Coinbase script parse failures

While writing the script engine for pynode, I ran a test to validate my script tokenizer -- a python script which does nothing more than split up scriptPubKey and scriptSig into component opcodes and data elements. No execution, just tokenization of the script's data stream.

Scanning the entire blockchain, my script found over 8,000 tokenization failures, and 100% of those were in coinbase transactions' scriptSig. The scripts used to generate this can be found at https://github.com/jgarzik/pynode

The following data dump are just the first few, and most recent few, of the invalid scripts I found in the blockchain:

Scanning block #142312 000000000000046acff93b0e76cd10490551bf871ce9ac9fad62e67a0
7ff1d1e (1 tx)
   TX 50cfd3361f7162b3c0c00dacd3d0e4ddf61e8ec0c51bfa54c4ca0e61876810a9
      txin 0 parse failed
Scanning block #142357 0000000000000743c432f84ad688b7b60d1474ccd7baa3d762df0b3f5
1205712 (1 tx)
   TX 587da4d4870515e57efc27623aa92fae0b7aef5908162de57fef0bbe6382be73
      txin 0 parse failed
Scanning block #143014 00000000000007fe6ecd20a8c454cd43c78d912b499c46a1179e30f7c
ff002b3 (1 tx)
   TX 4c8f43c5115c5f29f3761176fa59cde2de2ad976efcbc5faae8ee79fa5dd6264
      txin 0 parse failed
...
Scanning block #190315 00000000000006a0bc3be527033c02d3bcfa72af2f4213c4b0feec923
9573342 (336 tx)
   TX f0ba80ce080eb49148b69c47d744bbb85e4e07e4e4d0273b402c0989d79c359c
      txin 0 parse failed
Scanning block #190321 00000000000001c3bacc869917cacdafb6e00c552ac294835107b574a
44a0362 (38 tx)
   TX 4c91f5ad0616df92165819902d0b117d9e68345f5fe964de6146f89838b9295e
      txin 0 parse failed
Scanning block #190331 00000000000000e3d3eaf93600684b085df7d58f84ef952c91e84eb4a
251d5d8 (128 tx)
   TX 5ee371d65e323934570566b1d92dceb8456e887814da8ef2a53971683bd11da4
      txin 0 parse failed

Coinbase scriptSigs aren't required to be well-formed. They're never executed.

I mentioned this on IRC a week or so ago, noticing that though they are not executed and required to be well-formed, we still count any sigops that appear in them (which I guessed may be an interesting attack if you could get a miner to put a byte in there that is the equivalent of OP_CHECKSIG because we dont count the sigops in the coinbase scriptSig during mining, however luke pointed out that we always push the content of coinbase scriptSigs properly by default, and those modifying the code should spend time researching this stuff anyway, so if they break it, its their fault (and now they can find this email)).

Matt

On Mon, 2012-07-23 at 02:07 -0400, Jeff Garzik wrote:

> While writing the script engine for pynode, I ran a test to validate
> my script tokenizer -- a python script which does nothing more than
> split up scriptPubKey and scriptSig into component opcodes and data
> elements.  No execution, just tokenization of the script's data
> stream.
>
> Scanning the entire blockchain, my script found over 8,000
> tokenization failures, and 100% of those were in coinbase
> transactions' scriptSig.  The scripts used to generate this can be
> found at https://github.com/jgarzik/pynode
>
> The following data dump are just the first few, and most recent few,
> of the invalid scripts I found in the blockchain:
>
> Scanning block #142312 000000000000046acff93b0e76cd10490551bf871ce9ac9fad62e67a0
> 7ff1d1e (1 tx)
>    TX 50cfd3361f7162b3c0c00dacd3d0e4ddf61e8ec0c51bfa54c4ca0e61876810a9
>       txin 0 parse failed
> Scanning block #142357 0000000000000743c432f84ad688b7b60d1474ccd7baa3d762df0b3f5
> 1205712 (1 tx)
>    TX 587da4d4870515e57efc27623aa92fae0b7aef5908162de57fef0bbe6382be73
>       txin 0 parse failed
> Scanning block #143014 00000000000007fe6ecd20a8c454cd43c78d912b499c46a1179e30f7c
> ff002b3 (1 tx)
>    TX 4c8f43c5115c5f29f3761176fa59cde2de2ad976efcbc5faae8ee79fa5dd6264
>       txin 0 parse failed
> ...
> Scanning block #190315 00000000000006a0bc3be527033c02d3bcfa72af2f4213c4b0feec923
> 9573342 (336 tx)
>    TX f0ba80ce080eb49148b69c47d744bbb85e4e07e4e4d0273b402c0989d79c359c
>       txin 0 parse failed
> Scanning block #190321 00000000000001c3bacc869917cacdafb6e00c552ac294835107b574a
> 44a0362 (38 tx)
>    TX 4c91f5ad0616df92165819902d0b117d9e68345f5fe964de6146f89838b9295e
>       txin 0 parse failed
> Scanning block #190331 00000000000000e3d3eaf93600684b085df7d58f84ef952c91e84eb4a
> 251d5d8 (128 tx)
>    TX 5ee371d65e323934570566b1d92dceb8456e887814da8ef2a53971683bd11da4
>       txin 0 parse failed
>