rebase, still need a proper activation test

(this was actually missing one deactivation -- the per-tx limit, see D5076)

OK we can do these later...

no prob :)

linked to wrong PR # (transposed digits)

An alternative possibility is to keep the current performance hit of using HaveCoin, and put it to work: reintroduce BIP30 checking for all new blocks. Then we aren't relying on our confidence in BIP34 to save us from bad times (which we maybe shouldn't be so confident about -- I wonder if there might be yet something else that every has overlooked).

update

typo switcing -> switching

rebase (conflict with hidden args changes in D5054)

(reviewed it again myself, am happy with this)

update the ContextualCheckBlock comment (this diff adds a second
check of coinbase in ConnectBlock, so scary warning is not needed)

use static_assert

The (slight) danger of using AddCoins with check=false is that a block with a duplicate transaction will cause an exception if it gets to this point. I don't think this can really happen in practice: currently we have CheckBlock (ensures first tx is coinbase and nothing else is coinbase) and new blocks are checked with CTOR (excludes duplicate transactions among non-coinbases), and we have checkpoints to ensure alternative pre-CTOR blocks are not considered.

Ah OK, nevermind, as soon as I posted this I realized what's going on ... I get it now. :)

consted

In D4903#122571, @deadalnix wrote:

Split out the miner change and we can land the rest. Let's not delay due to open discussion.

use block flags

add comments

chmod

add -automaticunparking option

we will not be using ScriptExecutionMetrics

updated all tests to pass, still need a proper activation test

Blarg, compiler bugs, that is fun :D

add functional tests

small tweak - instead of using assert A.issubset(B), use
assert_equal(A.difference(B), set()) -- the latter one prints
out which A transactions are missing from B, if it fails.

rebase; tweak ATMP virtualsize computation (same result)

expand tests greatly -- more logging statements, test a package with a very low sat/vbyte parent, and add blockmintxfee tests.

Please link to PR in backports, it helps a lot with review.

OK, looks good. It looks like there are more backports to do, at least https://github.com/bitcoin/bitcoin/pull/16849 (possibly others too).