[sane-devel] imageClass bug

Povilas Kanapickas povilas at radix.lt
Mon Nov 16 23:34:45 GMT 2020

Hi Ralph,

On 11/14/20 9:03 AM, Ralph Little wrote:
> Hi,
> I am hoping to garner some opinions about an issue that I am having with
> testing a Canon imageClass machine with the pixma backend.
> I have an imageClass MF4770N machine that has always been a little
> problematic on the ADF and I have never really understood why.
> With Rolf we managed to get the machine fairly well supported but
> sometimes the machine would fail with a service code and abort requiring
> a power-cycle.
> I recently managed to track down the specific behaviour that caused this
> issue but I am not sure how to deal with it.
> When scanning on the ADF, if the user selects a scan area that is
> aligned to both the hard left and hard right of the scan area, then the
> machine scans OK. However, if the selection is away from the left area,
> the selection can not extend within 32 pixels of the right hand side.
> Within that restriction, I can scan whatever I want.
> In order to scan a selection that aligns with the right hand side, the
> select must align with the left hand side also.
> This, I think is a bug in the machine's firmware. I just checked and it
> is fully up-to-date so it looks like there are no fixes coming.
> On Windows, when using the ADF, one can only select a fixed media type
> that is centred in the scan area. One cannot arbitrarily select a scan
> area. So it is not possible to see what the Windows driver would do
> because the scenario is impossible.
> My problem is how to deal with the issue in the Pixma backend. The
> common scan scenario in ADF is to scan Letter media which is a full
> width scan. I cannot just make the maximum width 32 pixels less to avoid
> the problem. In 75dpi, 32 pixels is a significant chunk to miss from the
> right hand side.
> Any ideas?

Could it be that there's some alignment requirement? E.g. the scanner
requires that the left hand side pixel offset is a multiple of 32, or
maybe the pixel count must be a multiple of 32 or something like that?
I've seen similar things on the genesys backend.


More information about the sane-devel mailing list