backport #839 (DefiniteDescriptorKey fixes) to 9.x #843
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Aside from not being future-proof at all, it's actually extremely hard to pin dependencies for this version of the library to get ones that work with 9.x. We have an ancient version of `bitcoind` in our deps, which in turn unconditionally depends on several crates which all have bad behavior (at least, the versions that we depend on). For example, both `zip` many versions and `bitcoind` itself have maximum `flate2` versions which are incompatible with each other. The latest version of `zip` 0.5.x removes the max version, but has an incorrectly stated minimum version of `thiserror`, which manifests in compilation errors because some trait impls aren't created by the `thiserror` macro. (And of course, you can't just use the latest `thiserror` because that does not compile with 1.41.1 for many reasons.) This is just one example. There are probably 10 such cases. Anyway now we have a lockfile and we can just forget about all this crap forever. As a secondary change: change all the `deny`s in lib.rs, except the one about unsafe code which seems likely to be future-proof, to `warn`s. The library doesn't compile on nightly otherwise, since e.g. the definition of `dead_code` has expanded. On my system I can now compile the library with both 1.41.1 and nightly 2025-06-23, with the same lockfile.
Sigh, after all the work I did to get a lockfile for the previous commit, it turns out that all the bitcoind tests were disabled anyway, with the explanation "breaking change in bitcoind". This commit deletes them entirely. We have no plans to fix them. If we ever want to regression test miniscript 9.x, we should do it externally by adding a fuzztest in master which pulls it in and tries to parse stuff. Then it will be actively exercised and maintained and not require ancient broken dependencies. This commit is all red EXCEPT a change to a test in src/miniscript/mod.rs. What's happening here is that we had one test (asserting on the string representation of errors from rust-bitcoin) which fails when the std feature is off. As it turns out, `bitcoind` unconditionally turns on that feature. So by having `bitcoind` as a dev-dependency it was impossible for us to ever test the nostd feature properly. Sigh.
As in other PRs, the meaningful changes here are: 'A'..'Z' becoming 'A'..='Z', and the partial_cmp replacements.
Remove the deprecated (well, renamed) "allow broken intradoc links" cfg flag, which exposes a ton of broken doc links. Mostly just uses of [] or <> that aren't in backticks. Fix them all. The one morally questionable change is to change two doclink references to Policy::enumerate_policy_tree to no longer be doclinks. The problem is that this function only exists when the `compiler` flag is on. In master in ebaa31b we just removed these references as well, but arguably we should have fixed them. Anyway in 9.x we're certainly not going to bend over backward to fix stuff like this.
Split off because I imagine these ones will be easier to review.
We have a `pub mod tests` module which is commented as being public "in case dependent libraries want to use these functions". However, it's also gated by #[cfg(test)]. I am not aware of any way that dependent libraries could compile *this* one with cfg(test) on. So this module was, in effect, never public. Make it non-pub to make things clearer, and also to silence all the "missing docs" warnings that appear when running clippy with --all-targets. This is NOT an API-breaking change even though it removes a public module, because the module was never actually usable.
sanket1729
approved these changes
Jul 5, 2025
apoelstra
force-pushed
the
2025-07_more_key-9.x
branch
from
f3bcfab to
5777a05
Compare
|
Ah, yep, I accidentally deleted a character in the unit test. |
Our current API allows constructing a DefiniteDescriptorKey with hardened steps, which means that you can't actually derive a public key. However, the point of this type is to provide a DescriptorPublicKey which implements the ToPublicKey trait. This is a backport, so it introduces the `has_hardened_steps` helper function but does *not* make it pub, in the interest of avoiding gratuitious API changes.
apoelstra
force-pushed
the
2025-07_more_key-9.x
branch
from
5777a05 to
d34188c
Compare
|
Does this also count as an ACK on #837? (That is, can I just close that and merge this one?) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Builds on #837