I use testmempoolaccept for sanity-checking transactions I produce are valid in some app.
Wanted to propose a PR, but reading the original testmempoolaccept PR it seems to have some precedent, so asking here before about the conceptual validity : what do you think of making testmempoolaccept (accept a array which len >1 and) test each transaction passed in the array as if all valid precedent transactions of the array were into the mempool ?
(Edited after reading the doc more carefully..)
testmempoolaccept PR is #11742, and by precedent I meant that it was a lot of previous attempt to do this (see #11742 (comment) and the response) and apparently the scope of what it should “verify” / test. Maybe “precedent” was not the right term?
-chain or so, then they do.
The fact that MemPoolAccept::AcceptSingleTransaction exists and that the RPC accepts an array seems to suggest this is planned but I don’t know much about who’s working on/thinking about what. Is there ongoing work that would enable this as a side effect?
I think this would be very nice to have. Doing this for non-chained transactions would probably have little value (just make separate calls), but I definitely see why we’d want to test accept chained transactions.
I’m not 100% sure what needs to be done to implement this. My intuition is that this requires making a temporary copy of chainstate/mempool on which we apply each transaction so that we can use it for the ones that come after it. Also, won’t ordering matter? If a child comes before a parent in the array, should we make a temporary orphan pool too? Seems nontrivial 🤔
Leaning to concept nack, there are too many cases where mempool validity is not the thing you want to check (e.g., suppose an input has a 1 block relative sequence, such arrays will always fail).
Concept ACK on generally having an AnalyzePackage RPC to help a user figure out if all txns can go in within some time bound?