Yes, and it's nothing wrong with it, since `accept_message()` just sets gas limit to maximum available. Even more, if you place `accept_message()` after signatures' checks, smc may exceed gas limit processing messages signed by many participants, and these messages will just be dropped.

Yep, you're right. I expected that if smc failed processing message within initial gas limit, it won't pay for fuel even if `accept_message()` was already called. But that's wrong. And I had to check this assumption before submitting :)

Thanks for a feedback!

I agree, `accept_message()` is called too early and smc may throw after message was accepted which isn't good also.

But `seqno` is actually signed. In `code.fc` there is only signatures dict stripped from `in_msg` and the rest of it is participated in signatures' check later. Likewise in `sing-msg.fif` signatures dict at the beginning is cut, but the whole slice's remainder is signed.

And I don't get how gc could take O(N**2) even potentially. I'm building a completely new dictionary including only key-values from original one, but skipping expired orders. Could you please explain where is the problem in this approach? Except that I probably shouldn't run gc on every valid external message.