| | <<<Back 1 day (to 2012/12/03) | 2012/12/04 |
| Robin_Watts | heads off then. | 06:52.52 |
| kens | tor8 ping | 09:49.47 |
| tor8 | kens: hi | 09:49.55 |
| kens | could you take a quick look at this one please ?: | 09:50.07 |
| | http://stackoverflow.com/questions/13689129/mupdf-limits-h-no-such-file-or-directory-in-mac-os-x | 09:50.07 |
| | Is it an out of date version of FreeType causing teh problem ? | 09:50.33 |
| tor8 | kens: more likely an out of date version of Xcode or missing "SDK" packages for it | 09:50.52 |
| kens | Ah, could be, I'm afraid I know nothgin about XCode | 09:51.06 |
| tor8 | IIRC Xcode doesn't ship system headers by default anymore, you have to install them separately | 09:51.10 |
| kens | Hmm, he claimsd limits.h is installed | 09:51.22 |
| | path problem perhaps ? | 09:51.39 |
| tor8 | right. 10.7.5 is old by now, it may be that the paths have changed | 09:51.46 |
| | I think I had to update the build scripts to get the paths from different env variables when I upgraded a while ago | 09:52.03 |
| kens | OK I'll post that info, thanks | 09:52.12 |
| tor8 | kens: ios/build_libs.sh is the script that's used to invoke the external make target in the Xcode project | 09:52.42 |
| kens | OK makes sense. It looks like he can't build FreeTtype though | 09:53.08 |
| tor8 | kens: it worries me though, people are starting to use the iOS version... apple's app store doesn't allow GPL. | 09:53.13 |
| kens | Maybe we need a sterner warning | 09:53.27 |
| tor8 | kens: we build third party libraries ourselves | 09:53.32 |
| | oh wait! | 09:53.46 |
| | I didn't read his question properly, he's *not* using the Xcode project | 09:54.06 |
| kens | Note though that he says he wants it for Android, and is using iOS to generate the 'generated' files | 09:54.06 |
| tor8 | so ignore what I said about ios/build_scripts | 09:54.31 |
| kens | :-) | 09:54.35 |
| tor8 | does this line look fishy to you too? | 09:55.32 |
| | ./Developer/usr/bin/../lib/gcc/i686-apple-darwin9/4.2.1/include/limits.h:10:25: error: limits.h: No such file or directory | 09:55.35 |
| kens | THe error says that limits.h is not found in a specific location | 09:55.39 |
| | tor8 yes, that'ds what I meant | 09:55.46 |
| tor8 | well, he can't find string.h or stdio.h either so I think he's got a broken install | 09:56.26 |
| kens | broken gcc ? | 09:56.38 |
| tor8 | must be something like that; broken install, missing bits from the install, or broken paths | 09:57.37 |
| kens | OK I'll post some words, thanks. | 09:57.50 |
| tor8 | kens: I do all my android development on mac, so I know it works without doing anything fancy to the install | 09:58.36 |
| kens | OK thanks tor8 | 09:58.45 |
| | I'll add that too | 09:58.50 |
| paulgardiner | tor8: ping | 10:34.57 |
| tor8 | paulgardiner: hi | 10:42.01 |
| paulgardiner | Was just looking at this page-per-doc problem. | 10:42.31 |
| | It might not be hard to knock something up, although slightly wasteful in keeping 3 mupdf instances. Interested to know your thoughts though, and also how much effort we should expend on this. | 10:43.47 |
| tor8 | paulgardiner: somehow I think it'd be easier to knock it up in C | 10:55.41 |
| | just another fz_document that multiplexes, and keeps open only one sub-fz_document at a time (or a 3-slot cache of last recently accessed documents) | 10:56.50 |
| | the page counting might get tricky though | 10:57.04 |
| | so maybe I'm wrong and it's easier in the android app | 10:57.32 |
| paulgardiner | I guess it's influenced by which bits of code we know best. I can see quite clearly how to alter the java | 10:57.52 |
| tor8 | then go ahead :) | 11:01.09 |
