fix(deps): update rust crate mtp-rs to 0.25.0

[renovate]

This PR contains the following updates:

Package	Type	Update	Change
mtp-rs	dependencies	minor	0.23.0 → 0.25.0

---

Release Notes

vdavid/mtp-rs (mtp-rs)

v0.25.0

Compare Source

Library 0.25.0, CLI 0.7.0. Diagnostics so device-specific bugs can be captured by the reporter instead of reproduced on identical hardware (follow-up to the #​18 Samsung cancel-wedge work in 0.24.0).

Added

• [lib] Opt-in tracing feature for device-protocol diagnostics. Off by default (no dependency, no cost); enable it and install any tracing subscriber to emit events on the transaction and cancel/reset paths. Purpose-built so a bug reporter can capture what their device actually does (issue #​18) instead of us reproducing on identical hardware. The cancel/reset path logs at debug, per-operation execution at trace.
• [cli] --trace flag and richer doctor. mtp-rs --trace <cmd> prints the cancel/reset diagnostics to stderr (RUST_LOG overrides for finer control, e.g. mtp_rs=trace for per-operation detail). doctor now prints device capabilities, and doctor --probe-cancel runs a cancel-health check (download a file, cancel mid-stream) that classifies the device as healthy, wedged_recovered (#​18), or errored — the one-command artifact to attach to a freeze report.
• [lib] force_cancel_wedge(serial) virtual-device test hook (feature virtual-device). Arms a one-shot so the next cancel_transfer returns Error::DeviceReset, letting the #​18 cancel-wedge contract be regression-tested with no hardware.

v0.24.0

Compare Source

Library 0.24.0, CLI 0.6.0.

Fixed

• [lib] Recover from the Samsung "large-backlog cancel" wedge instead of hanging (#​18). Cancelling a held-open streaming download while the device still had a large bulk backlog queued (classically the first download right after a fresh USB connect) wedged the PTP session on Samsung phones (reproduced on a Galaxy S23 Ultra): the drain read the whole backlog and idled out without ever seeing the closing Response container, the device stopped answering, and cancel() returned a false success so the consumer's next call hung, looking like a mid-transfer freeze. cancel_transfer now detects the wedge (GET_DEVICE_STATUS timing out, distinct from an unsupported-op stall), issues a session-less USB DEVICE_RESET to un-stick the transport, and returns the new Error::DeviceReset instead of a false success. It deliberately does not reopen: post-reset these devices need a span of quiet to tear the old session down, so reopen is the caller's job (drop the device, wait a few seconds, reopen with idle-spaced backoff; hammering re-wedges it). download_windowed avoids the path entirely (no multi-MB backlog to cancel). Reported by @​qarmin.
• [lib] Ranged, windowed, and resumable downloads now work on devices that only advertise the 32-bit GetPartialObject. Previously download with a ByteRange::From/Range, download_windowed, and read_range all hard-required GetPartialObject64 (0x95C1), the 64-bit-offset op. Many PTP cameras (e.g. the Panasonic Lumix DMC-TZ61, #​12) only advertise the 32-bit GetPartialObject (0x101B), so these calls failed with Unsupported. The backend now falls back to the 32-bit op for any offset that fits in u32 (files up to 4 GiB); a resume past 4 GiB still needs the 64-bit op (returns Error::InvalidData), and a device with neither op returns Error::Unsupported. The op selection (plan_partial_read) is unit-tested, and the 32-bit path is covered end-to-end against a virtual device configured without the 64-bit op.

Changed

• [lib] Breaking: new PtpError::DeviceReset variant. The low-level ptp::PtpError enum is not #[non_exhaustive], so code that exhaustively matches it must add an arm. The neutral mtp::Error gains the same DeviceReset variant but is #[non_exhaustive], so matching it with a wildcard arm is unaffected. See the #​18 fix above for what the variant means.
• [workspace] Real-device debugging hub and test-harness hardening. docs/debugging.md is now a debugging hub (macOS ptpcamerad blocker, software-reset recovery, fast-fail timeouts, Samsung/Android gotchas), linked from AGENTS.md. Integration tests read an MTP_TEST_TIMEOUT_SECS override (default 30) so a wedged or absent device skips fast instead of stalling, and an opened device reporting zero storages (a half-authorized phone) now skips cleanly instead of panicking.
• [lib] Breaking (test support): VirtualDeviceConfig gains a supports_partial_object_64: bool field (feature virtual-device). Set it true to keep the previous behavior; set it false to model a camera that only implements the 32-bit GetPartialObject and exercise the fallback above. Only affects code that constructs VirtualDeviceConfig directly (test setups).
• [workspace] Opt-in integration-test env flags now test the value, not mere presence. MTP_RUN_SLOW_TESTS=0 and MTP_RUN_DROP_RECOVERY=0 used to enable the gated test (a bare "is the var defined" check); they now correctly mean "off". Only 1/true/yes/on enable. Reported by @​juleskers in #​12.

---

Configuration

📅 Schedule: (in timezone Europe/Budapest)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Enabled.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.