Rust: store replace on Windows tries POSIX-semantics rename first (store-windows-rs) #148

Merged

libby merged 6 commits from store-replace/windows-rs into main 2026-11-13 01:29:21 +00:00

Conversation 0 Commits 6 Files changed 2 +297 -51

libby commented 2026-11-12 21:58:28 +00:00

Member

Copy link

Cut variants:cut_spec:cut-store-windows-rs.r1, ruling variants:ruling:windows-replace-posix-first (operator, option c). Reports cut-store-windows-rs.h1 to .h3; Soul verdicts .s1 to .s3.

replace_with implements the ruled decision table over a ReplaceKernel seam (POSIX rename first; codes 1/50/87 fall back to write-through MoveFileExW for the rest of the call; 5/32/33 retry 25 x 20 ms; other codes error), tested with a fake kernel on Linux. The inline MoveFileExW loop is deleted.

Verified: Linux cargo test --locked; cargo check --locked --tests --target x86_64-pc-windows-gnu compiles the Windows FFI body and windows_replace_tests. Soul's Starfire probe confirmed the kernel behaviour the new delete-sharing-reader test relies on (POSIX rename returns 0, MoveFileExW returns 5).

Not verified: windows_replace_tests have never run on Windows (no Rust host on starfire), and the durability of the POSIX rename (no write-through flag) is unmeasured. Follow-up variants:follow_up:windows-replace-run-and-durability.

🤖 Generated with Claude Code

Cut `variants:cut_spec:cut-store-windows-rs.r1`, ruling `variants:ruling:windows-replace-posix-first` (operator, option c). Reports `cut-store-windows-rs.h1` to `.h3`; Soul verdicts `.s1` to `.s3`. `replace_with` implements the ruled decision table over a `ReplaceKernel` seam (POSIX rename first; codes 1/50/87 fall back to write-through MoveFileExW for the rest of the call; 5/32/33 retry 25 x 20 ms; other codes error), tested with a fake kernel on Linux. The inline MoveFileExW loop is deleted. **Verified:** Linux `cargo test --locked`; `cargo check --locked --tests --target x86_64-pc-windows-gnu` compiles the Windows FFI body and `windows_replace_tests`. Soul's Starfire probe confirmed the kernel behaviour the new delete-sharing-reader test relies on (POSIX rename returns 0, MoveFileExW returns 5). **Not verified:** `windows_replace_tests` have never run on Windows (no Rust host on starfire), and the durability of the POSIX rename (no write-through flag) is unmeasured. Follow-up `variants:follow_up:windows-replace-run-and-durability`. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

libby added 1 commit 2026-11-12 21:58:28 +00:00

cultcache-rs: Windows store replace tries a POSIX-semantics rename first through a ReplaceKernel seam ... 531ff6755b

replace_with owns the decision table (POSIX first; 1/50/87 fall back to
MoveFileExW write-through for the rest of the call; 5/32/33 retry 25 x 20 ms;
else error), tested with a fake kernel on every OS. WindowsKernel holds only the
two syscalls. The inline MoveFileExW loop is deleted.
(variants:ruling:windows-replace-posix-first)

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

libby added 1 commit 2026-11-12 22:15:38 +00:00

cultcache-rs: import FILE_RENAME_FLAG_* from Win32::System::WindowsProgramming, where windows-sys 0.59 defines them ... 7ed1647375

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

libby added 1 commit 2026-11-12 22:19:11 +00:00

cultcache-rs: the POSIX rename passes an absolute destination (FileName without RootDirectory must be a full path) ... 6094a7dba3

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

libby added 1 commit 2026-11-12 22:26:57 +00:00

cultcache-rs: revert the absolute() destination (a relative FileName works with a null RootDirectory); a delete-sharing reader test discriminates POSIX rename; doc names the ruling and where the table has landed ... 1d2ee73ce8

Soul's Starfire probe (build 26200) showed a relative FileName already works, so the reverted
claim was false. windows_replace_tests are compile-checked only and have never run on Windows.

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

libby added 1 commit 2026-11-12 22:33:08 +00:00

cultcache-rs: the Windows parent-sync comment no longer credits MoveFileExW alone with durability ... 7acc51eebb

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

libby added 1 commit 2026-11-12 23:06:25 +00:00

cultcache-rs: the replace doc says absence to a lock-free reader is rare on NTFS, not impossible ... b402757b27

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

libby force-pushed store-replace/windows-rs from b402757b27 to d265b7d647 2026-11-13 00:02:14 +00:00 Compare

libby merged commit 845c024379 into main 2026-11-13 01:29:21 +00:00

libby referenced this pull request from a commit 2026-11-13 01:29:23 +00:00

Rust: store replace on Windows tries POSIX-semantics rename first (store-windows-rs) (#148)

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels

No items

No labels

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

GameCult/CultLib!148

No description provided.