pdftopdf exits with status 1 (no error message) on PDFs with form fields lacking /N and /BBox

[hadisch]

pdftopdf exits with status 1 (no error message) on PDFs with form fields lacking /N and /BBox

Summary

pdftopdf terminates with exit status 1 when processing a PDF that contains
AcroForm fields whose appearance streams are missing the required /N entry
and /BBox. It emits only warnings, never an error, and gives no indication
that it is about to fail. The next filter in the chain (ghostscript) receives
an empty stream, reports "Input is empty, outputting empty file" and exits 0.
The job then fails with the generic message universal filter failed., which
points at the wrong component and makes the cause very hard to locate.

The same PDF prints fine after being rewritten by Ghostscript with
-dShowAnnots=false.

Versions

Package	Version
libcupsfilters	2.2.1
cups-filters	2.0.1
libppd	2.1.1
cups	2.4.19
ghostscript	10.07.1
Distribution	Artix Linux (Arch-based, rolling)

Printer: HP Smart Tank 7000 series, USB, hpcups driver with PPD.
The printer itself is not relevant — the failure happens before any
printer-specific filter is reached.

Steps to reproduce

Invoke the filter chain manually on an affected PDF:

sudo env \
  PPD=/etc/cups/ppd/HP-Smart-Tank-7000-series.ppd \
  CONTENT_TYPE=application/pdf \
  FINAL_CONTENT_TYPE=application/vnd.cups-raster \
  CUPS_DATADIR=/usr/share/cups \
  CUPS_SERVERBIN=/usr/lib/cups \
  /usr/lib/cups/filter/universal 1 user test 1 "" affected.pdf > /dev/null

Actual output

Abbreviated; the annotation block repeats once per form field:

DEBUG: page 1: Contains annotations
DEBUG: Opening field stream 1/4... with obj number 17
Key 'N' not found.
WARNING: Appearance stream is missing required /BBox.
DEBUG: special case ignore annotation with no appearance
...
DEBUG: Opening content stream 1/1...
DEBUG: Opened stream 1, resmap[1]=(nil)
DEBUG: cfFilterChain: pdftopdf completed with status 1.
DEBUG: cfFilterGhostscript: Input is empty, outputting empty file.
DEBUG: cfFilterChain: ghostscript completed with status 0.
ERROR: cfFilterChain: pdftopdf (PID 7025) stopped with status 1
INFO: cfFilterChain: ghostscript (PID 7026) exited with no errors.
ERROR: universal filter failed.

Note that pdftopdf walks both pages to completion and handles every
annotation via the "special case ignore annotation with no appearance"
path, so the failure does not appear to happen at the point where the
malformed annotations are encountered.

Expected behaviour

Either process the document (the malformed annotations are already being
skipped) or fail with a message that names the problem. The current
behaviour produces a misleading error that attributes the failure to the
universal filter.

Additional observations

• A minimal PDF generated with gs -sDEVICE=pdfwrite passes through the
  same chain without any problem, so the chain itself is intact.
• Rewriting the file with
  gs -sDEVICE=pdfwrite -dPrinted=false -dShowAnnots=false makes it print
  normally. This is what I am currently using as a workaround.
• qpdf --flatten-annotations=all does not help — qpdf exits without
  complaint but the malformed appearance streams survive and pdftopdf
  still fails.
• The failure is independent of the print path: it occurs identically via
  lp, via a PPD-based queue and via a driverless everywhere queue
  (the printer does not accept application/pdf, so PDF conversion runs
  in that case too).
• Job options make no difference; the failure reproduces with an empty
  option string.

Possible regression

Printing these documents worked on this machine before libcupsfilters was
upgraded from 2.1.1 to 2.2.1. No other component of the print stack changed
in the relevant time window (ghostscript, poppler, cups and cups-filters
were all older than the last known-good print).

I have not verified this by downgrading, so please treat the regression
claim as circumstantial rather than confirmed.

Sample file

The document that triggered this is a personal order confirmation, so I
cannot attach it. Any PDF with AcroForm fields whose appearance streams
lack /N and /BBox should reproduce the issue; the warnings quoted
above are the reliable indicator in the log.

[DeusProx]

Got a similiar problem with my new printer. All prints I did from chrome failed and get stuck in the print queue as failed prints.

Downgrading from 2.2.1-2 to 2.1.1-4 fixes this for me and the same prints from chrome work again.

