The GTK Printing code didn't support different preferences by printer, neither storing the last used printer. This patch adds the necessary changes to the GTK backend. * Ensures a printer name is set * Allow setting prefs by printer * Removes Postscript fallback pref mangling. * Store Paper sizes by GTK names, while still supporting old names. * Removes nsPaperPS.cpp * Page Setup preselects last used printer NOTE: The GTK PageSetup dialog always resets the paper size when passing a printer, I didn't found a workaround yet. The patch requires applying the patch in Bug 968141 first.

Comment on attachment 8371377 [details] [diff] [review] StoreGTKPrintSettingsByPrinter.patch This seems like a step in the right direction. Could you please explain: 1) Why do we still need the nsPaperSizePStoGTKList at all? 2) Why do we still need the code to strip "PostScript/" from printer names? (in InitPrintSettingsFromPrinter) 3) What exactly do you mean by "GTK seems to return custom sizes by display name?" (In general, we do not ever want to check in commented-out code; either it shouldn't be there at all or we should make it work correctly.) 4) You removed some #ifndef MOZ_WIDGET_X11 from nsPrintOptions. Does the Gtk back end actually support the code that was disabled for it, now?

I try to keep backward interface compatibility, meaning anything which makes sense to keep working. This means even adding some code for better compatibility. 1) Converting paper names. The paper list converts Gecko PS based names like A4 or Letter to PWG names GTK can understand. This means old settings or addons won't break. It would be nice to add js based interface tests for this. I need some help with this. (Technical on Linux you can still call the Printer Settings js api, there you can set a Gecko/PS name and the non writeable margins) 2. Postscript/ and Cups/ Because technical the current code if cups shim won't work or is disabled could return Postscript/ I plan to remove the PostScript CUPS shim and replace it with GTK code. Again if someone has PostScript or passes PostScript based names to the current Interface it shouldn't fail. However this part should be in a different Bug, so if it fails it can be reverted easily. 3. There seems to be a bug with GTK at least on a Ubuntu machine here. It is possible to create new Custom Paper names in the Page Setup UI. However instead of returning a PWG compatible name it returns a user defined Display Name. Anything you write there. This fails with GTKPageSetup horrible. I guess I need to write a platform independent PWG paper name parser and check what the API returns so we don't get zero width pages that makes PrintPreview very very unhappy. (I know your feelings about XUL Dialogs, however if I find more annoying bugs like this, my second strategy is to create the necessary interfaces to support a JS based solution and drive the code for Win32 and Linux with the same code and dialog) 4. Prefs by Printer. Yes, it now actually does. Again I need help with writing tests. The old code simply didn't set the Printer Name when doing Init from Printer. This means a lot of stuff screwed. Again some GTK Bugs are annoying. Specially the Page Setup won't allow to set a Printer and a Paper at the same time. But except more generally testing now the code should work as expected and the ifdef can go, at least from a GTK Linux standpoint. Could be that the OSX code path may not work as expected. But I think that should be fixed in OSX code and not in the platform independent code.

Pushed by ealvarez@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/db9899666fa8 Save per-printer settings on GTK. r=jwatt