| <<<Back 1 day (to 2013/05/13) | 2013/05/14 |
mvrhel_laptop | tkamppeter: is there place to upload my slides or do I send them to you? | 00:25.56 |
| good night | 06:00.49 |
kens | ping chrisl | 09:43.31 |
chrisl | kens: pong | 09:43.42 |
kens | got a minute to check me on something ? | 09:43.54 |
chrisl | Sure | 09:44.00 |
kens | Can you look at gxccache.c, gx_lookup_cached_char | 09:44.16 |
| around line 154 | 09:44.27 |
chrisl | Got it | 09:44.52 |
kens | I think there is a potential endless loop here | 09:45.05 |
| We have a large flat tbale with up to 0x3fff entries | 09:45.17 |
| We start by (in essence) hashing the glyph and fm_pair | 09:45.33 |
| and we look at that point in the table. | 09:45.41 |
| If its the cached character we want, we return | 09:45.53 |
| If its not, we add 1 to the index, and go round again | 09:46.04 |
| Note that the index is AND'ed with the max table size, so when we reach teh end of the table we go back to 0 | 09:46.30 |
chrisl | I *think* the final entry in the table is always NULL | 09:46.48 |
kens | THe only way to break out of this, is if we find the cached char, or an empty slot | 09:46.51 |
| I managed to create a situation where the table is full | 09:47.03 |
| Just by calling gx_add_cached_char | 09:47.15 |
| Admittedly my code was incorrect, but all it did was add the same characters to the cache all over again | 09:47.47 |
| SO the cache filled up | 09:47.56 |
chrisl | Yes, but isn't the caller of gx_add_cached_char() supposed to check whether the cache is "full" before calling it? | 09:48.01 |
kens | Is it ? Itsa not obvious | 09:48.14 |
| THe code in pdfwrite certainly never does | 09:48.53 |
chrisl | Well, actually, the decision is taken during the setcachedevice (and co) call | 09:49.01 |
kens | OK, and what if the glyph never calls setcachedevice ? | 09:49.19 |
chrisl | Then the glyph is wrong, surely? | 09:49.35 |
kens | PCL bitmap fonts can have a space glyph whcih doesn't call setcachedevice | 09:49.39 |
| It happens that's teh glyphs I was adding | 09:49.50 |
chrisl | Then I would suggest that the PCL interpreter is wrong, and should be fixed. Withouth the call the setcachedevice, the cache device should never be created, so there should be no glyph to add to the cache | 09:51.02 |
kens | Hmm, maybe I'll open a bug for Henry. It would certainly fix my pdfwrite code if the setcachedevice were called. I'll check it further, maybe it just evades my pdfwrite override somehow | 09:51.48 |
| (and I have a fix for my own problem already in testing ;-) | 09:52.08 |
| Yep. In pl_bitmap_build_char, if the 'cdata' (presumably the bitmap) is NULL,it simply returns 0 without calling setcachedevice | 09:54.45 |
chrisl | Well, my reading of gx_add_cached_char() is that it assumes available space in the cache has already been checked, so it *requires* the setcachedevice to check all the cache limits. | 09:55.46 |
kens | I'll raise a bug for henry, just going to poke it a little and see if I can get some more information | 09:56.26 |
| I guess that the PCL interpreter doesn't add it to the cache, but pdfwrite *needs* to, in order to detect redefined glyphs in a font. | 09:57.00 |
chrisl | kens: it's possible that the intention in PCL is to "render" the empty bitmap uncached so as not to clutter up the cache with a non-marking glyph - but that seems a pointless non-optimisation | 09:57.26 |
kens | chrisl, if there is no bitmap data, then the glyph has no width it seems, there is actually nothign to tell the interpreter what to do, which is why Henry returns I think. | 09:58.05 |
| Setting a '0' cache device entry would maek more sense I think. | 09:58.18 |
chrisl | Yes, sorry, I assumed it was a space glyph or something. Simply not calling setcachedevice is *not* sufficient in this case | 09:59.11 |
kens | Its debatable as to what it is exactly. The PXL file uses a space character, but we don't seem to have a space in the font definition. I think this works out because the text is drawn with a XSpacing, so each glyph is spaced individually. | 10:00.17 |
| Definitely one to chuck to Henry. | 10:00.38 |
| I'd guess that if the text wasn't drawn with individual spacing, the 'space'; would simply vanish | 10:01.04 |
chrisl | Could this be one of the uses of the "galley character"? (whose purpose has always escaped me) | 10:01.20 |
kens | Umm, dunno.... | 10:01.29 |
| I don't think so though. | 10:01.35 |
chrisl | This sounds like another case of PCL making invalid assumptions about the behaviour of the graphics library | 10:03.24 |
kens | Actually it loks vaguely like Henry (or someone) added the NULL check to get around a seg fualt, or a valgrind warning or somethign similar | 10:03.58 |
| Well if the PXL interpreter emits a setcachedevice, then my code works a lot better, without the hackery I've just added. | 10:06.35 |
chrisl | Either that or it needs to set the flag in the enumerator to say "do not cache", I'd think | 10:07.40 |
kens | I'd rather it called setcachedevice | 10:08.02 |
chrisl | That would be my recommendation, yes. | 10:08.15 |
