367433 - Change mac printing to use print manager instead of cups

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Stuart Parmenter]

This should fix a bunch of the mac printing bugs

Comment 1

[Stuart Parmenter]

Attachment: fix

first pass. this needs some better comments about the weird ordering as well as testing on windows and linux

Comment 2

[Colin Barrett [:cbarrett]]

Attachment: fix with 10.3 support

Adds printing support for 10.3, based on stuart's patch.

Comment 3

[Colin Barrett [:cbarrett]]

Uses the runtime version check methods from 367548

Comment 4

[Stuart Morgan]

When this lands I assume products can remove the -lcups we added when the last round of print stuff went in?

Comment 5

[Stuart Parmenter]

yeah this will fully remove the cups dependency.  I'm hoping to finish this up soon (just needs more comments and some build fixes on windows and linux)

Comment 6

[Josh Aas]

Comment on attachment 252126 [details] [diff] [review]
fix with 10.3 support

-#ifndef nsDeviceContextSpecX_h_
-#define nsDeviceContextSpecX_h_
+#ifndef nsDeviceContextSpecX_h___
+#define nsDeviceContextSpecX_h___
...
-#endif // nsDeviceContextSpecX_h_
+#endif

Why change that stuff? That is the standard format for every header file in cocoa widgets, exactly.

Comment 7

[Stuart Parmenter]

(In reply to comment #6)
> (From update of attachment 252126 [details] [diff] [review])
> -#ifndef nsDeviceContextSpecX_h_
> -#define nsDeviceContextSpecX_h_
> +#ifndef nsDeviceContextSpecX_h___
> +#define nsDeviceContextSpecX_h___
> ...
> -#endif // nsDeviceContextSpecX_h_
> +#endif
> 
> Why change that stuff? That is the standard format for every header file in
> cocoa widgets, exactly.
> 

just there because we had to revert your changes to that file to get the previous patch to apply...  can be changed before I check in.

Comment 8

[Colin Barrett [:cbarrett]]

Attachment: fix with 10.3 support & non-clobbered include guards

Reverted these include guards (IIRC they got changed because stuart was working on an earlier version of the tree).

Comment 9

[Colin Barrett [:cbarrett]]

Attachment: 10.3 compilation support

this should allow us to compile on 10.3. ardissone, can you test?

Comment 10

[Stuart Parmenter]

Attachment: now with proper size stuff and windows and linux fixes (i think)

this should solve the size problem and should fix the windows and linux pieces. testing that now.

Comment 11

[Colin Barrett [:cbarrett]]

Attachment: dynamically load 10.4 printing symols

Patch on top of stuart's patch. Generated with interdiff --no-revert-omitted between attachment 252716 [details] [diff] [review] and my changes. It should apply to a tree that has had stuart's patch applied to it.

It dynamically loads the printing symbols to use on Tiger -- it should allow us to compile against the 10.3.9 sdk. Testing would be appreciated, as cross compiling to PPC (which is necessary to test 10.3.9) is currently broken (in our build system).

Comment 12

[philippe (part-time)]

(In reply to comment #11)

> It dynamically loads the printing symbols to use on Tiger -- it should allow us
> to compile against the 10.3.9 sdk. Testing would be appreciated, as cross
> compiling to PPC (which is necessary to test 10.3.9) is currently broken (in
> our build system).
>

I could build Camino on PPC 10.4 with GCC 3.3 + 10.3.9 SDK (ppc build only)
(and I could print to PDF more or less correctly)
I'll spin a Minefield build with GCC 4.01 + 10.3.9 SDK later.

Comment 13

[Stuart Parmenter]

Attachment: hopefully final patch

i've tested this on windows and mac.. testing on linux now

Comment 14

[Stuart Parmenter]

Attachment: final patch.

this has been tested/works on all 3 platforms.

Comment 15

[Boris Zbarsky [:bzbarsky]]

So with this change, creating a surface with the same width/height as a given one is pretty painful, because gfxSize has float width/height, while gfxPlatform::CreateOffscreenSurface takes integer sizes.  Passing a float in to an int arg leads to compiler warnings; casting to get rid of those is a pain.

Could we perhaps restore the old Width()/Height() or have a version returning nsIntSize or something?

Comment 16

[Stuart Parmenter]

CreateOffscreenSurface should be changed to take floats.  We want to use structs for everything in thebes and we can't use nsIntSize because it lives in old gfx and we don't want that dependency.  We could have a gfxIntSize or somesuch (or do templates as I've suggested in the past) but I would rather avoid that for now and just use doubles.

Comment 17

[Boris Zbarsky [:bzbarsky]]

Sure, all I care about is that there not be an impedance mismatch between the APIs.