| <<<Back 1 day (to 2016/02/10) | 20160211 |
tor8 | Robin_Watts: hm, I'm seriously confused now. | 11:57.55 |
| I rewrote the opengl code to use fz_font and fz_render_glyph_pixmap rather than FT_Face and freetype directly | 11:58.17 |
Robin_Watts | tor8: ok... | 11:58.40 |
tor8 | now the glyph renders I'm getting out looks a *lot* lighter | 11:58.43 |
| and the spacing is all over the place, like the subpixel offsets are not working | 11:58.59 |
| do we do any gamma correction of the glyph data coming from freetype? | 11:59.17 |
Robin_Watts | We do not. | 11:59.22 |
tor8 | http://ghostscript.com/~tor/stuff/text-freetype.png vs http://ghostscript.com/~tor/stuff/text-fitz.png | 12:00.56 |
| flipping back and forth, it looks like some characters are spaced off by one entire pixel | 12:02.00 |
| and the outlines are rendered much lighter | 12:02.06 |
Robin_Watts | I think I prefer the lightness of the fitz rendering. | 12:03.07 |
tor8 | they should be IDENTICAL!!! | 12:03.13 |
Robin_Watts | but I can't offer any suggestions as to why... | 12:03.19 |
tor8 | which is driving me crazy | 12:03.24 |
| as far as I can trace the code, we're calling freetype the exact same way and I'm pulling out the exact same values and yet ... this! | 12:03.53 |
Robin_Watts | I'm having similar nutso things with gs and images/patterns at the moment. | 12:03.53 |
tor8 | time to add printf's everywhere... | 12:04.36 |
| Robin_Watts: okay, I've managed to match the lightness... | 12:22.16 |
| the fitz lightness is the more correct one | 12:22.53 |
Robin_Watts | cool. | 12:23.00 |
tor8 | the transform sent to freetype has a large effect | 12:23.04 |
| if you FT_Set_Char_Size to the actual pixel size as freetype expects the outlines get fattened | 12:23.19 |
| I chalk that down to lack of precision | 12:23.26 |
| mupdf thus sets it to 65536 and scales the matrix down instead | 12:23.40 |
Robin_Watts | ISTR that chrisl had to do some juggling with the rescaling factors in gs when calling ft to get sane results. | 12:23.44 |
tor8 | yeah. I'm thinking the scaling mupdf does may be losing precision the other way though... | 12:24.01 |
| we lose precision when setting the font size, but keep max precision in the outlines | 12:24.18 |
| bah, I forgot to extract the quantized e,f values from the subpixel adjusted matrix. that explains the different rounding causing some glyphs to be off. | 12:38.50 |
kens | Robin_Watts : ping | 13:56.15 |
Robin_Watts | pong | 13:56.19 |
kens | I want to use mutool clean -d for a single page of a PDF file | 13:56.31 |
| But if I specify just, say, 1 as the page number it gives me a PDF with 2 pages | 13:56.49 |
Robin_Watts | kens: You want to clean a single page of a pdf, and leave the rest intact ? | 13:56.54 |
kens | The second of whch is invalid | 13:56.55 |
Robin_Watts | Can't do that. | 13:56.57 |
kens | No I want to just pull 1 page out and decompress it | 13:57.06 |
Robin_Watts | mutool clean -d in.pdf out.pdf 1 | 13:57.14 |
kens | Yeah givres me a PDF with 2 pages | 13:57.23 |
Robin_Watts | If that's giving you a 2 page pdf at the end that's a bug. | 13:57.29 |
| Can we have a copy of the pdf please? | 13:57.33 |
kens | Where page 2 is the same object as page 1, but broken | 13:57.34 |
Robin_Watts | How up to date is your mupdf ? | 13:57.55 |
kens | You almost certainly don't really want the PODF but if you really want to look its bug #696567 | 13:57.56 |
| This is 1.8 | 13:58.02 |
| The file is nearly 1Gb declompressed | 13:58.12 |
Robin_Watts | from git or from the release? | 13:58.14 |
kens | From the release, I tested it on WIndows for Tor | 13:58.29 |
Robin_Watts | ok, so I can't blame tor8's fix from the other day. | 13:58.45 |
kens | Probably not :) | 13:58.57 |
