Repository navigation
Conversation
The changelog had no entry for reading volumes from the mount table, the README's drive panel section did not say how Linux cards are found, and CONTRIBUTING's list of hardware fakes lacked the mountinfo fixture. Recount tests.
owenpkent
left a comment
There was a problem hiding this comment.
One P3 documentation defect: the new text says network mounts are excluded, although the implementation retains and classifies them. The exact contradiction and correction are inline.
Checked all three changed documents against the identical parent #19 volume implementation, its tests and the drive-panel consumer. The overlay-root fix and directory-scan fallback are described consistently; only the network/system-volume exclusion wording needs correction.
Validation: all 23 inline relative links resolve and git diff --check passes. Parent #19's exact-head volume tests passed locally with one symlink skip, and this head's CI succeeds.
Limits: documentation-only checks. Published test counts were not independently regenerated and no real Linux mounting experiment was performed.
Pinned scope: declared base 2454dca81fdbe821c7e2e8a6888570b0c3de782a, merge base 2454dca81fdbe821c7e2e8a6888570b0c3de782a, head 82e2bd624d863f44b9fc6033ac7fabe2d0a91b1a. Existing discussion and current-head checks were reviewed; no threads were resolved and no formal approval or merge is being issued.
The drive panel keeps NFS and SMB mounts as network volumes so they can be chosen as destinations; only pseudo and system filesystems are filtered out, and network volumes are kept away from card detection. The README and the changelog entry said network mounts stayed out of the list.
Stacked on #19. Merge that first, or merge this into it.
#19 changed how Linux volumes are found but left the docs describing the old behaviour by omission.
/proc/self/mountinfo, the overlay-root case, and the directory-scan fallback.linux_mountsfixture.Docs only. Suite on this branch: 408 passed, 8 skipped.
Review follow-ups (2026-10-11)