Report non-mandatory script failures correctly

[sipa]

The same variable was reused for mandatory checking of script failures, so the reported error is always "No error". Use a new variable check2 so we don't overwrite the message produced by the first.

[jtimon]

This case only makes sense when in AcceptToMemoryPool anyway, no non-mandatory flgs will be used when called from connectBlock. We could move this special case to AcceptToMemoryPool as well.
Something like jtimon@dc0dab8 (although that commit is based on previous commits so it wouldn't be exactly like that if done alone).
Maybe we can fix this in that way (also helping libconsensus encapsulation by removing the STANDARD_NOT_MANDATORY_VERIFY_FLAGS [relay policy] dependency from CheckInputs() ).

Or, as always, "that can be done later".
Otherwise utACK, good catch.

[dcousens]

utACK

[maflcko]

Good catch! utACK 7ef8f3c

[petertodd]

utACK sipa@7ef8f3c

Could probably eventually split CheckInputs() into a function for value consistency, fees, etc., and a function that checks the scripts themselves. That'd make running the latter twice with different flags cheap.

[dcousens]

Actually just ran into this my self, tested ACK @ 7ef8f3c

[jonasschnelli]

utACK 7ef8f3c

[fanquake]

utACK

On Monday, 4 January 2016, Jonas Schnelli notifications@github.com wrote:

utACK 7ef8f3c
7ef8f3c

—
Reply to this email directly or view it on GitHub
#7276 (comment).

[laanwj]

utACK, this was sneaky

[jtimon]

@petertodd perhaphs is now time to reopen #6445 (there's a more up to date version after a big consensus moveonly in jtimon@babe715 )? I closed it because "those refactors/optimizations are going to interfere with the mempool limiting work", but those changes are done now (and it was trivial to rebase, meaning the concern wasn't really justified).