kens | If it calls it with all 0 then everything works fine for me | 10:08.30 |
henrys | oh my favorite subject caching bitmap characters | 13:04.29 |
kens | :-) | 13:04.36 |
henrys | kens, chrisl:I should probably review that - it seems that the problem should be caught earlier on and not even get to build_char but if you are in a hurry please go ahead and use your fix as a workaround. Can you assign the but to me? | 13:19.42 |
| s/but/bug | 13:20.04 |
chrisl | henrys: isn't it preferable to have it cached, so we don't have to mess around with the font on every invocation of the glyph? | 13:20.44 |
kens | henrys, I'm not in a hurry | 13:20.47 |
henrys | chris;have what cached? it's a broken font with no data. | 13:21.39 |
chrisl | henrys: ah okay, I'd been assuming it was a valid font with an empty glyph | 13:22.16 |
kens | I was unsure if this was valid or not | 13:22.39 |
| I can commit my fix for the %d in putput file, and leave this open, it causes glyphs to go missing | 13:23.59 |
| but only with this broken font | 13:24.51 |
henrys | chrisl:to get there we would have had to create a character slot for the glyph and then not filled that slot with data (all bitmaps are downloaded) which is is invalid. | 13:24.53 |
chrisl | henrys: fair enough - if it's invalid then I guess we're not so worried about it. | 13:26.18 |
henrys | kens:I'd rather see marcosw tools be more robust about errors but if he has a good reason this won't work, okay. We should expect some jobs will have many errors in the testing. | 13:28.04 |
kens | henrys, its worse than that :-) | 13:28.37 |
| If you ignore the errors, then all but one of the pages has no text | 13:28.47 |
| I have a fix for that, but that exposed this problem | 13:29.02 |
| I have a fix ford that too, but while working on it I noticed this cache quirk | 13:31.32 |
| I'll commit what I have that works, and open a new enhancement for you henry ? | 13:32.24 |
henrys | kens:okay not thrilled with the hack - another eyebrow raising snippet in the code to baffle the reader but I guess that's just the way it is. | 13:34.05 |
kens | henrys, I have code which works without changes to the PCL interpreter | 13:34.41 |
| Its all surprising code in pdfwrite | 13:34.55 |
| which is more 'expected' | 13:35.08 |
henrys | kens:oh so you don't need the plchar.c change? | 13:35.15 |
| wow adobe has really pissed everyone off with subscription creative cloud⦠petitions. | 13:36.51 |
chrisl | I can't say that I'm surprised: their stuff is stupid expensive, and past experience says this will just make it more stupid expensive | 13:38.40 |
kens | henrys I don't *need* the changes in plchar.c, but they make some code I added redundant, the thing is I spotted the potential cache problem while wortking on it | 13:41.49 |
| I'll commit what I have | 13:42.42 |
| which should resolve Marcos's bug | 13:42.48 |
chrisl | I need to head out for a while - I should be back (barring traffic problems) in plenty of time for the GS meeting. | 14:06.26 |
Robin_Watts_ | tor8: ping | 14:48.52 |
tor8 | Robin_Watts: pong | 14:49.29 |
Robin_Watts | I just updated the svgwrite to be " like this" rather than "like that " | 14:49.46 |
| s/svgwrite/svgwrite commit/ | 14:50.01 |
| Do you want to look it over again, or are you happy with it as is? | 14:50.18 |
tor8 | was it helpful? | 14:50.21 |
Robin_Watts | didn't save me anything, but you're right it's better in the long run. | 14:50.38 |
tor8 | I'm happy to let the svgwrite device in | 14:51.40 |
| an item for the mupdf meeting would be a decomposition device or layer | 14:52.14 |
Robin_Watts | tor8: yeah. | 14:52.26 |
| Let's see if we can remember that for 9 minutes :) | 14:52.39 |
tor8 | either by fallback hooks in dev_null or by explicit transformation devices | 14:52.46 |
henrys | 8 | 14:52.49 |
Robin_Watts | 9 here. Time zones, eh? :) | 14:53.10 |
tor8 | text/font -> paths (or images) and shadings -> images are high priority there | 14:53.41 |
| and maybe tile flattening | 14:53.47 |
Robin_Watts | tor8: yeah, the text/font -> paths/images is hopefully just a rejigging of the existing code. | 14:54.42 |
| I guess shadings/images are too. | 14:54.51 |
tor8 | getting texts to paths should be trivial | 14:55.02 |
| to images would need to pick a resolution and ask for a pixmap and wrap that in an image | 14:55.16 |
| I don't think we need to call the draw device for any of these | 14:55.45 |
Robin_Watts | tor8: I think we'd assume that the matrix in use was to device pixels. | 14:55.57 |
tor8 | for the mesh we can probably just set up a pixmap and call the paint functions | 14:56.11 |
Robin_Watts | The draw device already has code that goes from text -> paths and text -> images. | 14:56.16 |
| It's just that when we have the path/image we plot it by a direct call, rather than calling back through the device interface. | 14:56.39 |
tor8 | we should probably hoist that code out of the draw device as a utility function then, if you want to reuse it | 14:57.08 |
Robin_Watts | that was my thought. | 14:57.19 |
tor8 | using the ctm for device pixels would be bad for svg output though | 14:57.55 |
