[sebastianst]

Why

revm-precompile 42 requires sha2 = "0.11". SP1 accelerates SHA-256 by replacing the sha2
crate through [patch.crates-io], and a patch only applies to the exact version it names — so
patch-sha2-0.10.9-sp1-6.2.0 does nothing for a graph that resolves sha2 to 0.11. Programs that
pull in revm 42 silently drop to the soft backend and stop emitting SHA_EXTEND / SHA_COMPRESS.

The 0.11 patch itself landed as sp1-patches/RustCrypto-hashes#10 and is
tagged patch-sha2-0.11.0-sp1-6.2.0. This PR adds the patch-testing coverage for it, so 0.11 is
validated the same way every other sha2 version already is. Tracking issue on our side:
ethereum-optimism/optimism#22890.

What

A sha2_v0_11_0 case mirroring sha2_v0_10_9:

• sha/program/Cargo.toml: optional sha2-v0-11-0 dep, v0-11-0 feature, [[bin]], and the
  [patch.crates-io] entry that swaps in the patched crate for the guest.
• sha/program/bin/sha2.rs: the extern crate … as sha2 alias behind v0-11-0.
• sha/build.rs: build the guest for the new feature.
• sha/src/lib.rs: #[sp1_test("sha2_v0_11_0", syscalls = [SHA_COMPRESS, SHA_EXTEND], gpu, prove)].
• Cargo.toml / build-host/Cargo.toml: sha2-v0-11-0-patched, so the patch is also compiled for
  the host target alongside the other patched crates.

Both patch sources point at the released patch-sha2-0.11.0-sp1-6.2.0 tag, written the same way as
the neighbouring 0.10.9 entries.

sha2 0.11 moves to digest 0.11, but the surface this test uses is unchanged — Digest::new,
update, and finalize().into() to [u8; 32] all still work — so bin/sha2.rs needed nothing
beyond the new alias, and the host-side expected-digest helper is shared verbatim with the 0.10.9
case.

Validation

Run locally against this branch, SP1 toolchain from sp1up (cargo-prove sp1 58c4aea, succinct
rustc 1.96.0-dev), CPU executor, no proving:

cargo test --manifest-path patch-testing/Cargo.toml -p sha --release -- --test-threads=1

6 passed, 0 failed, as does cargo check -p build-host --release. The new case emits both
precompiles and matches the host digests (cycle counts vary slightly run to run, since part of the
corpus is random):

case	cycles
test_sha2_v0_11_0_expected_digest_lte_100_times	1,336,295
test_sha2_v0_10_9_expected_digest_lte_100_times	1,347,444

An extra ad-hoc run over a corpus covering the empty input, one-block, multi-block, and inputs
sliced at odd offsets (to hit the unaligned-first-block branch of the patch) gave byte-identical
digests to host sha2 0.11 and identical syscall counts for 0.11 and 0.10.9:
SHA_COMPRESS = 117, SHA_EXTEND = 117.

🤖 Co-created with Claude Fable 5.1

[fakedev9999]

lgtm