[gdmaster3]

Motivation

Downstream wallets that embed LDK together with SQLCipher (e.g. a mobile wallet using bundled-sqlcipher-vendored-openssl) need exactly one libsqlite3-sys in the final binary. Cargo's links = "sqlite3" rule makes that impossible unless every rusqlite in the graph resolves to the same sqlite3-sys generation, and a rusqlite 0.31 (v0.7.x) / rusqlite 0.37 (downstream pin) pair cannot co-resolve.

Full disclosure: main already loosened the requirement to >=0.31.0, <0.40.0 (Cargo.toml:109-110, from #1187), so this is not a hot-fix. It is a backport-willing maintenance contribution so the v0.7.x line can interoperate, together with the two dev-dependency cleanups that a rusqlite 0.37 resolution exposes.

What changed

Branch sw/rusqlite-0.37, based on tag v0.7.0, 5 commits:

• aef479a — Cargo.toml: rusqlite 0.31.0 -> 0.37.0 (bundled feature unchanged).
• 17be6be, e02a484 — Connection::pragma call sites: Some(rusqlite::DatabaseName::Main) -> Some("main") in src/io/sqlite_store/migrations.rs (2 sites) and src/io/sqlite_store/mod.rs (1 site). rusqlite 0.37 removed DatabaseName; the first parameter is now schema_name: Option<&str>.
• 26ba14b — drops [target.'cfg(vss_test)'.dev-dependencies] ldk-node-062 = { package = "ldk-node", version = "=0.6.2" }. That dev-dep pulls a second libsqlite3-sys into resolution (not merely the build), which is what breaks downstream unification. It is only used by the cfg-disabled VSS backwards-compat test that needs a live VSS server.
• d82a50e — [target.'cfg(not(no_download))'.dev-dependencies]: electrsd features reduced to ["legacy"], matching the existing cfg(no_download) entry. The esplora_* / corepc-node_* build scripts download release artifacts and hash-assert them; they panicked on upstream artifact drift here. This is the same effect as running with --cfg no_download, so the integration suites can still run with PATH-provided binaries.

Evidence

Downstream gate run (workspace build, clippy -D warnings, cargo tree single-copy assertion, cargo deny licenses, SQLCipher cipher test):
https://github.com/gdmaster3/SuperWallet/actions/runs/42470100999

Plus this fork's own cargo test --lib io::sqlite_store — PASS under the bumped pin (the module touched by the pragma change).

Caveats

• VSS cfg(vss_test) dev-dep parity: removed rather than re-vendored. If you want it kept, I can re-add it as an explicitly non-default compat path (e.g. a path-patched ldk-node 0.6.2 carrying its own rusqlite bump); it should not remain a plain version dependency, since that alone breaks links = "sqlite3" unification downstream.
• The electrsd / corepc-node edit is deliberately inert rather than a version bump: the observed failure was build-script artifact hash drift, not a compile error.

Notes for review

• Targeted at release/0.7 (this branch is based on tag v0.7.0, so the PR diff is the 5 commits / 3 files). main already carries the equivalent modernizations (pragma_update, legacy-only electrsd, no ldk-node-062 resolution dep), so none of this is needed there.
• Connection::pragma(Some("main"), ...) is the 0.37 signature (Option<&str>): https://docs.rs/rusqlite/0.37.0/rusqlite/struct.Connection.html#method.pragma
• No CONTRIBUTING.md or PR template exists in this repository (only .github/workflows), so no stated title/CLA convention applies; commit messages follow the existing build: / ci: prefixes.
• Happy to split this into separate PRs (bump vs. dev-dep cleanups) or to drop the dev-dep commits entirely if they are unwanted scope.

[ldk-reviews-bot]

I've assigned @tnull as a reviewer!
I'll wait for their review and will help manage the review process.
Once they submit their review, I'll check if a second reviewer would be helpful.

[tnull]

No, there is a reason we did this way in #1187. We can't just require a newer SQLite, which a) would be incompatible with other crates in the rust-bitcoin ecosystem b) would technically be a SemVer violation if done in a patch release.

Superseded by #1225