[hpoj-devel] What I did to get 0.7 to compile and work with linux-2.4.0-test12

[dmc <dm...@in...>]

NOTE:  I do not care about developing this driver, just using it.  I
don't currently, and don't plan on reading the hpoj-devel list, as
long as I can print and scan on my PSC500, which I can now do.  I'm
just posting this info so other people needn't do the same digging around
that I did.

Despite the release notes that came with 0.7, it appears that 0.7
is not yet compatable with the latest version of 2.4.0-test, namely test12.

I had to change 3 things in ieee12844/ieee12844pp.c to get it to compile
and work.

First I had to add

#define PAR_CONTROL_REVERSE_DATA 0x20

This is not the "right" solution, but it works.  I went through old
versions of linux/include/linux/parport.h (or parport_pc.h) and
discovered that PAR_CONTROL_REVERSE_DATA had been
removed.  However, the parts of the linux code that used to handle
it, still have a special case for 0x20, I guess in case 3rd party drivers
use it.  Thats why I was confident that my solution was adequate
for the timebeing.  I'm not a kernel hacker, and I have no interest
in developing this driver, I just wanted it to work for me, so thats
why I didn't investigate into a more proper solution.

Second, I changed

parport_write_econtrol

to

parport_write_control

because I got an unresolved symbol for parport_write_econtrol
during the insmod

Again, a guess, that seems to work for me so far.

Third, and most fundamental, I changed

queue_task (&l->tq, &tq_scheduler);

to

schedule_task(&l->tq);

After a bit of research I discovered that tq_scheduler got nuked
by linux-2.4.0-test12-pre7.  The changelog yielded

 - pre7:
        ...
    - Andrew Morton: "tq_scheduler" is no more. We have keventd.
        ...

Not knowing much about kernel internals, I diffed pre6 and pre7
and discovered that every instance of

queue_task(X, &tq_scheduler);

was replaced by either

schedule_task(X)

or

MOD_INC_USE_COUNT;
if (schedule_task(X) == 0)
     MOD_DEC_USE_COUNT;

Even though I used the first and it worked, my guess is that the latter
is used for kernel modules, and is therefore more correct.  I'll probably
switch to it if I encounter any problems.

Otherwise, I CAN SCAN ON MY PSC500 :).  Thanks very much for
not making me regret nuking my windows partition a few weeks ago :)

And I'll eagerly await 0.8 which will have better fixes than my hacks.

-dmc