[MarkFeder]

Linked issue

Fixes #787

Replaces #702, which the PR-hygiene bot closed before the issue was accepted; GitHub doesn't let the author reopen it.

What and why

The deploy npm scripts and cicd.sh in the native and Pinocchio examples ran:

solana program deploy ./program/target/so/program.so

cargo build-sbf --sbf-out-dir=./program/target/so names the artifact after the crate, not program (e.g. transfer_tokens_program.so), so the documented deploy step failed with "No such file".

Every deploy command now names the file the build actually writes, <crate lib name>.so, taken from cargo metadata and cross-checked against the fixture each example's tests load:

solana program deploy ./program/target/so/token_2022_transfer_fee_pinocchio_program.so

A *.so glob would be shorter, but pnpm runs package.json scripts through cmd.exe on Windows, which passes the glob through literally, so the explicit name is the only form that works on every OS.

Also fixed while checking every deploy path against its crate:

• Seven native cicd.sh built without --sbf-out-dir and deployed ./program/target/deploy/program.so, which a workspace member never produces. They now build into ./program/target/so like the rest.
• Four token-2022 native package.json deployed hello_solana_program.so (copy-paste).
• Seven Pinocchio cicd.sh used the deprecated --bpf-out-dir; Anza v4's cargo build-sbf only documents --sbf-out-dir.
• basics/cross-program-invocation/native deployed hand.so/lever.so (cicd.sh) and program.so (package.json). Both now deploy the two crate-named artifacts; the npm script chains them with &&, which works in cmd.exe too.
• cicd.sh gets set -e (and pipefail where it pipes into grep), so a failed build stops before deploying.
• basics/counter/native/README.md said pnpm deploy, which is pnpm's own workspace command, not the script; it now says pnpm run deploy.

basics/transfer-sol/asm builds with sbpf and is left alone.

Testing

• A script compared every solana program deploy path in native/Pinocchio package.json and cicd.sh (96 commands) with <--sbf-out-dir>/<crate lib name>.so from cargo metadata: 0 mismatches.
• In a Linux container, for all 48 native/Pinocchio examples with a cicd.sh: ran bash cicd.sh (real cargo build-sbf), then pnpm run deploy, with a stand-in solana that only accepts program deploy <path> when the path is an SBF ELF the build produced. The same harness on current main fails on the broken examples, so it does catch them.
• On Windows, a package.json script under pnpm 10.33.0 receives ./program/target/so/*.so literally (file not found), while the explicit path and the && chain resolve.
• prettier --check . passes. CI doesn't run deploy, so it is otherwise unaffected.

AI disclosure

Check exactly one. See CONTRIBUTING.md.

• No AI tooling was used beyond editor autocomplete.
• AI tooling was used. Tool and extent: Claude Code helped make the edits, build the verification scripts and write this description; I reviewed the changes.

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no actionable issues remain.

Summary

The PR points deployment commands at the actual crate-named build files.

• Example deploy commands now point to their crate-named program files.
• The cross-program example builds and deploys its hand and lever programs separately.

The earlier, unnumbered finding about deploying old builds is fixed: the affected scripts now stop before deployment when a build fails. No new actionable issues were found.

_{Reviews (3) · Last reviewed commit: "chore: deploy the real artifact name so ..." · Reviewed by Greptile}