| <<<Back 1 day (to 2014/12/01) | 20141202 |
termi525 | Hi all! o/ I have a mupdf /android related question. i have a sigsegv while using the updatePage method or at least it happens after i used it. I cannot find an entry point or where to look for the error. Used Version: mupdf 1.6 App Min SDK 14. | 11:47.12 |
| can someone help me? | 11:47.22 |
paulgardiner | Difficult to know what to suggest. There are so many possible causes. | 11:51.40 |
termi525 | sure. i can provide the log and some methods i use, if this helps | 11:55.45 |
Robin_Watts | termi525: As a rule, we don't get involved in debugging other peoples code. | 11:56.02 |
| We simply have too much to do. | 11:56.07 |
| If you can give us the smallest possible example that shows the problem, then you maximise your chances of us taking a look :) | 11:56.58 |
| Are you a supported customer, or a free user? | 11:57.06 |
termi525 | yeah i understand but the problem here is: its my fault / my code that causes this but it crashes in the ndk code. Sadly i'm not a supported customer :( | 11:58.18 |
Robin_Watts | termi525: Well, I would isolate your problem down to the simplest skeleton that can show it happening. | 12:18.52 |
| open a document, ..., update_page, ... crash | 12:19.23 |
termi525 | okay. the only thing i can tell is: rendering a normal page works like a charm, even 10 pages at once. it only happens during a page update and then switching / loading new pages. this ends almost guaranteed in a crash. | 12:23.23 |
tor8 | kens: Robin_Watts: paulgardiner: we're staying at the hilton on hyde park right? | 13:17.36 |
kens | correct | 13:17.44 |
tor8 | should be easy enough to get to, less than a kilometer walk from paddington station | 13:18.07 |
kens | Not the doubletree, the real hilton, right next door ro Queensway tube station | 13:18.08 |
tor8 | probably faster than navigating the tube :) | 13:18.25 |
Robin_Watts | tor8: yes. it's not a long walk, depending on your luggage. | 13:18.33 |
kens | Possibly yes | 13:18.34 |
tor8 | kens: oh, the doubletree is right next door | 13:19.14 |
| me? luggage? :) | 13:19.25 |
kens | Its pretty close, yes | 13:19.32 |
Robin_Watts | tor8: The doubletree is 200m away or so, I think. | 13:19.37 |
| close enough for it not to be a huge problem if you end up at the wrong one by mistake. | 13:20.07 |
tor8 | are we meeting up for dinner tomorrow night? | 13:23.31 |
kens | Not me, I',m coming up Thursday morning | 13:23.50 |
| But there will be other people there | 13:24.14 |
jogux | tor8: pete & I will arrive at hotel at 17:30 or so hopefully, Henry at least should arrive before us, plan is to convene on IRC and try to form a plan for dinner | 14:03.49 |
| I'm presuming we can get away without booking on a Wednesday evening. | 14:04.27 |
kens | I'm sure McDonalds can find you something :-) | 14:04.54 |
tor8 | jogux: sounds good. | 14:11.12 |
jogux | don't know quite how zonked the americans will be ;) | 14:11.30 |
tor8 | jogux: probably very, if they're getting in on wednesday! | 14:11.48 |
| I know how I am flying eastwards... | 14:11.55 |
Robin_Watts | Paul/Linda/Helen/I won't be there until more like 10pm, so I suspect we'll see people for breakfast the next morning. | 14:12.04 |
| Usually it's 8:30am for brekkie, but sometimes Miles thinks that 8 is a better idea. | 14:12.29 |
kens | I was assuming meeting start at 9:30, I won't make it to the hotel until 9 (assuming no problems) | 14:13.20 |
henrys | jogux tor8 : yeah I'll start looking at irc in the late afternoon. 5 or so. | 14:24.59 |
| not likely to be my usual chipper self ;-) | 14:27.43 |
jogux | :-) | 14:30.48 |
