Add order reclaim command to the CLI by kaze-cow · Pull Request #206 · cowprotocol/solana-programs

kaze-cow · 2026-10-16T05:51:23Z

Description

Adds cow reclaim <order>..., which closes up to 15 order PDAs in one transaction and refunds each one's rent to the order's created_by. It also ships the CLI Docker image workflow (from #190, now triggered on main and tags only) and a cow --version flag.

Motivation

The devops reclaim scripts need to sweep closed orders from the CLI image. This is the first of three PRs split out of #204: order reclaim here, then buffer create, then buffer reclaim.

Considerations

All or nothing

Every input is resolved and checked (it has to be an order owned by the program) before anything is sent. If one fails, the CLI lists the offending inputs and sends nothing. If the transaction fails on-chain, the failing instruction index is mapped back to its input and the SettlementError is decoded.

Batch size

MAX_ORDERS = 15 is the worst case where every order has a distinct creator. max_orders_is_the_most_that_always_fit pins that against the 1232-byte packet limit, which is why bincode comes in as a dev-dependency.

`From<InstructionError> for SettlementError` function

Interestingly, there doesn't seems to be a good way to convert from a InstructionError to a SettlementError. So I added this to the interface!

How to test

cargo test -p cow-test-cli
cargo run -p cow-test-cli -- reclaim --help

🤖 Generated with Claude Code

Also ships the CLI Docker image workflow and a --version flag. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Malformed order PDAs/UIDs are now rejected at argument parsing, before any keypair load or RPC call. The PDA is derived from the UID afterwards, since a value parser can't see `--program-id`. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@glokos

# Description Brings the `cow` CLI Docker image and its `CLI image` workflow into `main`. This is all @glokos's work: it builds `ghcr.io/cowprotocol/cow-cli` for amd64 and arm64, then merges both into one manifest tagged with the sha and branch/tag. The image includes the Solana CLI and `bc` that the buffer-reclaim script needs. The first three commits are Assem's original commits from #190 and #205, cherry-picked unchanged. The last one drops the temporary `cli-reclaim-buffer` push trigger, which #190 said to remove once the image landed in `main`. ## Motivation #190 and #205 were merged into the reclaim stack's branches instead of `main`, so the image work now shows up inside #206/#207/#204 and makes them harder to review. Landing it here on its own means those PRs only show CLI changes once they're rebased. ## Considerations ### Credit Please **rebase-merge** this PR rather than squash it, so that Assem's commits land on `main` with him as the author. The trigger commit lists him as co-author too. ### Parity with the stack `.github/workflows/cli-image.yml` and `test-cli/Dockerfile` on this branch are byte-for-byte identical to the versions on `cli-reclaim-buffer` (#204), so rebasing the stack onto `main` afterwards should drop these files from its diff. ## How to test This PR should be the same as those already included in the buffer reclaim changes we eventually expect to merge: ``` git diff origin/cli-reclaim-buffer -- .github/workflows/cli-image.yml test-cli/Dockerfile # empty ``` **Once merged**, any branch that contains the workflow file can be built by hand, via Actions → CLI image → Run workflow (pick the branch under "Use workflow from") or: ``` gh workflow run cli-image.yml --ref <branch> ``` Unfortunately it doesn't seem to be possible to workflow dispatch until this is actually merged, but an example of this dockerfile generating a successful build though CI is [already on its actions page](https://github.com/cowprotocol/solana-programs/actions/workflows/cli-image.yml). The image is tagged `cow-cli:<branch>` (slashes become `-`) and `cow-cli:sha-<short>`. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Assem Hasna <assem@cow.fi> Co-authored-by: Assem Hasna <210342530+glokos@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Federico Giacon <58218759+fedgiac@users.noreply.github.com>

the problem is TryFrom sort of implies that, if the `TryFrom` succeeds, it was a valid settlement error. but if the error was emitted by an unrelated program, this may be misleading. So use an explicit function

fedgiac

Looks good, just nits.

fedgiac · 2026-10-21T07:19:37Z

+/// recipients couldn't be read, or the account is not owned by the settlement
+/// program.
+pub fn run(ctx: Context, args: ReclaimArgs) -> anyhow::Result<()> {
+    let orders = &args.orders;


Should we try to dedup the orders first? I believe that reclaiming two times the same order would lead to failure.

yea my mentality here is to leave the failures to the settlement program/solana runtime if they are going to happen. this seems like a rather specific error (there are many ways to specify invalid inputs here), so I would rather leave this out for the same reason as other errors in the input.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

fedgiac

Nice!