Robin_Watts | the only worry with that, is that it might be slower to actually plot bitmaps calling back into the device rather than just doing it directly, as we have to check for the need to scale etc. | 14:57.57 |
tor8 | I'd pass a callback function to the utility function | 14:58.20 |
| and make the callback plot directly for the draw device | 14:58.29 |
Robin_Watts | tor8: it would. hence for svg we'd want to go to paths all the time or something. | 14:58.36 |
tor8 | and wrap it up in an image/path drawing call for the decomposition device | 14:58.41 |
Robin_Watts | tor8: yeah. that'd work. | 14:58.48 |
| for svg we'd want to go to an svgfont (like a type3 font), probably. | 14:59.17 |
tor8 | Robin_Watts: for svg output, can you make a "library" of resources? | 14:59.20 |
Robin_Watts | That might need a bit more thought. | 14:59.23 |
tor8 | font decomposition should ideally use some form of resource caching for the glyphs | 14:59.46 |
Robin_Watts | tor8: you can give objects id's, yes. | 14:59.47 |
| so yes, we could do each glyph as a pattern. | 15:00.05 |
| but svgfont would be neater. | 15:00.13 |
tor8 | we could drop all glyphs as images or paths into an svg "defs" section | 15:00.44 |
Robin_Watts | tor8: We don't need to put stuff in a defs section. | 15:01.00 |
tor8 | but that's going to mean a two-pass approach | 15:01.00 |
Robin_Watts | defs is optional. | 15:01.06 |
henrys | does anyone know what Raed expects? | 15:01.09 |
Robin_Watts | which is good, cos I'm opting to ignore it. | 15:01.16 |
| henrys: for svgwrite? No. | 15:01.23 |
tor8 | Robin_Watts: oh, so you can have "draw-this-and-save-as-defs-for-later-use" in svg? | 15:01.35 |
| it's been too long since I read the svg spec | 15:01.44 |
Robin_Watts | tor8: Yes, AIUI. | 15:02.05 |
henrys | I wonder if it would be useful if marcosw asked him to describe the product to assist us in steering the development. | 15:02.15 |
tor8 | anyway, I'd start by going all paths, no caching or reuse, for text decomposition. | 15:02.20 |
Robin_Watts | henrys: I was about to suggest that I mail him saying "This is what we have for now, feel free to try it out. Can you tell us more about what you want?" | 15:02.52 |
tor8 | but if that's true (define at first use by setting id#) we can save the id and get some reuse there, but the decomposition device would need to be smarter there... | 15:02.59 |
| ...so then we could just as well be creating svgfonts | 15:03.18 |
Robin_Watts | tor8: but the smartness can be in the callback handler, possibly. | 15:03.21 |
henrys | Robin_Watts:probably best for it to go through marcosw but either way. | 15:04.01 |
tor8 | Robin_Watts: we have resources in the store for images, that should be used for svgout to reuse images | 15:04.03 |
| but for paths we have no such mechanism | 15:04.08 |
Robin_Watts | tor8: We can't use the store to do reuse; in low memory conditions we'd end up evicting all the time, and nothing would be reused. | 15:04.54 |
| We need to track reuse in the device, I think. It's what I did in pdfwrite (I think). | 15:05.12 |
| for now, we don't reuse images in svgwrite. | 15:05.36 |
henrys | paulgardiner read the status - looks good, I wonder if signatures could be something for michael's talk. | 15:05.46 |
Robin_Watts | henrys: signatures were mentioned in the "coming soon" section of the stuff I sent to michael. | 15:06.26 |
henrys | Robin_Watts:ah must have missed it ⦠great. | 15:06.41 |
mvrhel_laptop | Yes. I have that listed. I don't know a lot of details about it though | 15:06.43 |
Robin_Watts | (I think I copied it to tech) | 15:06.44 |
paulgardiner | henrys: how long until the talk? | 15:07.17 |
mvrhel_laptop | the talk is tomorrow. I am heading off tonight | 15:07.40 |
henrys | mupdf'er please read about the open printing summit business here: http://www.linuxfoundation.org/collaborate/workgroups/openprinting/openprinting-summitpwg-meeting-cupertino-2013 | 15:07.55 |
mvrhel_laptop | Robin_Watts: I did have one question for you | 15:07.56 |
Robin_Watts | mvrhel_laptop: You can sign a PDF with a public/private key to say things like "This is an unchanged version of the original document" and "This is signed with all the keys of the people that have approved it" etc. | 15:07.57 |
mvrhel_laptop | On one of your headings you had the title "Complete" PDF support. I was wondering about the quotes.... | 15:08.33 |
Robin_Watts | so in a company you could write something, and sign it, and send it to your colleague who could see that it really was from you, and he could sign it and send it on to someone else, who can see that it's really from both of you and ... | 15:09.00 |
mvrhel_laptop | Robin_Watts: ok thanks for the note above on signatures | 15:09.01 |
paulgardiner | henrys: we are a long way off being able to sign documents. There's a lot of work in supporting PDF's incremental update. | 15:09.04 |
henrys | marcosw is already at the meeting | 15:09.33 |
| or summit | 15:09.41 |
mvrhel_laptop | paulgardiner: ok. so we are not able to do what Robin_Watts just described above but that is where we are heading | 15:09.41 |