kens | Aha! I believe that tests 29-07I.ps and 29-07Kps are currently incorrect with the psdcmyk device. So my changes are in fact progressions :-) | 14:31.22 |
chrisl | jogux, tor8, henrys: I should be up in London for late tomorrow afternoon, too..... | 14:38.56 |
henrys | so the best plan I guess is try and stay awake until regular bed time UTC time. right? asking the seasoned experts. | 14:41.13 |
kens | That's what I find the best approach, yes | 14:41.27 |
jogux | henrys: yeah. albeit I rarely make it past 8pm. | 14:41.55 |
kens | I usually make 10 pm | 14:42.08 |
tor8 | henrys: an hour or two earlier than usual, to recover a bit | 14:42.39 |
| but whatever you do, don't take a nap in the afternoon! | 14:43.12 |
henrys | I was thinking I should spend the afternoon outside in the sun but then I thought well outside should be good enough... | 14:44.59 |
jogux | hehe. at least forecast is dry for wednesday | 14:46.03 |
kens | Sunday was nice, we had some sun | 14:46.23 |
chrisl | Just wrap up warm! I always find I'm susceptible to colds etc after a long flights | 14:46.31 |
jogux | looks like there may be a bit of sun on Saturday. but it's going to be cold. | 14:47.06 |
kens | chrisl could you look at 29-07k.ps with the psdcmyk device ? Pages 5 & 6 look incorrect to me (all the framing is missing for example) and the text on page 7 is yellow, which also looks incorrect | 14:52.05 |
chrisl | kens: I'm not sure I have a way to view PSD files.... let me check..... | 14:52.46 |
kens | Ah, I wonder what Marcos uses for the cluster | 14:52.58 |
chrisl | I think bmpcmp decodes psd itself | 14:53.38 |
kens | Oh, I've been using a very elderly copy of Photoshop | 14:53.51 |
| Well I'm reasonably sure the current output is incorrect, maybe I can ask Ray in a bit, he might have a Photoshop to look at it with | 14:55.34 |
chrisl | I'm just going to check a Windows VM, I might have a version of Photoshop..... | 14:56.31 |
Robin_Watts | chrisl: There is an online psd viewer... | 14:56.34 |
kens | I can put Photoshop on my laptop and discuss it on Thursday, I'm just looking for a second opinion really | 14:57.26 |
Robin_Watts | oh, no. sorry, not psd :( | 14:57.41 |
henrys | gimp for windows is free? | 14:58.17 |
kens | Does it open .psd ? I don';t have it installed on this machine | 14:58.32 |
chrisl | I don't think gimp opens PSD, and certainly doesn't handle CMYK...... | 14:58.45 |
kens | its worse than that, its CMYK+spots, potentially | 14:59.06 |
chrisl | kens: pages 5 and 6 I only see the color bars, no other markings at all | 14:59.21 |
kens | Yeah that's what I see, and I think its incorrect. | 14:59.31 |
paulgardiner | I think GIMP does open PSDs but doesn't allow editing of all that's in the file | 14:59.31 |
kens | Also on page 7 the black content is rendered in yellow | 14:59.44 |
| According to the gimp website it handles .psd | 15:00.13 |
chrisl | Yes, indeed. Those are definitely wrong, compared to other devices | 15:00.21 |
henrys | pretty sure I've looked at psd's in the gimp but don't know about spot and cmyk | 15:00.21 |
kens | chrisl, then I'll assume the fact that my new output matches the otehr deives is in fact correct :-) | 15:00.48 |
chrisl | "GIMP Message: Opening '/home/cliddell/artifex/ghostpdl/gs/debugbin/stuff001.psd' failed: Error loading PSD file: Unsupported colour mode: CMYK" | 15:01.15 |
kens | Something I;ve done changed the output, but I've made so many changes I'm not sure what it was. | 15:01.22 |
| Yeah, I thought it wouldn't like CYK | 15:01.30 |
| CMYK* | 15:01.35 |
