lj.so blob init error (or not?)

[Krzysztof Kotlenga]

cups 1.4.2
hpcups 3.9.12, "lj.so" in place
LaserJet m1120N MFP

Problem:
prnt/hpcups/ModeJbig.cpp:138:if (!HPLJSoInit || (HPLJSoInit && !HPLJSoInit (1)))

The above test succeeds -> ModeJbig::Init() returns PLUGIN_LIBRARY_MISSING -> printer doesn't work.
HPLJSoInit() returns 0. What value is it supposed to return? Because changing this logic a bit:

- if (!HPLJSoInit || (HPLJSoInit && !HPLJSoInit (1)))
+ if ((HPLJSoInit == NULL) || (HPLJSoInit (1) != 0))

surprisingly results in a working printer. Is the logic broken or HPLJSoInit() fails when it shouldn't?

[Krzysztof Kotlenga (pocek)]

Oops. I mean "lj.so", obviously.

[Krzysztof Kotlenga (pocek)]

First, the hack above is obviously wrong.

It looks like HPLJSoInit() will check if DEVICE_URI environment variable (passed from cups) starts with 'hp:' and fail otherwise. It's very unfortunate, because cups will automatically detect and set up the cups-native 'socket' protocol. 'hp' protocol doesn't work for me for unknown reasons, so here's the new hack to make the socket protocol working:

Create a wrapper for /usr/lib/cups/filter/hpcups:

#!/bin/sh
export DEVICE_URI="hp:/whatever"
exec /usr/lib/cups/filter/hpcups.orig "$@"

(don't forget to mv hpcups hpcups.orig first)

And it will work! Hurray! Thank you HP for making my live hard.

[Sanjay Kumar (sanjay-kumar14)]

Marking it as invalid because the issue occurs on very old release.

[Krzysztof Kotlenga (pocek)]

Not really, last checked with hplip 3.12.2.

The fix would be to allow 'socket' protocol in "lj.so" or to remove that check entirely (even better - because I don't think "lj.so" has _ANY_ knowledge to decide whether given printer/protocol is supported or not, so it's most likely wrong half of the time, making hpcups filter exit with oh-so-informative error 48).