| | also, given his griping about lack of docs, maybe we should make very stripped to the bare bones sample viewers without all the fancy features, for doc purposes? | 11:01.53 |
| | well, that's for later anyway | 11:02.03 |
| paulgardiner | Yes, that would be good | 11:02.27 |
| | Are we usually this forthcoming with help where someone isn't offerring to pay for changes? | 11:03.48 |
| tor8 | paulgardiner: not really, no. | 11:05.01 |
| | paulgardiner: so don't spend too much time on it | 11:05.26 |
| paulgardiner | Ok. I'll have a quick stab at it and see how it goes. | 11:06.05 |
| henrys_mac | The dreaded tech agenda complete. | 14:42.35 |
| paulgardiner | tor8: I have something sort of working. Do we want to stick this on a branch, or should I just email the diffs to him? | 15:12.50 |
| tor8 | the latter, unless it's something you think is worthwhile to keep around for general use (which I guess you don't) | 15:27.27 |
| paulgardiner | Mmm. Probably not. | 15:28.35 |
| henrys | tor8, paulgardiner :anything for the meeting this time around? Robin is traveling | 15:57.49 |
| paulgardiner | Nothing I can think of. | 15:58.51 |
| henrys | paulgardiner:it would be good if you were around/available as much as possible in case Michael or Robin have demo issues. | 15:59.23 |
| kens | henrys I've been sent a PCL file which gives 'bad' results with pdfwrite, but the glyphs render badly in ghostpcl too. If I send it to you can you try it on a HP pritner and see what it does ? I think the problem is the hints are broken, and I wonder if HP pritners ever do hinting of TT fonts. | 16:00.31 |
| henrys | kens:ah interesting send it along | 16:00.56 |
| kens | On its way | 16:01.01 |
| paulgardiner | henrys: Sure, although I'm out this evening between 8pm and 11:30 | 16:01.16 |
| henrys | paulgardiner:that should be fine. I assume you know about the staff meeting so things will be quiet for a few days IRC wise | 16:03.13 |
| paulgardiner | Ah right. Maybe I should post my mobile number to tech for when I'm not near the talker. I'm happy to be contacted any time. | 16:04.11 |
| henrys | paulgardiner:you can send it to Joanne and she should publish a new contact sheet I guess we didn't do that for you. | 16:06.00 |
| paulgardiner | Ok. Will do | 16:07.08 |
| henrys | but yes we should have a cell number for you and you should have our info too, we agreed to that as a company policy some time ago and we must have let you slip through. | 16:07.32 |
| tor8 | henrys: no, I'm good for now | 16:07.56 |
| henrys | okay meeting adjourned | 16:08.39 |
| kens | are we having a tech meeting this week ? | 16:08.51 |
| paulgardiner | henrys: Is Joanne's email what I'd guess it to be? | 16:10.12 |
| kens | joann.owens | 16:10.29 |
| henrys | I'm thinking we can cancel it, Robin and Michael aren't there | 16:10.30 |
| paulgardiner | ta | 16:10.33 |
| henrys | s/there/here | 16:10.42 |
| kens | And we are all meeting Friday anyway | 16:10.50 |
| | I'll be here anyway though, fmaily is at stables until late | 16:11.11 |
| | Hi mvrhel_laptop how's it going ? | 16:11.22 |
| mvrhel_laptop | hi kens. good | 16:11.53 |
| | going to sit in on this embedded android workshop today | 16:12.06 |
| kens | show starts today ? | 16:12.08 |
| mvrhel_laptop | well there are some workshops today. tomorrow we set up the booth | 16:12.33 |
| henrys | with my new power line networking setup I have to go downstairs to get a print out ⦠good exercise. | 16:12.33 |
| mvrhel_laptop | and then thursday and friday we man the booth | 16:13.02 |
| kens | I always have to go downstairs to get printouts, the printer is on the family PC.... | 16:13.08 |
| mvrhel_laptop | when does Robin_Watts arrive? | 16:14.13 |