Trying to debug the conversion with one of the failed prints in the queue with the newest version 2.2.1-2 shows this:

% sudo cupsfilter -i application/pdf -m application/vnd.cups-pdf -p /etc/cups/ppd/EPSON_XP_5200_Series.ppd /var/spool/cups/d00006-001 > /dev/null                                         16:11:59
DEBUG: argv[0]="cupsfilter"
DEBUG: argv[1]="1"
DEBUG: argv[2]="root"
DEBUG: argv[3]="d00006-001"
DEBUG: argv[4]="1"
DEBUG: argv[5]=""
DEBUG: argv[6]="/var/spool/cups/d00006-001"
DEBUG: envp[0]="CONTENT_TYPE=application/pdf"
DEBUG: envp[1]="CUPS_DATADIR=/usr/share/cups"
DEBUG: envp[2]=""
DEBUG: envp[3]="CUPS_SERVERBIN=/usr/lib/cups"
DEBUG: envp[4]="CUPS_SERVERROOT=/etc/cups"
DEBUG: envp[5]="LANG=en_US.UTF8"
DEBUG: envp[6]="PATH=/usr/lib/cups/filter:/usr/bin:/usr/bin:/bin:/usr/bin"
DEBUG: envp[7]="PPD=/etc/cups/ppd/EPSON_XP_5200_Series.ppd"
DEBUG: envp[8]="PRINTER_INFO=cupsfilter"
DEBUG: envp[9]="PRINTER_LOCATION=Unknown"
DEBUG: envp[10]="PRINTER=cupsfilter"
DEBUG: envp[11]="USER=root"
DEBUG: envp[12]="CHARSET=utf-8"
DEBUG: envp[13]="FINAL_CONTENT_TYPE=application/vnd.cups-pdf"
INFO: pdftopdf (PID 648545) started.
DEBUG: ppdFilterLoadPPD: Color profile qualifier determined from job and PPD data 'RGB.Stationery.300dpi'
DEBUG: ppdFilterLoadPPD: No ICC profiles specified in PPD
DEBUG: ppdFilterLoadPPD: Searching for profile "-/Stationery"...
ERROR: ppdFilterLoadPPD: Last filter could not get determined, page logging by the PDF filter turned off.
DEBUG: ppdFilterLoadPPD: Last filter determined by the PPD: None; Final data MIME type: application/vnd.cups-pdf => PDF filter will not log pages in page_log.
INFO: ppdFilterEmitJCL: Filter function (PID 648546) started.
DEBUG: page 2: Contains annotations
DEBUG: page 3: Contains annotations
DEBUG: page_dict_cb(dict=0x55864edee050, key="ExtGState", outpage=0x55864edf0d00), type=5
DEBUG: page_dict_cb(dict=0x55864edee050, key="Font", outpage=0x55864edf0d00), type=5
DEBUG: page_dict_cb(dict=0x55864edee050, key="ProcSet", outpage=0x55864edf0d00), type=1
DEBUG: page_dict_cb(dict=0x55864edee050, key="XObject", outpage=0x55864edf0d00), type=5
DEBUG: Opening content stream 1/1...
DEBUG: Opened stream 1, resmap[0]=(nil)
DEBUG: page_dict_cb(dict=0x55864edeeb20, key="ExtGState", outpage=0x55864edf0d00), type=5
DEBUG: page_dict_cb(dict=0x55864edeeb20, key="Font", outpage=0x55864edf0d00), type=5
DEBUG: page_dict_cb(dict=0x55864edeeb20, key="ProcSet", outpage=0x55864edf0d00), type=1
DEBUG: Opening field stream 1/1... with obj number 15
Key 'N' not found.
WARNING: Appearance stream is missing required /BBox.
DEBUG: special case ignore annotation with no appearance
DEBUG: Opening content stream 1/1...
DEBUG: Opened stream 1, resmap[1]=(nil)
DEBUG: page_dict_cb(dict=0x55864edef560, key="ExtGState", outpage=0x55864edf0d00), type=5
DEBUG: page_dict_cb(dict=0x55864edef560, key="Font", outpage=0x55864edf0d00), type=5
DEBUG: page_dict_cb(dict=0x55864edef560, key="Pattern", outpage=0x55864edf0d00), type=5
DEBUG: page_dict_cb(dict=0x55864edef560, key="ProcSet", outpage=0x55864edf0d00), type=1
DEBUG: Opening field stream 1/3... with obj number 31
Key 'N' not found.
WARNING: Appearance stream is missing required /BBox.
DEBUG: special case ignore annotation with no appearance
DEBUG: Opening field stream 2/3... with obj number 32
Key 'N' not found.
WARNING: Appearance stream is missing required /BBox.
DEBUG: special case ignore annotation with no appearance
DEBUG: Opening field stream 3/3... with obj number 33
Key 'N' not found.
WARNING: Appearance stream is missing required /BBox.
DEBUG: special case ignore annotation with no appearance
DEBUG: Opening content stream 1/1...
DEBUG: Opened stream 1, resmap[2]=(nil)
DEBUG: ppdFilterEmitJCL: Filter function (PID 648546) exited with no errors.
ERROR: pdftopdf filter function failed.
ERROR: pdftopdf (PID 648545) stopped with status 1

