Remove rt-multi-thread dependency from AFC file descriptor drop#93
Merged
Merged
Conversation
AFC's close-on-drop used block_in_place + block_on to synchronously send a FileClose packet, requiring tokio's rt-multi-thread feature. This is heavyweight for a best-effort cleanup that already did nothing on wasm and single-threaded runtimes. Replace with a simple no-op drop that warns (debug_assert + println) if .close().await wasn't called. The device reclaims FDs when the AFC session ends regardless. Also fix the afc tool to explicitly close file descriptors after use.
Owner
|
I know @abdullah-albanna and I discussed this when adding it in the discord. Any thoughts? |
Owner
|
I've given it some thought, and I think adding the debug assert is a fine trade-off. The drop was added to catch dumb mistakes, so this is fine. |
maxslarsson
added a commit
to maxslarsson/idevice
that referenced
this pull request
May 7, 2026
…xson#93) AFC's close-on-drop used block_in_place + block_on to synchronously send a FileClose packet, requiring tokio's rt-multi-thread feature. This is heavyweight for a best-effort cleanup that already did nothing on wasm and single-threaded runtimes. Replace with a simple no-op drop that warns (debug_assert + println) if .close().await wasn't called. The device reclaims FDs when the AFC session ends regardless. Also fix the afc tool to explicitly close file descriptors after use.
jkcoxson
pushed a commit
that referenced
this pull request
May 8, 2026
…gClient (#96) * refactor: remove pairing_file field and state param from RemotePairingClient Move pairing_file from a struct field (&'a mut borrow) to a parameter on connect() and its callees. This removes the lifetime parameter from the struct and scopes the mutable borrow to the function call. Also remove the unused state: S generic from connect/pair/request_pair_consent, since every caller passed 0u8 and ignored it. Callers can capture state via move closures instead. * fix: feature gating for error conversions and dead deps (#95) - Add internal _serde_json and _reqwest features to gate From impls - Remove unused json and byteorder optional dependencies - Add missing xpc dependency to remote_pairing feature - Simplify map_err workarounds in tunnel code to use ? directly * fix: make openssl PairingFile escrow_bag field Optional (#97) The openssl variant of PairingFile had escrow_bag as Vec<u8> while RawPairingFile has it as Option<Data> (it's None on Apple Watch). This caused compile errors in the From/TryFrom impls when building with the openssl feature. The rustls variant already had the correct Option<Vec<u8>> type. * fix: improve openssl crypto backend (#98) * fix: make openssl crypto path a first-class backend Un-gate `ca.rs` and `pair()` from rustls — they only use pure Rust crates (rsa, x509-cert, sha2) so pairing now works with openssl. Always detect legacy devices instead of only with openssl feature. Add openssl feature to FFI crate. Clean up SNI string and gate rustls-only PEM helpers to fix dead code warnings. * refactor: return legacy flag from start_session, remove duplicate detection `LockdownClient::start_session` already queries ProductVersion to detect legacy devices. Four callers were making the same query beforehand. Now `start_session` returns the legacy bool so callers reuse it instead of round-tripping to the device twice. * Remove rt-multi-thread dependency from AFC file descriptor drop (#93) AFC's close-on-drop used block_in_place + block_on to synchronously send a FileClose packet, requiring tokio's rt-multi-thread feature. This is heavyweight for a best-effort cleanup that already did nothing on wasm and single-threaded runtimes. Replace with a simple no-op drop that warns (debug_assert + println) if .close().await wasn't called. The device reclaims FDs when the AFC session ends regardless. Also fix the afc tool to explicitly close file descriptors after use.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Hey! Love the project -- been using it and it's been great.
I ran into an issue where
rt-multi-threadgets pulled in as a transitive dependency just by depending onidevice. I really don't need, want, or can afford draggingrt-multi-threadinto my project. It also feels a bit odd for a library to pull in a runtime opinion like that, especially since it's only used in one place:block_in_placewhen closing an AFC file descriptor on drop, specifically when running on a multi-threaded runtime. And yetrt-multi-threadis pulled in even if you don't use AFC (which I don't) nor a multi-threaded runtime (which I'm not I'm using single-threaded).Since there's no async
Dropin Rust, there isn't a clean way to handle this. One option I explored was pulling inrtinstead ofrt-multi-thread(which feels a lot more reasonable but still a little weird for a library) and usingcrate::spawnto fire-and-forget the close packet instead ofblock_in_place. The upside is it works on any runtime flavour including single-threaded and wasm. The downside is the close becomes fire-and-forget -- you lose the guarantee that the device ack'd the close before theAfcClientis dropped, and if the runtime is shutting down the spawned task might get cancelled.But on wasm and single-threaded runtimes, the file descriptor was already just silently dropped without closing. So I figured it doesn't hurt that bad to just drop the file without sending the close, and require callers to explicitly call
.close().await. I added adebug_assert!so it panics in debug mode if someone forgets, and it also prints an error to stdout (matching the existing error printing pattern on line 392 of the same file). I also went through all the existing call sites in the repo (tools, tests, crashreportcopymobile) and added explicit.close().awaitcalls.Another option worth considering: we could add a
multi-threadcrate feature that enablesrt-multi-threadand brings back the defensiveblock_in_placeclose-on-drop behaviour for users who want that safety net and are already running on a multi-threaded runtime.Let me know what you think or if you'd like to see any changes -- I'd love to get
rt-multi-threadremoved as a hard dependency!