A rather interesting point brought up by these comments. When SigOps counting is removed, we should think carefully about this stuff (from what I can tell on first glance, it should be fine to move all SigOps counting from CheckBlock to ContextualCheckBlock, but need to think very carefully about this stuff).

Merge #11747: Fix: Open files read only if requested

OK tried out the test plan on commit before and after D3564 laned -- yes it did depend on D3564 to pass :-)

Anyone feel free to commandeer if they want to do similar.

Interesting, subtle use-after-scope there...

Is there a better way to do those else clauses? Otherwise all good...

I think these changes to Qt do deserve a release note. Since this is the final one for now (it looks like) I think this Diff would be a good place for the release note to be added.

Looks good I think ... this should change nothing of the behaviour when run with default settings (cashaddr on), right?

How can I test this -- just make check? Also, if rebased on the new master it shouldn't have this unrelated test error.

From the test plan, sounds like this depends on D3564? Or am I misunderstanding...

In D3540#83565, @deadalnix wrote:

Core has soem code doing this. cmake needs to be kept in sync with automake.

Even something so simple makes me nervous but note there are tests for STACK_SIZE error (and for max-sized stacks) in script_tests.json.

This is mostly an experiment to see what kind of optimization is possible with just a little work. Would be better if we could use TESTS= but my automake-fu is not good enough for that.

Fabien notes that the test summary actually prints out one test: the bitcoin-qt tests. For me it was 0 because I was using a non-qt build.

rebase (I rebase-conflicted myself with D3527, which (in the original PR) removes the "Verify valid transaction" test case added here... :facepalm:)

Note that if accepted, this (and the trivial fix mentioned) would bring us up to sync with Core regarding the way tests are done. All other changes in this makefile seem to be regarding fuzzing.

Concept ACK!

In D3537#83384, @deadalnix wrote:

Please remove travis altogether if you are going that road.

In D3528#83347, @deadalnix wrote:

The description is fubared

If we take this, we should also do this fix for OpenBSD compat (why not): https://github.com/bitcoin/bitcoin/pull/13355

This is pretty neat but it still has some issues.

(Similar to D3533 but this actually does result in a test failure if run individually)

I don't think this makes any difference for us, but see also:
https://github.com/bitcoin/bitcoin/pull/12150
https://github.com/bitcoin/bitcoin/pull/12408

rebase again :/

just rebase

minor changes

Note that this removes -whitelist , and adds this expect_disconnect feature. This is will let us do nicer tests of transaction banning/non-banning for mandatory flags.

Note to self: also backport #13048 after this

Self review the next morning shows only minor issues. Pointing out some of the changes that happened due to CTOR and the padding rule below, and some minor things to change...

shelving for now, until parents are updated

Dunno why it says "These changes have failed to build."

I still think "Bitcoin ABC Logo" is more appropriate (alt text is a substitute for the image), but OK :-)

This was, hands down, the most enjoyable backport ever.

fix lint-induced syntax errors and remove lint-identified unused imports

linter introduced syntax errors

fix

This is not complete, I am seeing some weird behaviour.