Office conversion fails (LibreOffice / unoconv incompatibilities)

[hackepeter]

Importing Office documents (XLSX, DOCX, DOC) via API, Webinterface and Consume folder fails on my Cloudron Instance with Paperless.

I found two compatibility issues in the Gotenberg/LibreOffice conversion process. Both could be resolved with a temporary workaround.

I'm not sure whether this affects all installations or just mine.

Environment

Component	Version
Cloudron package	1.66.0
Paperless-ngx	3.3.0
LibreOffice	26.2.6.3
unoconv	0.8.2
Python	3.14.4

1. LibreOffice 26.2 incompatibility

Gotenberg returns HTTP 500 when converting Office documents.

Debug logs:

File "/usr/bin/unoconverter", line 1012, in convert
    inputurl = unohelper.absolutize(...)

AttributeError: module 'unohelper' has no attribute 'absolutize'

The installed /usr/bin/unoconverter contains three references to this function:

1012: inputurl = unohelper.absolutize(...)
1026: templateurl = unohelper.absolutize(...)
1166: outputurl = unohelper.absolutize(...)

This appears to be a known LibreOffice 26.2 incompatibility:

https://bugs.debian.org/1143523

Workaround: Map the missing function to uno.absolutize().

After applying this fix, direct XLSX conversion via Gotenberg works (HTTP 200).

However, importing XLSX through Paperless still fails.

2. Unsupported unoconv argument

Paperless sends these parameters to Gotenberg:

pdfa: PDF/A-2b
updateIndexes: false

Gotenberg passes --disable-update-indexes to unoconv.

Debug output:

unoconv: option --disable-update-indexes not recognized,
try unoconv -h for a list of all the options

Result:

exit status 255
HTTP 500

Workaround: Remove the unsupported argument before unoconv processes it.

After applying this second fix, XLSX imports through Paperless work again.

Temporary workaround

Create /app/data/gotenberg-fix/sitecustomize.py:

import uno
import unohelper
import sys

unohelper.absolutize = uno.absolutize

if "--disable-update-indexes" in sys.argv:
    sys.argv.remove("--disable-update-indexes")

Make sure the file is readable by the cloudron user.

Stop Gotenberg:

supervisorctl stop gotenberg

Start it manually with the patch:

runuser -u cloudron -- env \
  HOME=/home/cloudron \
  PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin \
  PYTHONPATH=/app/data/gotenberg-fix \
  gotenberg --log-level=debug

This works but is not persistent across restarts.

The second patch also ignores the updateIndexes=false setting, so it's only a workaround.

Test results

Test	Before	After
Gotenberg health	HTTP 200	HTTP 200
Direct XLSX conversion	HTTP 500	HTTP 200
XLSX import in Paperless	Failed	Works

Direct LibreOffice conversion also works without the workaround.

Suggested fix

The Cloudron package likely needs two changes:

• Update or patch unoconv to replace unohelper.absolutize() with uno.absolutize().
• Fix the incompatibility between Gotenberg and unoconv regarding --disable-update-indexes.

Questions

• Can anyone reproduce this with Paperless-ngx package 1.66.0 on a fresh installation?
• Are LibreOffice and unoconv versions identical across installations?
• Could these two issues be addressed in the Cloudron package?

I haven't tested a fresh installation. The workaround fixes XLSX imports on my existing installation.

Related issue: https://forum.cloudron.io/topic/13545/gotenberg-permission-issue-and-large-log-volume