paulgardiner | mvrhel_laptop: yes exactly | 15:09.57 |
Robin_Watts | mvrhel_laptop: We don't support: 3d models, movies, pdf portfolios... | 15:10.07 |
tor8 | Robin_Watts: how do you deal with font reuse in pdfwrite? | 15:10.13 |
mvrhel_laptop | Robin_Watts: ok. fair enough | 15:10.27 |
Robin_Watts | tor8: /* FIXME: Get someone who understands fonts to fill in this bit. */ | 15:10.38 |
mvrhel_laptop | :) | 15:10.46 |
henrys | Robin_Watts: I like it. | 15:11.08 |
Robin_Watts | tor8: I suspect I currently take a reference to each font I see and keep it in a list. | 15:11.09 |
tor8 | Robin_Watts: which means they have to survive the store | 15:11.34 |
Robin_Watts | by us having a reference, they *will* survive the store. | 15:11.48 |
| The store can only ditch things where it's the only holder of a reference. | 15:12.42 |
henrys | of 693503 - this is a P1 with last comment in january? | 15:14.27 |
paulgardiner | Eek! git commit -a on casper has thrown me into nano. Was expecting vi. | 15:14.38 |
Robin_Watts | Ctrl-X | 15:14.49 |
| Ctrl-X then Ctrl-C, probably | 15:15.05 |
tor8 | Robin_Watts: right. so we can do the same for images. | 15:15.23 |
henrys | reboot casper ;-) | 15:15.24 |
paulgardiner | Something like that worked | 15:15.28 |
tor8 | to reuse them in pdfwrite and svgwrite | 15:15.30 |
Robin_Watts | tor8: yeah. | 15:15.48 |
| I just haven't coded that. | 15:15.55 |
tor8 | and add a list like that to the text decomposition | 15:15.56 |
Robin_Watts | henrys: Yeah, possibly that should be closed now. I'll recheck. | 15:16.08 |
tor8 | if it wants to reuse glyph images | 15:16.15 |
Robin_Watts | henrys: oh, wait, that's the bug for the fuzzying of mupdf. | 15:17.44 |
| It's all the openjpeg stuff still left. | 15:17.50 |
| I forget... did we have Shelly looking into using OpenJpeg 2 or something ? | 15:18.29 |
henrys | Robin_Watts:what about using luratech for commercial releases in mupdf? | 15:18.49 |
Robin_Watts | henrys: I have no experience of luratech, but I guess it's possible. | 15:19.16 |
henrys | Robin_Watts:either him or zeniko - it ought to be bountiable if you want somebody to look at it. | 15:19.21 |
Robin_Watts | henrys: Step 1) Get OpenJPEG 2 working as well as our patched OpenJPEG 1.5 currently does. | 15:19.57 |
| Step 2) fix any of the fuzzing bugs that are still outstanding. | 15:20.16 |
| Step 2 is bug 693503. Step 1 we'd need to create a new bug and make it bountiable - if there isn't one already. | 15:20.47 |
| So Michael takes the stage at 9pm tomorrow our time. | 15:22.24 |
mvrhel_laptop | Yes. So, I believe the winrt viewer is working pretty well now. The only things I need to do are add a progress bar during text search, store the state in case of suspension, password entry, clean up tab spacing issues. At that point perhaps we can see about having it merged into golden and maybe submitted as an app in windows store? Or do we want to wait for that? | 15:22.37 |
henrys | mvrhel_laptop: it's all ASAP I think, why sit on it. | 15:23.10 |
Robin_Watts | mvrhel_laptop: Have you pushed it to your repo? | 15:23.50 |
mvrhel_laptop | ok. sounds good to me. | 15:23.57 |
| Robin_Watts: yes. it is all up there | 15:24.15 |
Robin_Watts | cool. I'll try and look at that soon. | 15:24.28 |
mvrhel_laptop | Robin_Watts: thanks | 15:24.34 |
Robin_Watts | We had a volunteer to do the armeabi stuff with V8. | 15:24.47 |
paulgardiner_ | I just fell off IRC, so may have missed the last few messages | 15:25.05 |
Robin_Watts | he said he'd look this coming weekend, so maybe we'll have some news for next weeks meeting. | 15:25.12 |
henrys | mvrhel_laptop: are there build instructions in case we get volunteers who aren't as comfortable as the devels? | 15:25.31 |
Robin_Watts | Load VS2012 Project. Hit Make. ? | 15:25.50 |
mvrhel_laptop | that is pretty much it | 15:25.56 |
| I will add a few notes as to what is the default project, what is the winrt library | 15:26.30 |
| and I can add some notes about the different classes in the project | 15:26.41 |
henrys | mvrhel_laptop, Robin_Watts:I thought I'd have to install xaml libraries or some set of windows acronyms to get things going. | 15:27.07 |
Robin_Watts | Oh, a random thought I had... it would be nice to get an 'overview' document of the classes in the android viewer, and how they fit together. | 15:27.18 |
mvrhel_laptop | right now the viewer is in c++, but it might be nice to add a simple viewer (no frills just pages shown) in a c# and javascript or visual basic project | 15:27.23 |
henrys | Robin_Watts:I'd really like that - paulgardiner ? | 15:27.53 |
mvrhel_laptop | henrys: no if you are in windows8 and have visual studio 2012 it should just work | 15:28.06 |
Robin_Watts | To do that we'd need paulgardiner to take some time off digital signatures. I guess I should make an enhancement bug and assign it to him, so he can do it when he needs a break? | 15:28.08 |
paulgardiner | Sure. No problem | 15:28.10 |
henrys | Robin_Watts:sounds good. | 15:28.28 |
| for all the openjpeg stuff can we just make a bunch of bountiable stuff for zeniko and shelly? | 15:29.44 |
