Transaction Header should include upper limit on compute time

[bytemaster]

Currently it includes the upper limit of the context free compute; however, this requires us to recompute the exact instruction count on replay and does not allow the transaction creator to prevent rogue contracts (such as notify on transfer) from consuming an unexpected CPU instruction count.

[wanderingbort]

If we don't charge the max at all times then we still have to compute the exact instruction count on replay. A max would provide some defense against expensive notifies/inlines.

…

On Fri, Mar 30, 2018, 9:22 PM Daniel Larimer ***@***.***> wrote: Currently it includes the upper limit of the context free compute; however, this requires us to recompute the exact instruction count on replay and does not allow the transaction creator to prevent rogue contracts (such as notify on transfer) from consuming an unexpected CPU instruction count. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1998>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACYR4tIpX3ch3SeAogOUzlstm3IArmn4ks5tjtpTgaJpZM4TCR9I> .

[bytemaster]

We always charge the declared amount, we just assert the actual is less than declared.

[bytemaster]

Your use of std::forward<> is wrong, should be std::move()

[arhag]

I just noticed this issue. I posted a very similar one here: #1999.

[bytemaster]

Partial work on this in commit: 18009d1

[bytemaster]

Closing due to dupe #1999

[andriantolie]

Duplicate. Put the label here, so this issue's marked to be unneeded for test