| I'm just trying it with page 7 now since that's the (barking mad) page I actually need | 13:59.14 |
| And this time it came out with only 1 page | 13:59.33 |
| Must have been something I did wrong | 13:59.53 |
Robin_Watts | kens: phew. | 14:00.03 |
kens | Of coruse, its *still* nearly 1Gb :-( | 14:00.07 |
Robin_Watts | kens: Use -dif | 14:00.17 |
| That avoids uncompressing images and fonts | 14:00.24 |
kens | aha excellent plan | 14:00.32 |
| 300 Mb should be able to open that.... | 14:00.59 |
| bah but its the wrong page, need to try again | 14:01.38 |
| Oh yeah 447000 ExtGstates on one page | 14:05.56 |
| Almost all of which are : | 14:07.14 |
| << | 14:07.14 |
| >> | 14:07.14 |
| endobj | 14:07.14 |
| Bah | 14:07.18 |
| "<< | 14:07.23 |
| " /Type /ExtGState | 14:07.24 |
| ">> | 14:07.24 |
| "endobj | 14:07.24 |
| ie stupidly pointless | 14:07.30 |
chrisl | I'm now regretting letting Win7 update itself on a reboot. >10 mins and "Installing update 7 of 35"....... :-( | 14:12.54 |
| And now I have a laptop that won't boot..... f*ck!! | 15:26.43 |
kens | Probably 'upgraded' you to Windows 10 | 15:26.59 |
chrisl | It's just sat at "Starting Windows" and churning the disk..... | 15:27.55 |
Robin_Watts | chrisl: arse. let it churn. | 15:50.16 |
chrisl | Robin_Watts: I don't have much choice - keeping fingers crossed. It would be more worrying if the disk wasn't churning | 15:51.14 |
Robin_Watts | My PC blue screened repeatedly yesterday. I'm hoping it's a duff graphics card driver so I can update out of it when they release a new one. | 15:52.00 |
kens | has a new graphcis card to install, a job for the weekend | 15:53.10 |
chrisl | Aha! Laptop's back.... still, that can't be good - 50 minutes to boot up | 16:21.40 |
marcosw | chrisl: When I was young we used to _dream_ of it only taken 50 minutes to boot a laptop. You kids today don't know how lucky you have it. | 16:24.17 |
chrisl | marcosw: if it was *supposed* to take that long, it wouldn't bother me at all | 16:25.13 |
marcosw | It's windows, it will decide how long something will take and you will like it! Could be worse, Windows 8 won't let you shutdown without installing pending updates. | 16:26.53 |
malc_ | spent considerable amount of time lately trying to figure out why his machine takes so long too boot... 9 bloody seconds, shaved off only 2 seconds so far | 16:27.01 |
marcosw | never understood why windows users are so fixated on boot times. | 16:29.14 |
chrisl | Perhaps because, traditionally they've had to do it so often?? | 16:29.52 |
HenryStiles | marcosw: cause it keeps crashing? | 16:29.58 |
malc_ | is linux user though | 16:30.52 |
Robin_Watts | malc_: Never mind, we all have our crosses to bear. | 16:31.14 |
marcosw | OTOH, I seem to be obsessed with compile times. Have gotten it down to 7.9 seconds for Ghostscript/PCL/XPS by using ccache and bcache with a ridiculously fast M.2 drive. | 16:35.57 |
malc_ | marcosw: bcache? | 16:36.29 |
marcosw | Block layer cache for LInux. Similar to a Mac OS X Fusion drive. DOn't know what they are called on Windows. | 16:37.29 |
| Intel SRT? | 16:37.58 |
malc_ | marcosw: ssd in front of hdd thingie? | 16:38.06 |
marcosw | yeah. but not ssd anymore, M.2 | 16:38.18 |
| And not that old fangled SATA M.2, PCI-E M.2 | 16:39.24 |
malc_ | marcosw: m.2 = https://en.wikipedia.org/wiki/M.2 ? | 16:39.35 |
marcosw | yeah. | 16:40.08 |
malc_ | thank you | 16:41.06 |
marcosw | the hot setup is to use two M.2 cards in a RAID0 config. But my motherboard only has one M.2 socket and I don't want to bother with an adapter card. | 16:41.24 |
malc_ | everything is soldered here and completely irreplacable here :) makes me happy | 16:46.03 |
| Forward 1 day (to 2016/02/12)>>> | |