chrisl | It seems insane since the lack of CMYK and spot color support are probably the biggest impediments to using gimp in a pro setting | 15:02.34 |
kens | Yes, absoltuely true | 15:02.45 |
| Wihtout that, it has limited use | 15:03.00 |
Robin_Watts | chrisl: Will imagemagick read psd? | 15:03.20 |
henrys | google says: http://www.openmutual.org/2011/08/converting-cmyk-psd-files-for-use-in-gimp/ | 15:03.35 |
chrisl | Robin_Watts: Yes, but gets the CMYK colors wrong..... | 15:03.36 |
Robin_Watts | If so, convert from psd to pam, and then use: http://ghostscript.com/~robin/pamview.html | 15:03.53 |
| If you convert to pam it'll keep stuff as CMYK and so the colors don't matter. | 15:04.13 |
chrisl | My antique Photoshop did the job..... | 15:04.25 |
Robin_Watts | That viewer will let you turn on/off individual channels. | 15:04.27 |
kens | THe website henry pointed to suggests IM will open them | 15:04.27 |
chrisl | I only tried viewing with IM, I didn't try any of the conversions.... | 15:04.58 |
henrys | we should have a company subscription to CS. | 15:05.08 |
kens | For what I need to use it for, my old Photoshop works fine. | 15:05.41 |
chrisl | It's (I assume) quite a lot of money for the rare occasions we actually need it | 15:05.52 |
paulgardiner | Isn't CS3 free? | 15:07.09 |
kens | is doubtful... | 15:07.25 |
| Adobe doesn't give useful stuff away for free on the whole | 15:07.36 |
henrys | paulgardiner: you mean on piratebay? | 15:07.36 |
chrisl | There's a maybe an "Elements" version, or something? | 15:08.00 |
kens | Yeah, Melanie got a clopy of that with her graphics tablet | 15:08.19 |
jogux | paulgardiner: technically it's able to be activated without a purchased license, which isn't quite the same as free :-) | 15:08.20 |
paulgardiner | :-) No. When we were messing with the splash screen, we discovered that one of the old versions was avaialble free | 15:08.20 |
| Ah that might be it | 15:08.39 |
kens | Hmm, well amazon is selling them at £399 | 15:08.45 |
jogux | basically they switched off the CS3 license server, so made the non-license server protected downloads available to everyone iirc. | 15:09.02 |
paulgardiner | It was a very old copy | 15:09.04 |
kens | My Photshop is *much* older :-) | 15:09.19 |
jogux | technically you were only meant to use them if you had a valid license ;-) | 15:09.21 |
kens | At least mine is legal.... | 15:09.38 |
wordToDaBird | Who is the person that wrote/writes the Android wrapper for mupdf? | 15:35.30 |
Robin_Watts | wordToDaBird: Originally me, much improved by paulgardiner, and tweaked further by mattchz. | 15:38.13 |
wordToDaBird | ReaderView.getSelectedView(ReaderView.java:660) | 15:40.27 |
| at android.widget.AdapterView.onInitializeAccessibilityEvent(AdapterView.java:951) | 15:40.27 |
| at android.widget.AdapterView.onRequestSendAccessibilityEvent(AdapterView.java:926) | 15:40.27 |
| at | 15:40.27 |
| java.lang.UnsupportedOperationException: Not supported | 15:40.43 |
| I checked the source and ReaderView.getSelectedView throws an UnsupportedOperationException | 15:41.13 |
Robin_Watts | It does. | 15:41.24 |
wordToDaBird | I assume its called in the C++ code somewhere but haven't checked, when/why | 15:41.33 |
| what's getSelectedView a placeholder for? | 15:42.04 |
Robin_Watts | The C++ doesn't call it. | 15:42.29 |
| Android calls it. | 15:42.40 |
wordToDaBird | I didn't notice it being called anywhere in the source. | 15:42.45 |
| ohhh | 15:42.46 |
| why? | 15:43.11 |
Robin_Watts | possibly the right thing to do is simply to not throw, but I haven't had time/the inclination to read up on the the circumstances in which it is called. | 15:43.15 |
