texttopdf: comment token across a page boundary writes before the page buffer

Summary

cfFilterTextToPDF() writes two bytes before its heap-allocated page buffer
when C/C++ pretty printing recognizes /* at the exact automatic page
boundary. The syntax state is updated before wrapping, but the first comment
character is highlighted after wrapping with column - 1, when column has
already been reset to zero.

The failure is reachable through the normal standalone texttopdf filter
without a fuzzing harness, a PPD, or malformed text.

Reproduction

Final upstream recheck on 2026-08-02: OpenPrinting/cups-filters 11d1a190530f85a361a1b3835f57d635256dff1a, libcupsfilters 905fd94fb22a9298bc8e2c845eb7e04f7055b686, and libppd fc41539f761286396a7df8aeeda762070192e37e.

Validated revisions:

• cups-filters: 11d1a190530f85a361a1b3835f57d635256dff1a
• libcupsfilters: 905fd94fb22a9298bc8e2c845eb7e04f7055b686
• libppd: 522af8dd135f4dde66b1aac8b9d067808bbe122d

The 6,469-byte PoC is:

6467 bytes of "A" + "/*"

Its SHA-256 is
1f34d3d7cc9a19bd4b543ab3e64102ab9a738bc91737ee818d2cff6138b906ed.
With the fallback Letter page, cpi=12, lpi=8, and pretty-print header,
the page contains 84 columns by 77 lines. The / occupies the final cell and
the * crosses onto a freshly cleared page.

Run:

./reproduce.sh

This invokes the standard seven-argument CUPS filter interface through
ppdFilterCUPSWrapper() and then cfFilterTextToPDF().

The following commented Bash script constructs the exact PoC and control
inputs. Save it as make_poc.sh, then run bash make_poc.sh poc.

#!/usr/bin/env bash
set -euo pipefail

# PoC: texttopdf: comment token across a page boundary writes before the page buffer
# Finding ID: texttopdf-comment-page-boundary-oob
# Trigger: cfFilterTextToPDF() writes two bytes before its heap-allocated page
# buffer when C/C++ pretty printing recognizes /* at the exact automatic page
# boundary. The syntax state is updated before wrapping, but the first comment
# character is highlighted after wrapping with column - 1, when column has
# already been reset to zero.
#
# Binary PDFs, Raster files, images, and PPDs are stored as gzip-compressed
# base64 so embedded NUL bytes and exact parser offsets survive copy/paste.
# Every reconstructed file is checked before the vulnerable program is run.

OUTPUT_DIR="${1:-poc}"
mkdir -p "$OUTPUT_DIR"

for tool in base64 gzip sha256sum; do
  command -v "$tool" >/dev/null || {
    printf 'missing required tool: %s\n' "$tool" >&2
    exit 1
  }
done

write_file() {
  local name="$1"
  local expected_sha256="$2"
  local path="$OUTPUT_DIR/$name"
  local actual_sha256

  mkdir -p "$(dirname "$path")"
  base64 --decode | gzip --decompress > "$path"
  actual_sha256="$(sha256sum "$path")"
  actual_sha256="${actual_sha256%% *}"
  if [[ "$actual_sha256" != "$expected_sha256" ]]; then
    printf 'SHA-256 mismatch for %s\n' "$path" >&2
    return 1
  fi
  printf '%s  %s bytes  sha256=%s\n' \
    "$path" "$(wc -c < "$path")" "$actual_sha256"
}

# Malformed or boundary document consumed by the real filter/API.
# Output: document.txt (6469 bytes)
write_file document.txt 1f34d3d7cc9a19bd4b543ab3e64102ab9a738bc91737ee818d2cff6138b906ed <<'POC_PAYLOAD_0'
H4sIAAAAAAACA+3BMQEAAAwCoD4+ZrB/qf3LAGwAAAAAAAA8zQExUFYzRRkAAA==
POC_PAYLOAD_0

Result

ASan reports:

ERROR: AddressSanitizer: heap-buffer-overflow
WRITE of size 2
cfFilterTextToPDF .../cupsfilters/texttopdf.c:1433:40
0x52b0000001fe is located 2 bytes before 25872-byte region

The complete report is in asan.txt. The unsanitized build exits with status
134 after glibc reports double free or corruption (out); see baseline.txt
and baseline-stderr.txt.

The two UBSan diagnostics in sfnt-private.h occur earlier during font
embedding and are independent of this page-buffer write.

Cause and expected behavior

At texttopdf.c:1325-1334, the * and preceding / enter comment state.
At lines 1366-1380, wrapping resets column, line, and the page when the
current column is full. At lines 1425-1433, the comment highlighter then writes
doc.Page[line][column - 1].attr, which addresses the element before the new
page buffer.

The highlighter must retain the pre-wrap location of /, highlight it before
the cursor transition, or guard the previous-cell access across line and page
boundaries. A valid syntax token crossing a layout boundary must not access
outside doc.Page.

Start with cupsfilters/texttopdf.c at lines 1325-1433 and run reproduce.sh with the supplied 6469-byte document and sanitizer configuration. Trace the comment-state transition, page wrapping, and highlighter access across the boundary. Done means the standalone filter no longer reports the heap-buffer-overflow and preserves valid comment highlighting at the page boundary.