| kens | tonight our time i THINK | 16:14.27 |
| | assuming its the same time as our flight, 14:15 local time | 16:14.49 |
| | so will be with you at the hotel around 4pm I would think | 16:15.03 |
| mvrhel_laptop | ok. I will take him out for a milkshake | 16:15.20 |
| kens | suspects he will appreciate that | 16:15.32 |
| henrys | kens:the HP rendering is "expected" ours is not. | 16:16.07 |
| kens | henrys that is indeed interesting | 16:16.18 |
| | I'll need to consult with chrisl but that sounds to me like the HP is not using hints in TT fonts | 16:16.35 |
| | Mayeb we need a way to (at least optionally) disable the hints | 16:16.51 |
| | Of course, pdfwrite will still emit the font embedded, and Acrobat will still display it using the broken hints.... | 16:17.14 |
| henrys | so this is a regression with the free type + pcl | 16:17.29 |
| kens | In essence yes, because we never used to use hinting | 16:17.40 |
| | (I think) | 16:17.47 |
| | Of course, with properly hinted fonts, the result should be better with hinting turned on.... | 16:18.15 |
| henrys | If it isn't then you should probably make a pcl bug and I'll look at it first it could be something else. | 16:18.24 |
| kens | henrys, I'll talk with chrisl about ti first. | 16:18.40 |
| henrys | okay | 16:18.47 |
| kens | I've just discovered that pdfwrrite is actually emitting type 3 fotns, not TT which is a surprise | 16:19.01 |
| henrys | are they bitmaps? | 16:20.03 |
| kens | In the pdfwrite output ? Yes I believe so, I'm just checking | 16:20.18 |
| | Yes, they are | 16:20.57 |
| henrys | kens:did chrisl have a switch to disable the new font scaler stuff? | 16:20.58 |
| | I thought he did | 16:21.08 |
| kens | henrys, I'm not sujre, presumably I can still se -dDisableFAPI | 16:21.15 |
| henrys | mvrhel_laptop: we decided to cancel the meeting | 16:21.44 |
| kens | yes, with -dDisableFAPI, the results with pdfwrite are as I expect | 16:21.53 |
| henrys | oh good ;-) | 16:22.05 |
| kens | the GhostPCL rendering seems to be OK too | 16:22.35 |
| | So I'll talk to chrisl about it when he comes back online | 16:23.09 |
| | Speak of the devil | 16:24.19 |
| | chrisl, can you look at the last bit of the logs please | 16:24.33 |
| henrys | let me know if you want a simple file. | 16:24.50 |
| mvrhel_laptop | henrys: Thanks. I figured that was the case | 16:25.33 |
| chrisl | kens: so what is the actual problem? | 16:26.18 |
| kens | chrisl it looks like the font has broken hinting | 16:26.30 |
| chrisl | "broken" or "poor"? | 16:26.53 |
| kens | THe old non FAPI code, and an HP printer do not exhibit a problem (suggesting they are not hinting the glyphs) the new code shows malformed glyphs | 16:26.57 |
| | chrisl defintiely malformed | 16:27.11 |
| | and highly symptomatic of incorect hints | 16:27.22 |
| | also possibly a non-zero versus even-odd winding rule problem | 16:27.36 |
| chrisl | kens: is this with pdfwrite, or other devices, too? | 16:27.36 |
| henrys | but a bitmap font wouldn't have hints | 16:27.43 |
| kens | chrisl pdfwrite is embedding bitmaps, and the display is incorrect. | 16:27.59 |
| | henrys, the bitmaps are caused by rendering the glyphs in GhostPCL | 16:28.10 |
| | Instead of embedding a TT font | 16:28.29 |
| | We are going through the 'fallback' code, I don't know why at the moment | 16:28.40 |
| chrisl | Well, we can't spot poorly hinted glyphs. | 16:28.56 |
| kens | Possibly we don't have enough of a TT font to succesfully embed it | 16:28.58 |
| | chrisl I was thinking that we might need a way to turn off hinting | 16:29.11 |
| | Possibly even *always* render TT fonts in PCL unhinted | 16:29.35 |
| henrys | kens:where did you get that pcl file? | 16:29.51 |