| parcel it up into bounty size pieces? | 15:30.02 |
Robin_Watts | henrys: yeah, the problem is in knowing enough to know how to bundle it :( | 15:30.35 |
| I have no idea if it's a 5 minute job, or a 5 week job. | 15:30.54 |
| It's possible I did at one point, but it's all fallen out of my head. | 15:31.11 |
henrys | Robin_Watts:I was just thinking that. We could make on big bug and just tell them we'll assess upon code receipt and they'll just have to trust us? | 15:31.38 |
Robin_Watts | henrys: That would seem reasonable to me. | 15:31.56 |
| If they identify large separable areas we can break out bugs along the way. | 15:32.11 |
henrys | once we see the work it is fairly straightforward to asses the difficulty and the proper reward .. | 15:32.23 |
Robin_Watts | seems fair to me. I'll open a bug and send an email out. | 15:32.47 |
henrys | great | 15:32.55 |
Robin_Watts | I've been chasing a problem with pattern handling in mupdf all week. | 15:33.24 |
henrys | I was going to go over the milestones in the schedule miles gave raph but we're already over let's do it another time. | 15:33.28 |
Robin_Watts | I've got it down to hopefully the last few diffs now. | 15:33.40 |
| The test file that kens supplied kills gs. | 15:34.04 |
paulgardiner | I need some help deciding how to include openssl in mupdf, but that can be handled outside the meeting | 15:34.34 |
henrys | oh there is one final mupdf issue we need a third party list like ghostscript has maybe tor8 can do this or you might be able to convince chrisl who did ghostscript's. | 15:34.38 |
tor8 | third party list? | 15:34.51 |
henrys | tor8:see gs/doc/thirdpaty.htm | 15:35.42 |
Robin_Watts | henrys: With mupdf, isn't that "everything that's in the thirdparty directory" | 15:36.09 |
henrys | or thirdparty.htm would be a better thing to look at. | 15:36.10 |
Robin_Watts | ? | 15:36.11 |
mvrhel_laptop | brb | 15:36.33 |
Robin_Watts | henrys: editing the gs one into a mupdf one would be simple enough I think. | 15:37.16 |
| I'll do it if no one else wants to. | 15:37.25 |
henrys | okay, you may want to ask chrisl how he keeps up with versions. | 15:38.21 |
| and new libraries | 15:38.35 |
| seems like chrisl should own all of it. one person is more likely to keep it up to date but I guess it isn't that important. | 15:39.25 |
Robin_Watts | Can we claim that the git submodule definition file is our thirdparty.htm? :) | 15:39.59 |
paulgardiner | Robin_Watts: the same thought struck me | 15:40.22 |
Robin_Watts | paulgardiner: bug 694073 is all yours. | 15:40.41 |
paulgardiner | Your so kind! | 15:41.00 |
| You're even | 15:41.05 |
henrys | Robin_Watts: uh no. | 15:41.15 |
| ah I notice hin-tak will be at the summit | 15:42.49 |
| I wish they had a webcast or something. | 15:43.35 |
ray_laptop | I thought they had previously | 15:44.03 |
Robin_Watts | They are at Apple. I can't believe that Apple aren't set up for that. | 15:44.21 |
chrisl | Apple are probably worried there'd be a glimpse of something super-secret | 15:45.35 |
Robin_Watts | Aha, a chrisl. I'm going to make tea before the next meeting, but see above for discussion of the thirdparty.htm thing. Do you want to do it? If not, do you have a cunning system for catching version changes etc? | 15:46.37 |
henrys | they do have call in numbers. | 15:46.41 |
chrisl | Robin_Watts: no magic, I'm afraid, they don't change that often. I'm not fussy about doing it or not, it's not a big job, on the whole | 15:47.54 |
henrys | I see here you can joint the meetings:http://www.pwg.org/chair/meeting-info/may_2013_cupertino.html | 15:48.58 |
marcosw_ | morning | 15:50.11 |
chrisl | marcosw_: morning - thought you weren't going to be around? | 15:50.44 |
ray_laptop | marcosw: I thought you weren't going to be here | 15:50.52 |
| now we can't say nasty things about you ;-) | 15:51.07 |
chrisl | It doesn't usually stop us ;-) | 15:51.26 |
marcosw_ | meeting hasn't started yet. | 15:51.26 |
| btw, hin-tak says hello. Â First time I've met him in person. | 15:51.38 |
henrys | marcosw_:yes say hi for us too, I just noticed him on the web page. | 15:52.07 |
| are you attending the pwg plenary marcosw_? | 15:52.26 |
chrisl | paulgardiner: I reckon I can "fix" the libtiff and libpng WinRT problems without actually touching their source - I'll have a proper play with that tomorrow | 15:54.52 |
marcosw_ | henrys: yeas, my meeting at uni this morning was cancelled, so I'll be here until lunch. | 15:55.04 |
paulgardiner | chrisl: excellent | 15:55.12 |
henrys | marcosw_:"plenary" sounds highfalutin | 15:55.29 |
Robin_Watts | henrys: Will weed help, do you think? | 15:56.00 |
mvrhel_laptop | only in Colorado or Washington | 15:57.14 |
henrys | they'd have to move it here where it is legal. | 15:57.16 |
ray_laptop | plenary is just a fancy way of saying 'bull session', right ? | 15:57.33 |
| hi, mvrhel_laptop | 15:58.12 |
| when is your talk ? | 15:58.20 |
Robin_Watts | http://www.linuxfoundation.org/collaborate/workgroups/openprinting/openprinting-summitpwg-meeting-cupertino-2013 | 15:58.48 |