# Description Adds the `cow buffer` subcommand group with its shared `send` helper, and `cow buffer create <mint>...`, which creates the missing buffer PDAs of up to 14 mints in one transaction. Mints whose buffer already exists are skipped and reported. ## Motivation Second of the three PRs split out of #204, stacked on #206. This lays the groundwork that `buffer reclaim` (#204) builds on. ## Considerations ### One instruction per token program `CreateBuffers` serves a single token program, so mints are grouped and get one instruction per program. `MAX_MINTS = 14` is the worst case, where the mints span both programs. `max_mints_is_the_most_that_always_fit` pins it. ### Validation As with order reclaim, every mint is checked first: it has to exist, be owned by a supported token program and decode as a mint. If any fail, nothing is sent. ## How to test ``` cargo test -p cow-test-cli cargo run -p cow-test-cli -- buffer create --help ``` 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Federico Giacon <58218759+fedgiac@users.noreply.github.com>

# Description Adds `cow buffer reclaim <target> <buffer>[:<max burn>]...`, which closes up to 11 buffer PDAs in one transaction, burns whatever they hold up to a per-buffer limit, and sends their rent to `target`. The CLI image also gets the Solana CLI and `bc` (#205), which the devops buffer-reclaim script uses. ## Motivation This is used by the reclaim scripts on the devops side. It's the last of the three PRs split out of the original version of this one, stacked on #207 (buffer create), which is stacked on #206 (order reclaim). ## Considerations ### Older deployments Buffer and state PDAs are seeded with the `<major>.<minor>` release, so `--settlement-version` lets the CLI derive them for a deployment of an older release (e.g. `0.4`). The interface gains `settlement_seed_for_version` for this, and the CLI wraps it in `SettlementSeed`, which validates the version string. ### Burn limits A buffer holding more than its burn limit is left open and listed instead of failing the batch, since the program would skip it anyway. Any other bad buffer (wrong owner, not at its mint's PDA, not a token account) aborts the whole batch before anything is sent. ### Batch size `MAX_BUFFERS = 11` is the worst case: both token programs, and a reclaim authority signing separately from the payer. ## Refactor for Keypair Arg Its possible to have a key that only signs the transaction without actually holding any SOL. To allow for this. capability, a new `KeypairArg` parsable thing was added to clap and it allows for a file input which can then be resolved once the signer is needed. ## How to test ``` cargo test -p cow-test-cli -p cow-settlement-interface cargo run -p cow-test-cli -- buffer reclaim --help ``` 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Assem Hasna <210342530+glokos@users.noreply.github.com> Co-authored-by: Federico Giacon <58218759+fedgiac@users.noreply.github.com>

Add order reclaim command to the CLI

9fb7350

Also ships the CLI Docker image workflow and a --version flag. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

kaze-cow mentioned this pull request Oct 16, 2026

Add buffer create command to the CLI #207

Merged

kaze-cow requested a review from a team as a code owner October 16, 2026 05:51

kaze-cow mentioned this pull request Oct 16, 2026

Add buffer reclaim command to the CLI #204

Merged

kaze-cow added 2 commits October 16, 2026 17:06

refactor some unnecessary code or clean it up as needed.

1b77fc8

update parsing

ad58e27

kaze-cow commented Oct 16, 2026

View reviewed changes

Comment thread test-cli/src/main.rs

kaze-cow and others added 3 commits October 19, 2026 14:57

remove the unnecessary order reclaim argument cap

d3c53ec

bail on the first bad order in order reclaim

38f2670

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

kaze-cow mentioned this pull request Oct 19, 2026

Add the CLI Docker image and its build workflow #209

Merged

kaze-cow and others added 3 commits October 19, 2026 17:24

Merge remote-tracking branch 'origin/main' into cli-order-reclaim

ed323c3

move away from try from for instruction error

5f66a26

the problem is TryFrom sort of implies that, if the `TryFrom` succeeds, it was a valid settlement error. but if the error was emitted by an unrelated program, this may be misleading. So use an explicit function

Try to avoid separate parallel vectors

7b51b0d

fedgiac reviewed Oct 21, 2026

View reviewed changes

kaze-cow and others added 2 commits October 21, 2026 17:10

switch to using proptest for the parsing units

05c1411

Simplify print_failures to take string slices

4a89b05

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

kaze-cow requested a review from fedgiac October 21, 2026 08:20

fedgiac approved these changes Oct 21, 2026

View reviewed changes

kaze-cow merged commit ebe1584 into main Oct 21, 2026
18 checks passed

kaze-cow deleted the cli-order-reclaim branch October 21, 2026 08:23

Navigation Menu

Repository navigation

Add order reclaim command to the CLI - #206

Add order reclaim command to the CLI#206
kaze-cow merged 11 commits into
mainfrom
cli-order-reclaim

kaze-cow commented Oct 16, 2026 •

edited

Loading

Uh oh!

Uh oh!

fedgiac left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

fedgiac Oct 21, 2026

Uh oh!

kaze-cow Oct 21, 2026 •

edited

Loading

Uh oh!

Uh oh!

fedgiac left a comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

kaze-cow commented Oct 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Motivation

Considerations

All or nothing

Batch size

From<InstructionError> for SettlementError function

How to test

Uh oh!

Uh oh!

fedgiac left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

fedgiac Oct 21, 2026

Choose a reason for hiding this comment

Uh oh!

kaze-cow Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

fedgiac left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

kaze-cow commented Oct 16, 2026 •

edited

Loading

`From<InstructionError> for SettlementError` function

kaze-cow Oct 21, 2026 •

edited

Loading