wordToDaBird | yea, I thought about that, but I figure if its being thrown there is a reason. | 15:44.09 |
| I am trying to discern why | 15:44.14 |
| why does Android call it, the entire method is empty other than the throw | 15:44.47 |
Robin_Watts | wordToDaBird: I don't know. You'd need to do some reading. | 15:54.19 |
wordToDaBird | Robin_Watts have suggestion on what/where/when/??? Just point me in the direction with as much info as you can. | 15:56.06 |
Robin_Watts | wordToDaBird: No suggestions other than to google the elements of that backtrace. | 15:56.30 |
henrys | Robin_Watts: wow looks like good progress on the 64 bit changes. | 15:57.27 |
Robin_Watts | henrys: Windows appears to run (at least superficially) | 15:57.43 |
| linux and ios aren't building yet, but it's progressing. | 15:58.09 |
henrys | Robin_Watts: I guess android 64 is coming soon, there are phones with 64 bit hardware but I read google has only put out an android preview 64 and it's x86 | 16:11.44 |
| not arm | 16:11.51 |
Robin_Watts | henrys: Android L is out for some(*) phones now. | 16:12.02 |
| (* where some == at least one) | 16:12.14 |
| and those are ARM phones. | 16:12.20 |
| oh, sorry, 64, not L. | 16:12.34 |
| The phone I bought the other day (LG G3) has a 64bit processor in it, but runs a 32bit android at the moment. | 16:14.35 |
| It's only got 3Gig of RAM (and that's more than most phones out there), so it doesn't need 64bit for that. | 16:15.14 |
henrys | Robin_Watts: right I thought all the 64 bit phones were still running 32 android, but I'm probably not up to date. | 16:16.55 |
Robin_Watts | henrys: I think you're right. | 16:17.11 |
| My point is that I'm not sure 64bit Android will actually help with the current generation of phones. | 16:17.36 |
| It's only when you need 1 process to be able to access more than about 2Gig of RAM that 64bits becomes compelling. | 16:18.52 |
henrys | Robin_Watts: agreed so are you doing that concurrently now? the android specific stuff or waiting for a good platform? | 16:19.48 |
jogux | wordToDaBird: getSelectedView is part of the Android View API, or something like that | 16:19.57 |
Robin_Watts | henrys: Android L is done. We're on Google Play and Samsung Apps in Android L compatible form. | 16:20.14 |
jogux | oops, robin already answered. my IRC stopped scrolling for some reason. | 16:20.42 |
Robin_Watts | 64bit Android can wait til we get the platforms which are actually shipping 64bit sorted first :) | 16:20.54 |
jogux | robin_watts: still amazon to put android L onto. | 16:20.57 |
Robin_Watts | jogux: I've done the amazon build. | 16:21.09 |
| I haven't uploaded it cos I haven't had the strength to properly test it. | 16:21.22 |
| Yes, I am crap. | 16:21.35 |
jogux | ah :-) | 16:21.41 |
Robin_Watts | And do we have the amazon log in details? | 16:21.48 |
henrys | Robin_Watts: I doubt google and company will be as demanding as apple | 16:21.56 |
jogux | Paul was asking that the other day. I presume he found them. | 16:21.57 |
Robin_Watts | henrys: I would imagine that 64bit android will be "easy" once the others are done. | 16:22.28 |
jogux | indeed | 16:22.38 |
paulgardiner | I have the login details | 16:22.42 |
Robin_Watts | paulgardiner: Did you update the file on casper with them? | 16:23.05 |
jogux | robin: I read somewhere that ARM64 is more a pull than AMD64, there's advantages beyond 64bit addressing. but I forget the details :-) | 16:23.14 |
| (certainly Apple are pushing it hard despite the devices only having 1G or 2G of RAM...) | 16:23.27 |
paulgardiner | Robin_Watts: No. I didn't know of the file on casper | 16:23.43 |
