foo2xqx produces garbled print result on HP LaserJet Pro M1136 MFP

[Icenowy]

I'm using foo2xqx on a HP LaserJet Pro M1136 MFP using the M1132s configuration (which seems to be the same printer with M1132s MFP, they share resources on HP websites). It usually works fine but sometimes the printed pages are garbled.

An example is attached (this is derived from some document from DigiKey). The XQX file is generated with foo2xqx-wrapper -r1200x600 -p9 -T3 -m1 -s7 -d1 -n1 .

bug4.ps.gz
bug4.xqx.gz

[Icenowy]

HPLIP (with proprietary plugins) can print this file w/o issues, and here's the result of printing this XQX file then scan the result with HPLIP (scanned with a low resolution to reduce file size).

[Icenowy]

Strangely, with HPLIP proprietary plugins, the emitted stream seems to be ZJS instead of XQX.

A dumped ZJS stream (dumped with Wireshark, with PJL snippets) are attached.

raw.zjs.gz

[ValdikSS]

There used to be some models which support both, yes.
I'm not a fan of debugging such issues without the printer. If I manage to get it from somewhere, I'll take a look.
For now, I guess, you can see which model is similar to the ZJS stream HPLIP has generated, and if it works in foo2zjs, we can just replace the driver.

[Icenowy]

I tested to set MX value to 0 instead of 16 (because the dumped ZJS stream has MX=0), and the resulting XQX file can be correctly printed.

[Icenowy]

I tried -z1 and -z2 of foo2zjs, neither of the result can be handled by the printer and the printer resets when receiving them.

[ValdikSS]

May I ask you to test MX of HPLIP for all other XQX models? If it's all 0, we can just change to this value. If it's different, I'll make a special case for this model only.

[Icenowy]

I don't know how to let HPLIP emit XQX without real printers...

[Icenowy]

I doubt for all HP printers the MX value is 0 -- only 1 functions are called by hpcups filter for JBIG (another one is called for JPEG), the prototype are below:

int hp_encode_bits_to_jbig(int iWidth, int iHeight, unsigned char **pBuff,
                        HPLJZjcBuff *pOutBuff, HPLJZjsJbgEncSt *pJbgEncSt);

The pJbgEncSt struct, which contains the JBIG encoding setting, is an output instead of an input -- none of its fields are set before calling the function.

[Icenowy]

Some interesting point: for M1136 MFP with hpijs, hpijs is called with LaserJet 1018 as parameter -- maybe ZJS streams produced by -z1 is okay but PJL preamble is not correct.

[ValdikSS]

A dumped ZJS stream (dumped with Wireshark, with PJL snippets) are attached.

Oh, you don't know how to print into a file?

1. /etc/cups/cups-files.conf FileDevice Yes
2. Add printer as file:///tmp/print-output

[Icenowy]

Yes, with PJL preamble from foo2zjs replaced with the one from hplip, ZJS stream (generated with foo2zjs-wrapper -r600x600 -z1 -p9 -T3 -m1 -s7 -d1 -n1 -P) could be printed -- but because of wrong MX value the printed result is incorrect too.

Maybe other HP printers should be tested with this PS file? (I printed it with an A4 paper)

[ValdikSS]

I'll be able to test on M1005 tomorrow.

[Icenowy]

I tested with a "Virtual P1505", and the resulting XQX file has

   11c: XQX_START_PLANE(5), 4 items
   124:         XQXI_0x80000000, 64 (0x40)
   130:         XQXI_0x40000000, 0 (0x0)
   13c:         XQXI_BIH(0x40000002)
                DL = 0, D = 0, P = 1, - = 0, XY = 9472 x 6791
                L0 = 128, MX = 0, MY = 0
                Order   = 3  ILEAVE SMID
                Options = 92  LRLTWO TPDON TPBON DPON
                54 stripes, 0 layers, 1 planes

   158:         XQXI_END, 3735928559 (0xdeadbeef)

[Icenowy]

A "Virtual P2014" produces the same values.

[Icenowy]

So does P1006. (The hpcups.drv.in file in hplip seems to have two classes for "ljzxstream" printer language, one is for m1005/p1505/p1505n/p2014/p2014n, another is for p1005/p1006/p1007/p1008/p1009).