[0-andy-0]

Same problem here - suddenly I was unable to print from the browser and tracked it down to chromium producing PDFs that the cups filter rejected, as described above. Rather than downgrade I've added a pdftocairo wrapper around my /usr/lib/cups/filter/pdftopdf to restore my system to a sane state.

[uddhavphatak]

@DeusProx, could you please share me a input file with this, I an unable to get a PDF with form field lacking /N and /BBox. It would become very helpful for me.

[rob215x]

I can reproduce this on Manjaro, and I can confirm it's a regression in 2.2.1: downgrading libcupsfilters alone fixes it. The trigger here is ordinary link annotations from Chrome, not only form fields. The comment above reports the same failure and the same fix with a different printer (Epson XP-5200), so it isn't specific to one printer.

About getting an affected file: the failing PDFs here contain no form fields at all. They are ordinary Chrome 153 print outputs whose only annotations are /Subtype /Link entries with no /AP, and they produce exactly the warnings quoted in this issue (Key 'N' not found, Appearance stream is missing required /BBox). So printing a page with several links from Chrome may be an easier way to get an affected file than hunting for a PDF form. I've hit it with a web email and with Google's homepage. I can't attach mine because they contain personal account data.

Environment: Manjaro (Arch-based), cups 2.4.19, cups-filters 2.0.1-2, libcupsfilters 2.2.1-2, libppd 2.1.1-2, pdfio 1.6.5-1, qpdf 12.4.1-1. Network IPP printer (Brother DCP-L2540DW), PPD queue, final format image/urf.

Symptom: Printing from Google Chrome 153 (Skia/PDF m153) fails. The same pages print fine from Firefox and Chromium, and from Chrome's "Save as PDF" when the file is opened in a viewer.

Findings:

• The annotations are /Subtype /Link with /A, /Border, /F, /Rect and /StructParent. There are no /AP entries.
• A one-link test page (printed from Chrome) passes. A web email (2 pages, 13 link annotations) and Google's homepage (1 page, 13 link annotations) fail. On the Google PDF, keeping only the first 1 or 2 annotations passes and keeping the first 3 or more fails. I haven't isolated whether it's the count or one specific link.
• Stripping /Annots from a failing file (using pikepdf) makes it print.
• Downgrade confirmed: rolling back only libcupsfilters from 2.2.1-2 to 2.1.1-4 (qpdf stays at 12.4.1-1) makes the failing files and direct Chrome printing work again. That rules out the qpdf update.

Reproduction: with the PPD copied to a readable path:
PPD=/tmp/b.ppd CONTENT_TYPE=application/pdf FINAL_CONTENT_TYPE=image/urf /usr/lib/cups/filter/universal 1 user t 1 "" affected.pdf > /dev/null; echo $?
It exits with 1 on 2.2.1 and 0 on 2.1.1.

Workaround for Arch/Manjaro users: install the older package from the pacman cache (sudo pacman -U /var/cache/pacman/pkg/libcupsfilters-2.1.1-4-*.pkg.tar.zst) and add libcupsfilters to IgnorePkg in /etc/pacman.conf.

[lostmind84]

Confirmed on another Arch-based Omarchy system:

• cups 2:2.4.19-1, cups-filters 2.0.1-2, libcupsfilters 2.2.1-2, poppler 26.08.0-1
• HP Smart Tank 7000 series over driverless IPP
• Printing a browser-generated Gmail PDF failed with repeated Appearance stream is missing required /BBox warnings, followed by cfFilterChain: pdftopdf ... stopped with status 1 and universal filter failed.

Downgrading to the signed Arch package libcupsfilters 2.1.1-4, restarting CUPS, and restarting the preserved job fixed it. The PDF completed successfully (CUPS reported 4 sheets). A CUPS test page that had failed with the same pdftopdf error before the downgrade also printed successfully afterward (1 sheet, confirmed physically).

This reproduces the regression and confirms the downgrade workaround on an HP Smart Tank 7000.