Robin_Watts | see skype. | 16:23.56 |
paulgardiner | I can do that now | 16:23.57 |
jogux | I'll feel a lot happier about 64 bit once we've got some bitmaps out of ATS. | 16:25.33 |
| which is hopefully close. | 16:25.44 |
henrys | jogux: ars reports 30% faster, that's huge: http://arstechnica.com/apple/2014/10/apple-beginning-february-2015-app-store-submissions-need-to-be-64-bit/ | 16:26.14 |
paulgardiner | Updated | 16:26.58 |
Robin_Watts | I haven't read up on the ARM64 architecture in depth (I'm taking the view that if I ignore it, it'll go away). | 16:27.57 |
| I suspect they've done away with the lovely "condition codes in the top nibble of instructions" stuff, which is hard for highly pipelined implementations. | 16:28.55 |
| and hard for compilers to target efficiently, cos compilers are crap. | 16:29.14 |
henrys | I do wonder if that is just peculiar to ios code and android will not see similar improvements. That's a much large improvement than expected. | 16:29.44 |
Robin_Watts | dunno if ARM64 has the jazelle stuff in... | 16:30.11 |
jogux | henrys: I'd vaguely suspect that clang/llvm used for iOS are way better at ARM64 than the gcc used on Android.... | 16:30.59 |
Robin_Watts | More registers on ARM64 (because of the ditching of the condition codes, apparently) | 16:31.31 |
henrys | jogux, Robin_Watts even so ... 30% wow... | 16:32.11 |
Robin_Watts | Aha, and jazelle is in ARMv8 (the 64bit stuff). Which means android might gain more, as it's more dependent on java. | 16:33.02 |
henrys | starts packing up... see you guys soon | 16:34.47 |
Robin_Watts | henrys: see you! | 16:35.07 |
chrisl | I've begin my preparations, too - I've just restrung a guitar..... ;-) | 16:37.37 |
Robin_Watts | chrisl: Are you bringing a buitar? | 16:39.02 |
| s/b/g/ | 16:39.04 |
jogux | henrys: have a good flight etc :) | 16:39.08 |
chrisl | Robin_Watts: Yeh, I'm bringing my Faith acoustic for Miles and Scott to have a shot at | 16:39.39 |
chrisl | thinks: I'm sure the hotel will be *thrilled* when I rock up carrying a guitar! | 16:41.02 |
Robin_Watts | ah, acoustic. I won't bother bringing the rocksmith cable then :) | 16:41.05 |
chrisl | Robin_Watts: ah well, I wouldn't want to embarrass myself, anyway.... | 16:41.56 |
kens | Hmm, looks like our Times-Roman doesn't contain a /gbreve glyph | 16:52.40 |
chrisl | I can't remember if that a glyph we listed as missing | 16:53.33 |
kens | Me neither, but if I try to 'glyphshow' it I get nothing | 16:53.52 |
| Looks like the glyphs that the free user is complaining about are all missing. Of course, if he embedded the font...... | 16:54.19 |
chrisl | Hmm, I don't have it as "missing" from Times-Roman, but I do have it missing from other fonts | 16:56.15 |
kens | Yes, there's no gbreve in the CHarstrings dictionary | 16:56.16 |
| THe font is NimbusRomNo9L-Regu | 16:56.35 |
| THere's a /breve, bit no /gbreve | 16:57.10 |
chrisl | Yeh, the fact I don't have it listed as missing means it isn't in the Monotype version, either | 16:57.40 |
kens | Well I'm just going to close the bug to be honest, he's using a non-embedded font wioth a customer encoding, thre's absolutely no guarantee it will work. | 16:58.11 |
chrisl | kens: I had it listed as missing from Nimbus_Roman_No9_Regular but not from Nimbus_Roman_Regular - I can't remember what the difference was....... | 17:05.31 |
kens | Well, I guess you have it as missing from the font I'musing. I'm not inclined to go raking through the document and font to see what other glyphs are missing though. IMO this is a consequence of not embedding the font you are using, and he gets what he deserves. | 17:06.44 |
