--- Log opened Thu Feb 17 00:00:01 2022
03:59 < darosior> This branch is the signing PR rebased on the 2 first PRs with the latest modifications (but your signing algorithm simplification which iirc you haven't shared): https://github.com/darosior/bitcoin/commits/miniscript_random_fuzz. It contains in order the introduction of the random fuzz target, your modification to make it non-recursive, and my
03:59 < darosior> re-modification on top making it strictly encoded
06:34 < darosior> I ran a coverage report of the corpus for miniscript_random, and it's not great. But i haven't left it run very long too. When i find the time i might code an encoder for the former random tests to seed the corpus.
09:21 < darosior> Thanks for the commit review, you found the cause of this ^
11:56 -!- arnab [~arnab@2401:4900:3bed:8274:d1db:b2dc:307e:64d3] has joined ##miniscript
11:56 -!- arnab [~arnab@2401:4900:3bed:8274:d1db:b2dc:307e:64d3] has quit [Client Quit]
11:57 -!- arnab [~arnab@2401:4900:3bed:8274:d1db:b2dc:307e:64d3] has joined ##miniscript
11:58 -!- arnab [~arnab@2401:4900:3bed:8274:d1db:b2dc:307e:64d3] has quit [Client Quit]
12:30 < andytoshi> sanket1729: in https://gist.github.com/sanket1729/b1ed8ed776c7ac798d5e33e40f5d996d you say that rawtr has an `explicit_script` but i think this isn't true
12:31 < andytoshi> this is the only row where tr and rawtr differ
12:32 < andytoshi> in general, the two "obvious" separations i see here are (a) "single-key" vs "script-containing" (which taproot kinda blurs), and (b) "pre-taproot" vs "taproot"
12:32 < andytoshi> you suggest addressable vs non-addressable, and solvable vs non-solvable, but i think these aren't good categorizations for the reasons you say in the doc
12:34 < andytoshi> there's something of an argument for "partial" vs "full", but i don't think it would work well in practice ... like, it's not clear to me what the difference in API would actually be
12:35 <@sipa> I think the justification for having partial descriptors in the first place is exactly that there are already many instances of practically partial/unsolvable descriptors, which cannot be distinguished from others.
12:35 < andytoshi> agreed
12:35 <@sipa> (like you're using a derived public key, but not providing origin info for it... a signer might be unable to sign in that case)
12:36 <@sipa> but there may also be cases where that key isn't derived from anything else
12:36 < andytoshi> yep
12:36 <@sipa> and hey, raw(51) is arguably solvable :p
12:37 < andytoshi> :)
12:37 < andytoshi> yeah, arguably we couldc do something smart with raw(<script that is valid miniscript>)
12:38 <@sipa> agreed (though i'm strongly opposed to actually doing that)
12:38 < andytoshi> e.g. we could try to parse a script as miniscript -- if it succeeds, that's a "bare" descriptor, if not it's a "raw" one ... and both would serialize as raw(...)
12:38 < andytoshi> ok :)
12:38 < andytoshi> i don't have very strong feelings about it. it feels dangerous but i can't put my finger on why
12:39 <@sipa> i guess there could be sort of a "smarten" transformation on descriptors, which could try to extract structure which is implicit
12:39 <@sipa> e.g. it could convert a raw() that actually contains valid miniscript to a corresponding miniscript descriptor
12:41 < andytoshi> yeah, i think in our API we could put a special method on the raw type for this
12:41 < andytoshi> but i agree it should not happen automatically or silently, that would cause weird surprising behavior
12:41 <@sipa> indeed
12:41 < andytoshi> e.g. somebody who desn't know anything about miniscript and just creates raw scripts, finds our "magic library" that can just deal with it
12:42 < andytoshi> then later adds one opcode somewhere we don't expect and everything breaks completely
12:42 <@sipa> there are also no bounds to it... what if you manage to hardcode the P2SH address of OP_TRUE, and then turn addr(that address) into sh(1).
12:43 < andytoshi> yeah, that makes sense
15:03 <@sipa> darosior: I have something that seems to work (merging the malleable and non-malleable signer), letting the fuzz tester run on 16 cores overnight to make sure.
15:04 <@sipa> Then I'll work on integrating it into your latest branch.
15:52 < sanket1729> andytoshi: I can see how you can say that rawtr does not have a explicit/inner script. But it's partial so cannot know. I guess erroring is the correct way
22:14 -!- meshcollider [meshcollid@meshcollider.jujube.ircnow.org] has quit [Ping timeout: 252 seconds]
22:15 -!- meshcollider [meshcollid@jujube.ircnow.org] has joined ##miniscript
--- Log closed Fri Feb 18 00:00:02 2022