henrys | ray_laptop:it is actually a good use of the word I looked it up: 2 (of a meeting) to be attended by all participants at a conference or assembly, who otherwise meet in smaller groups: a plenary session of the European Parliament. | 15:58.52 |
Robin_Watts | Tomorrow 1pm | 15:58.56 |
mvrhel_laptop | ray_laptop: tomorrow | 15:59.01 |
henrys | oh the ghostscript meeting. | 16:00.29 |
| mvrhel_laptop:694067 - can we tell the customer he shouldn't be doing that and make this a non customer issue. | 16:02.39 |
marcosw_ | there are only about 10 people here in person. Â If everyone calls in via webex we'll have them outnumbered. Â BTW, tkamppeter is not here in person, he's on the phone. | 16:02.44 |
mvrhel_laptop | henrys: ok I was wondering about that | 16:02.58 |
| seriously til is not there? | 16:03.24 |
henrys | marcosw_:can you see the people webex'd? | 16:03.46 |
| a room of talking heads? | 16:03.58 |
marcosw_ | henrys: no, it's voice only. | 16:04.04 |
ray_laptop | on the page Robin_Watts provided the link to, it says he'll call in | 16:04.08 |
marcosw_ | mvrhel_laptop: yup, he's on the phone. | 16:04.18 |
Robin_Watts | webexed: me, Michael Sweet (host), David L. Zwang, Nancy Chen | 16:04.44 |
henrys | so should we ditch the meeting and webex in? | 16:05.16 |
Robin_Watts | I can see a slide, but there is no audio or video (yet) | 16:05.20 |
marcosw_ | you may have to call in to get audio:Â +1-619-326-2730 | 16:06.51 |
| Access code: 491659# | 16:07.09 |
| ooops, | 16:07.11 |
kens | Don't fancy internationall call | 16:07.19 |
Robin_Watts | I might do that for mvrhels talk tomorrow. | 16:07.23 |
marcosw_ | til says he couldn't come because of a visa issue. | 16:07.30 |
Robin_Watts | kens: telediscount.co.uk <- 1p a minute. | 16:07.34 |
henrys | mvrhel_laptop:I just had something similar and I assigned it back to marcosw_ to contact the customer then he can give it back to you as a non customer bug ⦠keeps the ball rolling and stuff off engineer's plates. | 16:07.56 |
mvrhel_laptop | henrys: ok I will do that same thanks | 16:08.13 |
marcosw_ | oops, I gave out the wrong number, that was for the pre-meeting. | 16:08.36 |
henrys | the only other thing I had for the meeting is what to do about all these fuzzing bugs and should we fix the static analysis bugs first assuming they are easier and will cause some of the fuzzing bugs? | 16:10.16 |
| anybody else have stuff for the ghostscript meeting? | 16:10.34 |
kens | Not today | 16:10.51 |
Robin_Watts | kens: your patterns.pdf file kills gs. | 16:11.03 |
kens | You said | 16:11.09 |
| But tha'ts one for alex :-) | 16:11.16 |
| Its a barking mad test file to be honest | 16:11.29 |
alexcher__ | henrys: make the fuzzing bugs bountiable ? | 16:11.33 |
marcosw_ | henrys:the bad news is that the fuzzing bugs that I've entered are only a subset, the 64 bit build issues that cause segfaults. Â I have 64 bit valgrind issues that don't segfault and 32 bit issues as well. | 16:11.35 |
Robin_Watts | kens: right, but opening a bug and attaching the file is not an option, right ? | 16:11.53 |
alexcher__ | kens: please assing this bug to me. | 16:12.12 |
kens | I'd prefer not to add that file to our repository of tests, nor make it public | 16:12.15 |
| alexcher, there is as yet no bug | 16:12.26 |
Robin_Watts | kens: but as a private attachment on a bug is alright? | 16:12.33 |
henrys | so I guess we all skeptical that working on scan build problems will fix fuzzing problems? | 16:12.36 |
chrisl | kens: you can add it a private attachment, and note in the bug that it shouldn't be added to the test repo | 16:12.51 |
kens | Robin_Watts : as long as its clear that it gets deleted, and not added to the test suite | 16:12.53 |
Robin_Watts | kens: Once added to a bug it can't (easily) be deleted. | 16:13.13 |
kens | THen I'd rather send the file internally as required | 16:13.29 |
chrisl | henrys: I'm skeptical about that, none of the ones I've looked at would fall into that category | 16:13.36 |
kens | I can trust you all to delete the file, right ? :-) | 16:13.46 |
marcosw_ | Robin_Watts: yes it can, I've enabled attachment deletion in bugzilla (you might have to have admin priv). | 16:13.50 |
henrys | chrisl:wishful thinking on my part. | 16:13.58 |
mvrhel_laptop | oh. does anyone have anything interesting in the trunk (or for the next release) that they want mentioned at the meeting tomorrow? | 16:14.01 |
| kens: perhaps your work with pdfwrite color? | 16:14.13 |
| will that be in the august release? | 16:14.24 |
| I only heard back from ray_laptop in my email | 16:14.44 |
kens | mvrhel_laptop : I would like to think so, but I am not promising anything at the moment | 16:14.48 |
chrisl | henrys: it's possible there will be overlap, given the sheer number of both | 16:14.49 |
Robin_Watts | mvrhel_laptop: I looked back over the history, and all my commits have been windows specific or SEGVs fixes - so no new features for linux users. | 16:14.51 |
marcosw_ | mvrhel_laptop: good news for you, the August open printing meeting in August will be in Washington State. | 16:15.09 |
