Re: [hpoj-devel] CVS hpoj code works (mostly)

[<pa...@rc...>]

Daniel Gun wrote:
> I checked out and compiled the hpoj cvs code of about March 13 or 14. 
> Everything compiled and installed just fine.  I have the redhat 2.2.17-14
> kernel and am using a PSC500 on the parallel port.
> 
> I was able to scan using sane 1.0.4 and ptal-connect directly.  I scanned a
> few MB images and also a 48MB image, and had no problems.  I was also able
> to print using ptal-connect, though I haven't tried ptal-printd.
Hi, Daniel.  Thanks for the success report.  I'm glad it works well for you.

> Thanks to David for getting this working!
You're quite welcome.  Of course some of the credit should go to my management
for being supportive of my involvement with this project (given that I don't
actually work for the all-in-one division) and for agreeing to GPL the code
that went into ptal-mlcd.

> As soon as version 1.0 gets
> released, I can start recommending HP all-in-ones to my linux friends.
That's certainly my goal.  :-)  What constitutes "version 1.0" to you?
I expect I will be making incremental improvements to the software for
quite some time, so I wasn't planning on bumping the version number to 1.0
any time soon.  Hopefully you'll recommend based on the functionality rather
than the version number.  :-)

> An interesting problem, though... when I tried "scanimage --test", the test
> appeared to pass without problem, but the scanner hung and ptal-mlcd died. 
> This is a repeatable problem for me.
Thanks for noticing that.  I guess I forgot to test with "scanimage --test".
:-)  I think I know what the problem is (a race condition between closing
the scan connection due to the aborted scan and data that's still moving
around in the daemon).  I'll let you know when I check in a fix, which
should be sometime this weekend.

> (The beauty of a user-mode driver is
> that I just restarted it and reset the PSC500 and everything worked fine
> again.)
That's one of several reasons I chose to rewrite the driver in user mode.
It's also much easier for me to debug and maintain.

> I haven't tried running it with debugging turned on, but want to
> know if anyone else has seen this problem.
Interestingly enough, I was able to reproduce this problem with a parallel-
connected R80 but not a USB-connected PSC750, so it makes sense that this
problem would be a race condition, given that USB and parallel have
different timing.  The exit is due to an unhandled signal (SIGPIPE),
although since you downloaded the code I added a handler for this signal
for an unrelated reason, so in my case I saw a bunch of error messages and
debug console resets without the exit (although the scanner still got stuck).

David