cairo_native icon indicating copy to clipboard operation
cairo_native copied to clipboard

Error in tx 0x6c51758aa1ae9506602fffb9194da427fe948314b74eb93cdc9570558d4a88d, Killed process

Open pefontana opened this issue 1 year ago • 3 comments

When you execute the 0x6c51758aa1ae9506602fffb9194da427fe948314b74eb93cdc9570558d4a88d in starknet-replay, the process gets Killed In starknet-replay, in a x86 computer

cargo run --release tx 0x6c51758aa1ae9506602fffb9194da427fe948314b74eb93cdc9570558d4a88d mainnet 648461
INFO replay: starting execution
    in replay::transaction with hash: "0x6c51758aa1ae9506602fffb9194da427fe948314b74eb93cdc9570558d4a88d", chain: "mainnet"

Killed

pefontana avatar Jul 05 '24 17:07 pefontana

The problem lies in MLIR's canonicalization pass. If you run mlir-opt --canonicalize contract.mlir it'll crash with an out of memory.

Disclaimer: Test the bug at your own risk, OOM errors can crash not only themselves but other apps too. contract.mlir.txt

azteca1998 avatar Jul 05 '24 19:07 azteca1998

Adding:

  • block: 642001
  • transaction: 0x07666fcaa051e318762020635b0e344fef06bd71f3b05df8ce4e7349b6989345
  • class hash: 0x05b3656430eebda05c8b2722a7e13206d2afb49dcca0be85b545b90f35a3728d

igaray avatar Jul 19 '24 10:07 igaray

llvm-19.1.0-rc2 seems to fix this bug. I tested it by running mlir-opt with both contracts mentioned in this issue, and both of them finished successfully

JulianGCalderon avatar Aug 09 '24 20:08 JulianGCalderon

Fixed with https://github.com/lambdaclass/cairo_native/pull/763

pefontana avatar Sep 27 '24 18:09 pefontana