[BUG] Valid exFAT upcase table is rejected at U+FFFF #17
Labels
No labels
alpha-1
bug
ci
docs
duplicate
enhancement
help wanted
invalid
performance
phase-6
question
refactor
security
wontfix
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
erix/fatd#17
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
Fatd rejects a valid exFAT upcase table whose final word maps
U+FFFFto itself. This includes the recommended table produced bymkfs.exfat, so ordinary mount admission fails.Reproduction
Format the existing Integration 64 MiB regular-file fixture with
mkfs.exfat -L ERIXLIVE. Its 5,836-byte upcase table has checksumE619D30Dand ends with literalFFFF.An independent smaller table contains the mandatory first 128 mappings followed by
FFFF FF7F FFFF. The counted identity run ends atU+FFFE; the final word supplies the remaining mapping.Expected and actual behavior
Mount admission and filename conversion should accept the complete table. The original parser treats the final
FFFFas another compression marker, demands a nonexistent count and returnsINVALID_ARGUMENT. Interior markers and malformed runs must retain their existing checks.Environment and identity
084143d1f9ae6406d909240fdbf50e5d4da66d3e.6f20e691142054bfae65eaf52d0fa5c3788879f8; x86-64, one CPU, release image with development runtime.appliance-disk-image-realm-filesystem-exfat-positive.exfatprogs 1.4.3; original catalog and undo journal retained.Evidence and prior checks
Integration CI 2870 exits 41 instead of 33 before storage readiness. That coarse guest outcome does not identify a parser step. Separately, compiled controls against the original parser reproduce both final-literal refusals, while three companion controls pass.
The exFAT specification, section 7.2.5.1 specifies the recommended table and checksum. The repair covers that table and independently authored boundary cases; exact-source guest acceptance remains open.
Tracking
Implementation: PR #3. Phase scope: Integration #65.
The repair is published in 3a4b046c, and Integration 30580361 selects it in both catalogs. The original parser reproduces both valid-table refusals; the repaired provider passes all five boundary/table controls and the complete local matrix. Source and local criteria are checked. The original guest workload remains open.