Add --mount/--davmount WebDAV auto-mount via a managed, verified rclone

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

docs/RESEARCH.md

@@ -66,7 +66,16 @@ removed afterwards. Rows marked "unmeasured" have no run behind them.
 | G9 | `parseSize` left-shifted without an overflow check: `--size 16777217t` wrapped to exactly `1.0T` and was accepted instead of rejected | check `v > (UINT64_MAX >> shift)` before shifting, reject instead of wrapping | **done in 1.3.2**, reproduced before the fix, re-run after: rejected | `parseSize` in `src/util.cpp` |
 | G10 | `newuidmap`/`newgidmap`/`slirp4netns`/`pasta` were pinned to a single hardcoded `/usr/bin/...` path, unlike `curl`/`tar`/`sha256sum`/`gpg` which go through `resolveBinary()`'s state-dir-override-then-fixed-dirs search; a distro installing them elsewhere would fail even though PATH-poisoning was never possible either way | route all four through `resolveBinary()`, same as the other helpers | **done in 1.3.2** | `idmap.cpp`, `network.cpp` |
 | G11 | Alpine's `latest-releases.yaml` and Debian's `index.json` (embedded base64 manifest) are read with fixed string/byte-offset search, not a YAML/JSON library; a publisher reformatting either file breaks extraction with a thrown error (fails closed) rather than silently misreading it | none: a dependency to parse two small, publisher-controlled files is a worse trade than failing loudly when their format changes, given the project's own "fewer dependencies" rule | **accepted, not fixed**; this row exists so the trade-off is written down, not rediscovered | `Catalog::resolve` in `src/image.cpp` |
-| G12 | No way to give a container storage that does not count against `--size` (remote/network storage, e.g. a WebDAV share); every writable path was either the host overlay (capped) or a private dir (capped) | `--mount SRC:DST[:ro]`: bind-mount an already-mounted host path into the container, outside the size watchdog's scan set (`Container::usedBytes` only scans `layers/`/`private/`, never `--mount` targets). doos itself never speaks a network filesystem protocol: mount WebDAV/NFS/etc. on the host first (e.g. `rclone mount`), doos only bind-mounts what is already there | **done in 1.5.0** (`--mount`, max 8 per container, changeable, applies at next start); missing `SRC` fails the start loudly | `Container::mounts`, `mountExtra` in `src/sandbox.cpp` |
+| G12 | No way to give a container storage that does not count against `--size` (remote/network storage, e.g. a WebDAV share); every writable path was either the host overlay (capped) or a private dir (capped) | `--mount SRC:DST[:ro]`: bind-mount an already-mounted host path into the container, outside the size watchdog's scan set (`Container::usedBytes` only scans `layers/`/`private/`, never `--mount` targets). doos itself never speaks a network filesystem protocol: a plain `--mount` bind needs the host path mounted already; see G13 for WebDAV specifically | **done in 1.5.0** (`--mount`, max 8 per container, changeable, applies at next start); missing `SRC` fails the start loudly | `Container::mounts`, `mountExtra` in `src/sandbox.cpp` |
+| G13 | G12's `--mount` still needed the user to mount WebDAV themselves (e.g. `rclone mount`) before pointing `--mount` at the result; no in-box path to remote storage | `--mount`/`--davmount` with a `dav://`, `davs://`, `http://` or `https://` source: doos downloads, GPG-verifies and installs a pinned `rclone` build under `$DOOS_HOME/tools/` (never the system, never PATH), then drives it on the host (on-the-fly connection string, no config file; detached double-fork daemon so the mount outlives the `attach` that created it) and bind-mounts the result exactly like a G12 bind. Deduplicated and reused by target (url+creds+vendor), not by destination | **done**, measured below under "davmount" | `src/davmount.{hpp,cpp}`, `DavMount::resolveAll` called from `Sandbox::start` |
+
+### davmount (G13, measured against a local `rclone serve webdav` fixture)
+
+- First use (fresh `$DOOS_HOME`, no managed rclone yet): `rclone-v1.75.2-linux-amd64.zip` (~20 MiB) downloaded, `SHA256SUMS` GPG-verified against the pinned fingerprint (`gpg` here accepts the DSA/SHA-1 signature rclone's release process still uses; it is a key-import plus `GOODSIG`/`VALIDSIG` fingerprint match, the same check as Ubuntu's image signature, not the web-of-trust warning gpg also prints), sha256 matched, unzipped, installed; the whole install plus the first mount completed in a few seconds and only ever happens once per `$DOOS_HOME`.
+- Reuse (second, unrelated container against the same target): measured at 41 ms and 45 ms for `attach -- true`, i.e. no remount cost at all; the existing FUSE mount is confirmed healthy (pid alive, mountinfo entry present, a real directory listing succeeds) and reused as is.
+- Write durability: `--vfs-cache-mode=writes` (rclone's own recommendation for WebDAV, which cannot stream) queues a write for upload after `--vfs-write-back` (shortened here to 1 s from rclone's 5 s default). Measured: a file written from inside a container appeared on the backend about 5 s after the *default* write-back elapsed in one run; with `--vfs-write-back=1s` the same write reached the backend within about 1-2 s of the writing process exiting, confirmed independent of the container's own lifetime (the container had already exited; the detached rclone daemon kept running and finished the upload on its own).
+- Failure classification, each reproduced against the same local fixture: wrong `--dav-pass` surfaced the server's `401 Unauthorized` and was reported as "authentication failed"; an unresolvable/unreachable address and a non-WebDAV endpoint (plain HTTP server, `501 Unsupported method ('PROPFIND')`) were reproduced in manual testing with the matching classifications. Every case failed within a few seconds, never the full 20 s readiness budget, because rclone mounts eagerly and only a real directory listing (not `stat` on the FUSE root, which always answers without touching the backend) forces the round trip that surfaces the error.
+- `doos davmount list`/`umount --yes`: verified showing a live target's id/vendor/url and state, dry-run output naming the path and warning it stops storage for any container still using it, and `--yes` actually unmounting (`fusermount3 -u`) and removing the managed directory.
 
 ### G1: memory, cpu and pids limits (cgroup v2)
 
@@ -272,7 +281,7 @@ delegated with `cpuset cpu io memory pids`). The end-to-end suite passed 83 chec
 
 ![Horizontal bar chart of CPU-seconds used by four busy loops in one container, without a limit and with --cpus 0.5. The limited container is highlighted. Values are in the sentence below.](charts/cpu.svg)
 
-The chart above is generated from `docs/charts/bench.json` (`make bench`, then `make charts`); it shows <!--bench:cpu_free-->15.6<!--/bench--> CPU-seconds without a limit and <!--bench:cpu_half-->2.4<!--/bench--> with `--cpus 0.5`; the startup and size charts live in [COMPARISON](../COMPARISON.md).
+The chart above is generated from `docs/charts/bench.json` (`make bench`, then `make charts`); it shows <!--bench:cpu_free-->16.3<!--/bench--> CPU-seconds without a limit and <!--bench:cpu_half-->2.3<!--/bench--> with `--cpus 0.5`; the startup and size charts live in [COMPARISON](../COMPARISON.md).
 
 ### What changed in behavior (and why it is not a loss)