mvrhel_laptop | marcosw_: I was not aware there was an August meeting | 16:15.28 |
ray_laptop | marcosw: I didn't see the weekly run last Friday with NumRenderingThreads. Did it not run for some reason ? | 16:15.46 |
henrys | mvrhel_laptop:did you harvest everything out of the newsletter? | 16:15.56 |
mvrhel_laptop | henrys: old stuff I have | 16:16.08 |
| yes | 16:16.09 |
marcosw_ | mvrhel_laptop:sorry, it's the PWG meeting. | 16:16.16 |
mvrhel_laptop | Just new stuff that has not yet been released | 16:16.22 |
ray_laptop | the updates I sent to mvrhel_laptop were after 9.07 | 16:16.22 |
mvrhel_laptop | yes. I want after 9.07 | 16:16.30 |
henrys | mvrhel_laptop:the xps writer is coming along. | 16:16.40 |
mvrhel_laptop | I will mention that. and they will ask, what is xps? | 16:16.59 |
| :) | 16:17.05 |
alexcher | mvrhel_laptop: we have conducted a systematic search for SEGVs, Valgrind warnings, and fuzzing problems. | 16:17.14 |
mvrhel_laptop | alexcher, yes, I have that in the talk | 16:17.24 |
| thanks | 16:17.26 |
chrisl | We haven't fixed them, but we've looked for them.... ;-) | 16:17.46 |
henrys | we can juggle, do card tricks and handstands. | 16:17.52 |
mvrhel_laptop | yes. the usual circus acts | 16:18.05 |
ray_laptop | and chrisl plays the guitar | 16:18.10 |
chrisl | And I can sing Ave Maria whilst standing on one leg in a bucket of custard (for any fans of Spike Milligan out there!) | 16:19.19 |
marcosw_ | ray_laptop:the NumRenderingThreads test email should have gone out on Friday afternoon.  Let me check if I received it. | 16:19.34 |
mvrhel_laptop | marcosw_ : did you see my email to you about the windowsRT mudpf project yesterday? | 16:20.44 |
marcosw_ | ray_laptop: I received the email, the subject is: "weekly regression report - Multithreaded and backgorund rendering - 2013-05-10-16:45:30" | 16:20.54 |
| there were 2000+ bitmap differences, but I believe there were all caused by a different change. Â I haven't looked at the bitmaps yet, but will post bug(s) if background printing broke something. | 16:21.43 |
ray_laptop | marcosw: strange. It's not in my inbox or my trash (on T'bird). I'll check gmail directly | 16:22.05 |
mvrhel_laptop | ray_laptop: good grief the customer's definition of color is really strict | 16:22.28 |
ray_laptop | mvrhel_laptop: yes, probably overly so | 16:22.48 |
mvrhel_laptop | they don't want neutral pages printed | 16:23.00 |
| it probably costs them money | 16:23.08 |
marcosw_ | mvrhel_laptop:No, I didn't see the windowsRT mupdf email. Â Let me look now. | 16:23.13 |
mvrhel_laptop | i.e. they make more if it is printed in color | 16:23.19 |
marcosw_ | mvrhel_laptop:turns out I did see your email and replied. | 16:24.01 |
henrys | marcosw_:there must be a lot of duplication in the fuzzing - I would think that one say missing error return would result in many problems. Any ideas how we can recheck all the problems per fix? | 16:24.06 |
ray_laptop | mvrhel_laptop: you are probably right. | 16:24.17 |
| at least Phil was satisfied with the FitPage fix | 16:25.36 |
tkamppeter | marcosw_, only 10 persons at Apple? Where are all the others? Did they got denied entry to the US, too? | 16:26.27 |
mvrhel_laptop | ray_laptop: if they want to have R=G=B strictly then they need to set the DEV_NEUTRAL_16 = 0 and DEV_NEUTRAL_8 = 0 and AB_NEUTRAL_8 = 0 and AB_NEUTRAL_16 =0 in gscms.h | 16:26.59 |
| AB_NEUTRAL_16 = 0 | 16:27.21 |
| marcosw_: ok thanks | 16:27.48 |
| Robin_Watts: do you have the VS version for which the ghostscript solution is made? | 16:28.54 |
henrys | ray_laptop:I think you have an old bountiable bug about romfs compression that needs to be settled if you don't have time just accept it and let's pay him and we can review it later. | 16:30.02 |
mvrhel_laptop | Robin_Watts: if it is easy for you to do, can you add gsicc_monitorcm.c to base/color/icc . I could try to do if from this 2012 version but would likely mess something up | 16:30.21 |
henrys | kens:you had a bountiable one as well. | 16:30.48 |
Robin_Watts | mvrhel_laptop: I do, and I will. | 16:31.12 |
mvrhel_laptop | Robin_Watts: thanks | 16:31.23 |
chrisl | Ah, good, Robin_Watts got there - FWIW, I have the appropriate VS as well | 16:32.01 |
henrys | Robin_Watts:is the webex good? | 16:32.03 |
marcosw_ | tkamppeter: I don't know where everyone is but some people came in late. Â There are now 14 in the room now. | 16:34.02 |
Robin_Watts | henrys: it's pointless without audio, I reckon. | 16:34.26 |
marcosw_ | Robin_Watts you can ask tkamppeter what the international dial in number and access code is. | 16:35.06 |
kens | henrys which boutiable bug are you referring to ? | 16:35.43 |
| I accepted teh bicubic one | 16:36.03 |
henrys | kens:ah that was it. | 16:36.36 |
kens | OK Math said he knew what to do, so I left it with him | 16:36.49 |
ray_laptop | henrys: right, I remember that. It's to do the lzma compression. Sorry that I didn't finish the review, but as long as we are OK with taking the entire 7zip archive, I guess it's OK. It _would_ be nicer to just take the part that we need, however | 16:36.58 |
