| <<<Back 1 day (to 2018/05/09) | 20180510 |
Robin_Watts | Ok, modulo rop differences, I think the pixel_patch stuff is there. | 13:05.35 |
| I'd really like someone else to look this over, if only to get some comments like "that doesn't look like gs naming" etc. | 13:06.13 |
| This is the branch in question: http://git.ghostscript.com/?p=user/robin/ghostpdl.git;a=shortlog;h=refs/heads/no_rects | 13:08.12 |
| The "Fix nonsensical ROP expression" one can be ignored (HenryStiles is looking at a proper version of that) | 13:08.43 |
| The "Fix problem with transparency and ROPs" will need to be evaluated in light of what HenryStiles comes up with for the previous one. | 13:09.32 |
| The rest are good to go. | 13:09.48 |
| I suspect I should update the gs docs with information about this new dev function. | 13:10.07 |
HenryStiles | Robin_Watts: I'm sorry I will get to that today. I had another diff with my fix to be investigated. | 13:10.54 |
kens | I n ok at it, but I doubt my ability to contribute anything meaningful. Maybe wait for Chisl and/or Ray | 13:10.56 |
Robin_Watts | HenryStiles: No worries, that wasn't a complaint. | 13:11.03 |
| kens: Well, I'd like you in particular to look at it because a) you expressed an opinion about enumerators etc, b) because there are some small pdfwrite/pswrite changes in there, and c) because I probably need to do something about subclassing. | 13:12.09 |
HenryStiles | Robin_Watts: I think it is a big change and should have assigned review, I'm nominating both ray_laptop and chrisl | 13:12.26 |
Robin_Watts | Makes sense. | 13:12.37 |
| I'm generally in favour of reviews. Spread the blame around :) | 13:13.03 |
kens | Robin_Watts : I'd expect the subclasing to 'just work', easy test, run a file which triggers your code, and add a -dFirstPage/LastPage | 13:14.56 |
| Typing is tricky, I have a hell script running tests and teh GS window keeps grabbing the focus and eating my keypresses | 13:15.36 |
| s/hell/shell/, maybe :-) | 13:15.47 |
Robin_Watts | kens: I was imagining I'd need to have a device function in the subclassing device that captures calls to pixel_patch and swallows them for unwanted pages, but I haven't looked at it at all. | 13:16.44 |
kens | I think you need to update the gdevflp.c firstpage lastpage device rather than 'subclassing' as such | 13:17.43 |
| ie its a specific case rather than general | 13:17.52 |
| Give me a minut till thi test completes and my computer is usable again :-( | 13:18.52 |
| OK we talked about the lcvd hideousness already, so that's OK | 13:20.18 |
Robin_Watts | yeah, could well be a specific case. | 13:21.25 |
| (or every specific case, of which there is only one, currently) | 13:21.34 |
kens | Well the first page/last page device does need to process *every* device method. Which is why you get complaints in debug builds that it implements obselete methods | 13:22.09 |
| So it will need modified yes. | 13:22.21 |
| I think you'll need to look at the device (name forgoteen, I'll check) which drops images too | 13:22.38 |
| Presuambly if we aren't rendering images, we don't want to do pixel_patch ? | 13:22.56 |
Robin_Watts | If begin_image drops stuff, then pixel_patch will never be called. | 13:24.17 |
kens | So I'm probably reading this wrong, but you modified the lcvd device to have a pixel_patch method, but it returns '0' in response to copy_color_is_fast. | 13:24.34 |
Robin_Watts | oh, so maybe we're OK for firstpage/lastpage because of that. | 13:24.34 |
kens | Is that the right way round ? | 13:24.44 |
| And yes, first page last page will not pass on the begin image | 13:25.00 |
| If they are dropping the page | 13:25.06 |
Robin_Watts | copy_color_is_fast returns non-zero if copy_color does not resolve to rectangles. | 13:25.20 |
kens | In which case both these will be OK, but it might be better (belt *and* braces) to add handlers | 13:25.24 |
| Then that makes sense, OK, I wasn't sure what the sense of the test was. | 13:25.52 |
chrisl | I kind of assumed we already had code that degenerates images into fills | 13:26.18 |
Robin_Watts | chrisl: We do. It's the image_render_color_portrait_icc stuff. | 13:26.41 |
| and thats the stuff that's had its guts ripped out to do pixel_patch. | 13:26.56 |
| So we have the same code, just in a slightly different place now, and in a way that can be overridden with "better" versions. | 13:27.18 |
chrisl | Okay, I see | 13:27.59 |
