Re: [Gimp-print-devel] Canon PIXMA mp640 default media size

[Gernot H. <aik...@gm...>]

On Thu, Mar 2, 2017 at 2:36 AM, Solomon Peachy <pi...@sh...> wrote:
> On Thu, Mar 02, 2017 at 12:24:31AM +0900, Gernot Hassenpflug wrote:
>> Done except for Oufku Card, since the defined orientation in
>> papers.xml is wrong: the Canon driver wants it to be w567h420 rather
>> than w420h567 (well, in units of 1/600 inches).
>
> It's a lot of work in the short term, but in the longer term it may make
> more sense to have the Canon driver do its own paper management, perhaps
> similar-ish to how the Dyesub driver does things.  Effectively it means
> unique paper types for the Canon driver, but they can re-use standard
> names.

Hello Solomon and all,

Here is my report for why the Canon driver (on my system at least)
seems to be behaving so weirdly with papersizes.
In fact, there does not seem to be any problem as long as the
papersize is correctly passed in to gutenprint, and here it gets
weird, because while gutenprint papers.xml defines the names that the
canon driver uses, nor even seem to relate to names in CUPS standard
definitions (/usr/share/cups/ppds/media.defs).

So the situation is that the names displayed in the CUPS interface are
often not the gutenprint names at all (as seen in the PPD), and in the
worst case the displayed name is in fact the wrong one (e.g., B5
displayed for B5 ISO), while the real name to use is unrecognizable
(e.g., JB5 for B5 JIS).

A related issue is that although papers.xml seems to have separate
papersizes for landscape for several papersizes, very few of them show
up in the papersize selection. I have no idea what logic is used to
build the papersize selection that the CUPS interface uses.

I thought that, also from Solomon's explanation above, that the
gutenprint driver could reuse standard names, but it seems that the
names the gutenprint driver uses are overwritten by something else.
If the user manages to find the right papersize name in this muddle,
then the Canon driver gets the correct papersize input (i.e.,
gutenprint name from papers.xml), else the Canon driver determines
that the papersize is custom (which is correct behaviour).

It does appear that some papersizes, although selected in the CUPS
interface, get replaced by other ones somewhere in the process of
reaching gutenprint. For instance, selecting JB5 results in a name of
B5 reaching gutenprint, which corresponds to the B5 JIS seen in the
PPD. On the other hand, selecting Hagaki Card or Oufuku Card results
in the Postcard or Wide Postcard (actually hagaki (postcard) or Wide
Postcard in the CUPS selection) page dimensions (which are slightly
different from the Hagaki Card or Oufuku Card dimensions), but the
papersize string is given as "Letter".

> Basically, the dimensions the printer needs should be reflected
> accurately in the ppds so that CUPS will supply data with the correct
> aspect ratio.  Many full-bleed dyesub printers proved to be particularly
> sensitive to this..

In this case the names are muddled and although the sizes are (mostly)
present it is really hard to figure out which to choose since many
names are. I am sure doing what Solomon did with the dye-sub driver
would be a good way to control things, but first I need to understand
what on earth is involved in building the papersize selection, and
where the various names and aliases are coming from.
And then why something as innocuous as Hagaki Card could possibly
become in dimensions a closely-matching other papersize (Postcard) but
be diagnosed as Letter in the papersize string.

I'll post a list of the comparitive names seen in the gutenprint
papers.xml, PPD (and CUPS media.defs) and actual CUPS interface GUI
later tomorrow.

Best regards,
Gernot