| kens | Note that -dDisableFAPI works 'as expected' | 16:29.52 |
| chrisl | Well, I did ask whether HP printers honoured ttf hinting a couple of weeks ago | 16:30.09 |
| kens | henrys Scott forwarded it to me frommeone asking about PCL->PDF conversion and text extraction | 16:30.26 |
| | chrisl looks like the answer is 'no' | 16:30.36 |
| | But it might be best if you look at the font and satisfy yourself that it is broken hinting and not something else. | 16:30.56 |
| | I will sned the file if you like | 16:31.02 |
| | send* | 16:31.08 |
| henrys | yeah but it might just be a particular instruction is ignored. | 16:31.09 |
| kens | henrys I see at least two distinct problems | 16:31.25 |
| chrisl | kens: thanks. | 16:31.29 |
| | kens: I take it you used the display device to preview it? | 16:31.53 |
| kens | chrisl yes indeed | 16:32.03 |
| | file on the way | 16:33.00 |
| chrisl | I *think* freetype still has the "non-patented instructions only" setting, we *might* be able to use that..... | 16:33.13 |
| kens | Worth a try maybe, see what you think | 16:33.28 |
| henrys | The fact that PCL is almost always 600 dpi and quality expectations aren't as high in business graphics (pcl's world) and performance is premium is probably reason enough to turn it off. But it would be interesting to see what's going on with this file. | 16:33.46 |
| chrisl | henrys: the problem is the not-inconsiderable number of fonts out there use the bytecode for things other than hinting...... | 16:34.56 |
| kens | ROFL : | 16:35.28 |
| | http://www.bbc.co.uk/news/magazine-20536090 | 16:35.28 |
| henrys | chrisl:If you have sample fonts I can convert them to download pcl fonts and have a go on the printer. | 16:39.30 |
| chrisl | henrys: they are embedded in tests, and I can't remember which tests. If we have to, I can probably hack a font | 16:41.03 |
| henrys | okay | 16:41.24 |
| | I remember there was an entire collection of fonts that infamously abused hints but I can't remember which ones. | 16:42.27 |
| chrisl | henrys: Dynalab? | 16:44.05 |
| henrys | yes that's it | 16:44.12 |
| | ray_laptop: we canceled the meeting | 16:44.29 |
| kens | THIs one says (C) Microsoft | 16:44.41 |
| chrisl | MS has shipped fonts with broken hints, too :-( | 16:45.06 |
| ray_laptop | henrys: OK, thanks for letting me know. | 16:58.23 |
| chrisl | kens: yeh, definitely hinting. I'm not sure whether the hinting bytecode is broken, or it's a bug in freetype...... | 17:00.27 |
| kens | classic symptoms | 17:01.06 |
| chrisl | kens: FWIW, U F S T's rendering agrees with freetype's, so that's strong evidence that the font is wrong | 17:26.50 |
| kens | And also that HP printers ddo not apply hints | 17:28.32 |
| chrisl | kens: and, btw, these seem to be "XL" format embedded TTF's, so they have "required" tables missing, which is probably why pdfwrite type 3s them | 17:29.16 |
| kens | chrisl yes I was suspicious this was the case, that's what I was alluding to above. | 17:29.44 |
| chrisl | Yeh, I just thought I would confirm it | 17:30.01 |
| | There some other hinting issues in our regressions tests I need to investigate - I'm wondering if they are all XL fonts...... | 17:30.50 |
| kens | Could well be | 17:30.58 |
| chrisl | There's a problem with providing a command line option to disable hinting, because we don't get a pcl context passed into where I would need it to pass it into FAPI :-( | 17:31.51 |
| kens | Hmm, that's awkward | 17:32.12 |
| chrisl | That's another reason I implemented "DisableFAPI" in the rather hacky fashion I did | 17:32.48 |
| henrys | moving to the coffee shop bbiab | 17:51.29 |
| | Forward 1 day (to 2012/12/05)>>> | |
IRC Logs