| but the romfs part looks OK. | 16:37.09 |
Robin_Watts | mvrhel_laptop: http://www.softpedia.com/progDownload/Microsoft-Visual-C-Toolkit-Download-11595.html | 16:37.14 |
ray_laptop | I'm fine with giving him the bounty. It's good work | 16:37.29 |
Robin_Watts | I think that's a download for VS2005 Express Edition, which should enable you to make changes like this in the future. | 16:37.31 |
mvrhel_laptop | Robin_Watts: ah ok. thanks | 16:37.42 |
henrys | ray_laptop:you can add "please collect the bounty" to the bug or I'll do it, either way is fine by me. | 16:38.14 |
tkamppeter | marcosw_, henrys, Robin_Watts: I have sent you the number now. | 16:38.24 |
alexcher | ray_laptop: I've checked PDF collection code. Currently, temporary files have to survive --restore-- there. What can be done about it? | 16:40.45 |
Robin_Watts | mvrhel_laptop: Done and pushed. | 16:41.03 |
mvrhel_laptop | Robin_Watts: thanks | 16:41.17 |
chrisl | alexcher: could the files be opened in global VM? | 16:41.19 |
Robin_Watts | np. | 16:41.21 |
ray_laptop | henrys: I updated the bug 690969 with the 'approval' | 16:41.37 |
alexcher | chrisl: probably, i'll try this. | 16:43.00 |
tkamppeter | marcosw_, there are 19 who announced themselves to come in person, but there are perhaps some not coming all days. | 16:43.17 |
ray_laptop | henrys: you've already communincated with him, right ? | 16:44.15 |
| henrys: I'm fine with you contacting him | 16:44.37 |
chrisl | ray_laptop: persumably we only need the content of the "C" directory in the SDK archive? | 16:45.12 |
ray_laptop | henrys: I need to run a paper over to the school. anything else for me? Anyone ? | 16:45.17 |
henrys | ray_laptop:he knows to go to miles then miles contact me and we verify the bookkeeping is right | 16:45.26 |
| yeah meeting over sorry | 16:45.32 |
ray_laptop | henrys: OK. Thanks. | 16:45.43 |
henrys | 10:30 it's over ! | 16:45.43 |
kens | Ah cool, I will be off then. GOodnight all | 16:45.49 |
henrys | 9:30 Pacific time. | 16:46.05 |
ray_laptop | chrisl: yeah, just taking the C tree is probably enough | 16:46.20 |
chrisl | ray_laptop: if you want to assign it to me as a "build" thingy | 16:46.51 |
ray_laptop | mvrhel_laptop: Thanks. I'll tell them that as well. And tell them the reason for the 'tolerance' | 16:49.51 |
| chrisl: OK. Thanks. | 16:50.10 |
| mvrhel_laptop: oops. old message. ignore it | 16:50.34 |
| for some reason, gmail classifies some of the regression emails as Spam. That's what happened to the Multithreaded one from 5/10 | 17:25.16 |
Robin_Watts | bah. Why do gutters only go wrong when it's raining? | 17:42.31 |
marcosw_ | Robin_Watts:http://www.ghostscript.com/coverage/mupdf/ | 18:46.27 |
Robin_Watts | Thanks. | 18:50.13 |
| So... none of the test files ever throws an error. | 18:50.24 |
| or uses the context. | 18:50.27 |
| I think gcov has some issues :( | 18:50.33 |
| http://www.ghostscript.com/coverage/mupdf/fitz/base_context.c.html | 18:50.53 |
| Are you planning to put gs/pcl etc in the same place? | 18:51.23 |
| I'll add links to the dashboard if I know where to point them. | 18:51.38 |
| tor8: ping | 18:59.32 |
| tor8: http://git.ghostscript.com/?p=user/robin/mupdf.git;a=commitdiff;h=1d83fde9932b2102e50a72d7fdf70201aedbc7d5 | 19:00.14 |
sebras_ | Robin_Watts: what files did you use for that coverage measurement? | 19:11.24 |
Robin_Watts | sebras: I didn't, marcosw did. | 19:16.37 |
| If you look in the files, some of them have lines in yellow. | 19:16.59 |
sebras | Robin_Watts: yes.. are those misses? | 19:17.11 |
Robin_Watts | mouse over those lines and it gives you a few files. click on the line and it gives you the whole list of files. | 19:17.20 |
| no, those are the hits. | 19:17.25 |
sebras | aha, so the private suite. | 19:18.18 |
| Robin_Watts: I get much better coverage at home, so something is definitely breaking gcov here... | 19:18.43 |
| Robin_Watts: what's the difference between yellow and black lines? | 19:20.40 |
| obviously the grey ones have not been hit. | 19:20.51 |
Robin_Watts | sebras: can you give me an example? | 19:21.03 |
sebras | http://www.ghostscript.com/coverage/mupdf/fitz/filt_predict.c.html | 19:21.15 |
Robin_Watts | I don't know. marcosw? | 19:22.12 |
sebras | Robin_Watts: even looking at the html-source doesn't reveal anything... | 19:25.03 |
marcosw_ | The yellow highlighted lines have a different set of tests covering them than the previously covered line. | 19:42.42 |
| sebras: I'm willing to believe that I have errors in the coverage analysis. Â An example of a test and line would be helpful. | 19:44.06 |
Robin_Watts | As an example, every single file must call fz_new_context | 19:44.53 |
marcosw_ | Robin_Watts: Yes, I'm preparing the ghostpcl coverage analysis now and it will go in the analogous location. | 19:45.25 |
Robin_Watts | marcosw: but according to http://www.ghostscript.com/coverage/mupdf/fitz/base_context.c.html nothing calls it. | 19:45.32 |
marcosw_ | Robin_Watts: thanks, I'll take a look. | 19:46.16 |
| Forward 1 day (to 2013/05/15)>>> | |