Douglas,
>
> Abel,
> I forwarded the this post onto Juergen Fischer. Unfortunately
> I'm probably getting a bit confused between the aha152x
> [which is mainatined by Juergen] and the aha1542 [Eric
> Youngdale?].
>
> Anyway, which kernel version are you testing? Where does that
> "--enable-scsibuffersize=32768" go (is it backend specific)?
I'm using 2.2.13 as distributed by Suse with their Linux 6.3 package.
But the crash also occurs with a 2.2.14 kernel.
"--enable-scsibuffersize" is used within sanei_scsi.c, so that it
applies to most backends for SCSI scanners. The exceptions are the Umax
and the Sharp backends, which use a new call to select the SCSI buffer
size during program runtime (sanei_scsi_open_extended instead of the old
sanei_scsi_open). I introduced "--enable-scsibuffersize" so that one can
change the default buffer size at installation time. Practically, it is
a kind of replacement for the old SG_BIG_BUFF: The Sane SCSI library has
an internal static variable sanei_scsi_max_request_size, which was
initialized to SG_BIG_BUFF. Since this constant does not make much sense
with the new SG driver, I replaced it with the configure option; you can
also set the environment variable SANE_SG_BUFFERSIZE to change the
maximum buffer size. (Sorry, forgot to mention the latter option in my
last mail.)
>
> I have a aha1542b, a UMAX 1220S and sane 1.0.2 installed.
The Umax backend is not the most trivial piece of software; while it
tries to allocate a SCSI buffer of 128 kB when opening the SG file, I
could not figure out, how much data is actually transferred with a
single SCSI command. (Oliver, can you give a hint??)
Abel
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
This archive was generated by hypermail 2b29 : Tue Apr 11 2000 - 07:55:24 PDT