Robin_Watts | I'd understand if you guys would rather hold off until I've written the documentation. | 13:30.25 |
kens | So it looks like you need a new 'pixel_patch' method for general subclassing in gdevsclass.c, and I think you need to add one for gdevflp.c and gdevoflt.c when those devices are not swalllowing images. At least I'd be happier if those all gained methods to handle this. | 13:32.29 |
Robin_Watts | kens: OK, will do, thanks. | 13:32.49 |
kens | NP, it shoudl be easy enough to implement following the code already there I think | 13:33.05 |
chrisl | Well, beyond not being keen on "pixel_patch" as a name, I doubt there's much I can contribute - I'm guessing the cluster reports a lot of diffs :-( | 13:33.35 |
kens | I wasn't going to mention the name :-) | 13:33.56 |
chrisl | Trouble is, I'm not really sure of a better oner | 13:34.22 |
kens | pixel_copy ? | 13:34.49 |
| pixel_blit ? | 13:34.55 |
Robin_Watts | chrisl: No. the cluster reports almost no diffs. | 13:35.29 |
kens | From my POV 'patch' makes me think of Coons or Tensor patches | 13:35.31 |
chrisl | Same here | 13:35.42 |
Robin_Watts | The point is that the "patch" of pixels passed to this function might be portrait, or landscape, or even skewed. | 13:36.15 |
| "color_lookup_and_affine_texture_map_2d_array_of_pixels" seemed like a mouthful. | 13:36.52 |
| blit and copy both imply 1:1 to me. | 13:37.22 |
| chrisl: The only diffs you get with this are down to ROPs. | 13:37.52 |
kens | patch certainly doesn't imply what you say it does though, at least not to me | 13:37.54 |
Robin_Watts | it's a 2d region of pixels to get transformed onto the output device. | 13:38.17 |
| transform_pixel_patch ? | 13:38.30 |
kens | how about pixel_region then ? | 13:38.32 |
| I'd prefer to avoid patch because PostScript and PDF already use the word | 13:38.49 |
Robin_Watts | I could live with pixel_region if we have a consensus on that. | 13:39.10 |
chrisl | I prefer the lack of name clash - I still don't especially like it, though.... I wish I could say something more useful than just complaining about it | 13:40.09 |
kens | I wasn't going to complain, so obviously its not something I'm wildly concerned about, but I'd prefer region to patch | 13:41.02 |
Robin_Watts | I'm always prepared to revisit naming if it can be clearer. | 13:43.08 |
| transform_pixel_region would be my favored one I think. | 13:43.20 |
kens | I'm happy with that | 13:43.30 |
chrisl | I definitely prefer that | 13:44.36 |
chrisl | has to reboot | 14:14.18 |
louis___ | Hi, is it normal if I use sdevice=pxlcolor to print pdf document which contain 1 page and I see 20 pages at the print queue? The output at the printer is only 1 pages then the problem is just at the spooler level | 14:15.23 |
kens | You haven't said which operating system you are using, nor how you are transferring the PXL output file to the printer, so we can't really comment | 14:16.20 |
| Also, that's not really a question about Ghostscript, its your OS, whatever that is | 14:16.34 |
louis___ | windows server 2012 and I use gswin64 | 14:17.01 |
kens | And how are you transferring the PXL output file to the printer ? | 14:17.22 |
louis___ | yes with pxlcolor -sOutputFile=%printer%\\ argument | 14:17.57 |
kens | Well the only way the print spooler is going to be able to tell how many pages tehre are is if it can: | 14:19.03 |
| 1) Fully interpret the page description language | 14:19.03 |
| 2) prescan the whole input file to determine where the pages begin and end | 14:19.04 |
| The Windows print spooler is not capable of either of these feats | 14:19.17 |
| It just accepts a stream of bytes to send to the printer | 14:19.30 |
| So I suspect that its simply guessing when it says there are 20 pages, in the absence of any actual information | 14:20.01 |
louis___ | ok thanks | 14:20.15 |
kens | If you think tha'ts wrong, then I suggest you take it up with Microsoft | 14:20.17 |
louis___ | :) | 14:20.30 |
Robin_Watts | ray_laptop, chrisl, anyone else interested: I've updated the no_rects branch on my repo with the renamed device call. | 19:11.07 |
| ray_laptop: Dunno if you saw the discussion earlier, but my optimised image plotting changes are ready for review, and HenryStiles nominated you and chrisl. | 19:11.47 |
ray_laptop | Robin_Watts: sorry I missed the comment about reviewing your patch. I have some questions/concerns... | 23:59.11 |
| but I'll wait until you reply and keep slogging through. | 23:59.41 |
| Forward 1 day (to 2018/05/11)>>> | |