viaIR input if compiler does not support it
* Improve test output a little; string modes and list ignored tests
* Move Mode to common crate
* constants.mod, and Display for CaseIdx to use it
* fmt
* Rename ModePipeline::E/Y
* Re-arrange Mode things; ParsedMode in format and Mode etc in common
* Move compile check to prepare_tests
* Remove now-unused deps
* clippy nits
* Update fallback tx weights to avoid out of gas errors
* Update kitchensink weights too and fmt
* Bump default geth timeout to 10s
* 30s timeout
* Improve geth stdout logging on failure
* fix line logging
* remove --networkid and arg, back to 5s timeout for geth
* Use channels to send data to reporting thread and avoid hangs / mutex / duration. Limit max concurrent tasks to avoid too many open files
* More appropriate name for dirver/reporter task fns
* Back to parallelise individual cases, report individual cases, address grumbles
* newline before 'Failures' title in report
* Make metadata serializable
* Refactor tests to use steps
* Add a balance assertion test step
* Test balance deserialization
* Box the test steps
* Permit size difference in step output
* Added a resolver tied to a specific block
* Increase the number of private keys
* Increase kitchensink wait time to 60 seconds
* Add a case ignore flag
* Parallelize over cases
* Rename the state and driver
* Parallelize execution
* Update the default config of the tool
* Make codebase async
* Fix machete
* Fix tests & clear node directories before startup
* Cleanup the cleanup logic
* Rename geth node
* Allow for downloader to use version requirements.
We will soon add support for the compiler version requirement from the
metadata files to be honored. The compiler version is specified in the
solc modes section of the file and its specified as a `VersionReq` and
not as a version.
Therefore, we need to have the ability to honor this version requirement
and find the best version that satisfies the requirement.
* Request `VersionOrRequirement` in compiler interface
* Honor the compiler version requirement in metadata
This commit honors the compiler version requirement listed in the solc
modes of the metadata file. If this version requirement is provided then
it overrides what was passed in the CLI. Otherwise, the CLI version will
be used.
* Make compiler IO completely generic.
Before this commit, the types that were used for the compiler input and
output were the resolc compiler types which was a leaky abstraction as
we have traits to abstract the compilers away but we expose their
internal types out to other crates.
This commit did the following:
1. Made the compiler IO types fully generic so that all of the logic for
constructing the map of compiled contracts is all done by the
compiler implementation and not by the consuming code.
2. Changed the input types used for Solc to be the forge standard JSON
types for Solc instead of resolc.
* Fix machete
* Add resolc to CI
* Add resolc to CI
* Add resolc to CI
* Add resolc to CI
* Allow for files to be specified in the corpus file
* Attempt to improve the geth tx indexing issue.
We're facing an issue where Geth transaction indexing can sometimes stall
on some of the nodes we're running. The logs show that for all transactions
we always need 1 second of waiting time. However, during certain runs we
sometimes run into an issue with some of the nodes where it seems like
their transaction indexer fails (either at the start or after some amount
of time) which leads us to never get the receipts back from these specific
nodes.
This is not a load issue as it appears like all of the other nodes handle
it just fine. However, it looks like once a node gets into this state it
can not get out of it and its bricked for the entire run.
This commit adds some more command line arguments to the geth command in
hopes of improving this issue.
* Re-order the input file.
This commit reorders the input file such that we have a definitions
section and an implementations section and such that the the order of
the items in both sections is the same.
* Implement a reverse polish calculator for calldata arithmetic
James Wilson