| Ity looks like he used MS Word to create the document and something bizarre (the PDF Producer is a string of Chinese) to modify it and add the annotation. I ocudl probably solve it for him by telling him to use a Windows TTF font for Times-Roman, but I honestly can't be arsed | 17:08.22 |
chrisl | Speaking of that kind of crap, it looks like I'm going to have to rewrite the glyph name aliasing code for TT fonts loaded from disc..... again :-( | 17:09.23 |
kens | For Helmut ? :-( | 17:09.33 |
chrisl | Yeh, but not just him. It looks like the newer Windows fonts defeat the heuristic I wrote before, so others will stumble across it :-( | 17:10.45 |
kens | RIght, I guessed it was a font difference. | 17:11.01 |
chrisl | The problem is, I don't think there's any way to handle this without looping at least once through the entire AGL | 17:12.10 |
kens | Oh, and presumably that needs to be done each time the font is loaded, which is once per page its used for PDF | 17:12.39 |
chrisl | Yes, but only for non-embedded fonts...... | 17:13.16 |
kens | Which seems to be Helmut's main input, from what he's sent so far | 17:13.40 |
| Isn't this likely to affect Len as well ? | 17:14.25 |
chrisl | No, definitely not | 17:14.57 |
kens | Oh well that's something at least | 17:15.06 |
chrisl | And loading an entire TTF from disc will *always* be significantly slower than loading a subset from an already "open" PDF, so...... | 17:15.24 |
kens | Sure, its just that form what we've seen it will impact a bunch of people, oh well, more impetus for Mooscript | 17:15.55 |
chrisl | I'm not sure that mooscript actually solves this problem..... | 17:16.57 |
kens | At least we could do it in C instead of PostScript | 17:17.15 |
| Or is it already in C ? | 17:17.30 |
kens | has kind of lost track of that stuff | 17:17.41 |
chrisl | No, this is in PS. | 17:17.45 |
kens | Well, C would be quicker I guess | 17:17.55 |
chrisl | Possibly, but not massively..... | 17:18.25 |
kens | OK I'm going to give up in confusion over my current problem. Goodnight all | 17:18.28 |
chrisl | Goodnight kens | 17:18.40 |
rayjj | hmm... I don't see the info about which hotel we are going to. I just scanned through all of the emails from Miles and I see "hilton" mentioned, but not where. Does anyone know when the info was sent ? | 17:43.23 |
| I see all the info about cars, but presumably I need to tell the driver where to go :-) | 17:43.45 |
| oh, nm. Finally found it in the last staff meeting notes from Miles (Bayswater Hilton near Hyde park) (in case anyone else is lame enough to not already know ;-) ) | 17:48.13 |
chrisl | It's this one, isn't it: http://www3.hilton.com/en/hotels/united-kingdom/hilton-london-hyde-park-LONHPHN/index.html | 17:48.57 |
Robin_Watts | rayjj: The important thing to know is the postcode, I guess. | 17:51.07 |
| Cos the quote from the taxis are postcode based. | 17:51.24 |
| It's cheaper being in W2 than it is being further down the road, even though further down the road is closer to heathrow. | 17:51.50 |
| Don't ask me to explain that :) | 17:51.54 |
| Failing that, it's next to queensway tube station :) | 17:53.09 |
rayjj | Robin_Watts: so W2 4RJ is the postal code ? | 17:56.43 |
Robin_Watts | Yes. | 17:56.49 |
| W2 identifies the area of london (West 2), and 4RJ gets it to within less than "a block". | 17:57.14 |
mvrhel_laptop | Robin_Watts: ping | 20:05.52 |
| or paulgardiner | 20:15.36 |
wordToDaBird | Robin_Watts http://bugs.ghostscript.com/show_bug.cgi?id=695132#c1 | 22:55.16 |
| is that the preferred fix? | 22:55.16 |
| Forward 1 day (to 2014/12/03)>>> | |