| <<<Back 1 day (to 2012/10/16) | 2012/10/17 |
mvrhel_laptop | henrys: is the cups device not included by default in linux? | 02:30.03 |
| I did the fix for the issue that Robin_Watts and I were talking about this morning and the cluster push has a bunch of missing files with the cups device at 300 dpi | 02:30.38 |
| this is turning into a headache. I was hoping to get this stuff done for customer 330 and seem to be heading down on a detour now | 02:31.16 |
| thought I would check out the cups issue on linux since I thought or sure it was included by default | 02:31.53 |
| but alas no | 02:31.59 |
| hmm I thought it used to be included in the windows build | 02:35.51 |
| aha. a debug-cups version in the visual studio solution | 03:07.45 |
| what was the motivation for doing that | 03:08.03 |
henrys | mvrhel_laptop:I thought we ported cups to windows so you could work on color issues. | 04:04.57 |
mvrhel_laptop | yes. but it appears that changes were made with respect to cups being built with windows | 04:56.17 |
| I see all those configurations now in visual studio. for example a cups-debug configuration | 04:56.50 |
| and when I build that and the specify -sDEVICE=cups it does not seem to work | 04:58.03 |
| henrys: so I am not sure what is going on | 04:59.04 |
| well there is this one from chrisl | 04:59.53 |
| http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=b3ffd243ae375288c7c08607056dd08dae51018f | 04:59.55 |
| so it looks like that will do it | 05:00.14 |
| and it looks like Robin_Watts added something here related to the "Debug-cups configuration" | 05:01.37 |
| http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=7012db5af074788bf7bf14ef3718fe72eae3e701 | 05:01.39 |
| which I would have assumed would build with the cups device | 05:01.50 |
| It looks like that should work, but it does not seem to work for me | 05:03.18 |
| perhaps chrisl and Robin_Watts will read the logs and check that using the debug-cups configuration in the visual studio solution actually includes the cups device | 05:04.03 |
| I am done for the night | 05:04.16 |
kens | Robin_Watts : tor8, 'metta' who shoed up late last nioght in irc with a MuPDF buil;d problem raised it on Stack Overflow here : | 07:00.30 |
| http://stackoverflow.com/questions/12925208/stdio-h-file-not-found-mupdf | 07:00.30 |
| If you know what's going on could you answer it there too please ? | 07:00.45 |
| NB since its tagged with iphone I assume he's building on iOS so this may be one for tor... | 07:01.25 |
| chrisl did you see that mail to gs-bugs ? | 07:56.10 |
| about the inkcov device | 07:56.18 |
chrisl | kens: replying to it now | 07:56.22 |
kens | aha, thanks | 07:56.52 |
chrisl | Hmm, well, the Debug-cups configuration in the VC project works as expected for me - no idea what mvrhel's malfunction was..... | 08:11.09 |
kens | Er,, never tried it... | 08:13.38 |
chrisl | All it does is add the WITH_CUPS=1 option to the nmake command line, which tells the build to include the cups device | 08:14.48 |
kens | That's what I would expect, certainly. | 08:16.49 |
chrisl | <sigh> reboot required...... | 08:30.27 |
tor8 | paulgardiner: Robin_Watts: fitz/doc_document.c:194:9: warning: implicit declaration of function 'pdf_js_supported' | 09:05.34 |
Robin_Watts | tor8: Yes, there is a commit on robin/master for that. | 09:06.22 |
tor8 | Robin_Watts: okay, just ran across it building on my mac | 09:06.51 |
| trying to figure out that thing kens mentioned about Xcode not finding stdio.h | 09:07.08 |
| kens: I have no idea what problem 'metta' has run into. but it's not related to the thirdparty stuff (since that'd show up much later as the usual 'make' errors you get with missing files). I'd chalk it down to user error, not knowing how to use Xcode | 09:13.10 |
| oh for f.... now the "About This Mac" menu option takes over a minute to pop up the dialog. this mountain lion stuff -- it's not very good. | 09:16.27 |
| did I mention how iTunes starts playing silence randomly? | 09:16.41 |
paulgardiner | Robin_Watts, tor8: Ping. Question about how transparency impinges on separate pdf_run_page calls for main page and annotations. | 12:51.14 |
Robin_Watts | paulgardiner: PDF uses painters algorithm. | 12:51.41 |
| Page stream, then annotations. | 12:51.47 |
paulgardiner | But pdf_run_page_with_usage begins with a call to fz_begin_group conditional on page->transparency | 12:52.37 |
Robin_Watts | Right. | 12:52.49 |
paulgardiner | Do I leave that out when just painting annotations? | 12:53.03 |
| Will it matter if that is called twice when painting both layers? | 12:53.29 |
Robin_Watts | pixmaps in mupdf are n+1 planes deep, so rgba or grey+alpha etc. | 12:53.36 |
| I suspect that call is to do with the possibility of rendering pdf onto a pixmap that's not cleared. | 12:54.06 |
tor8 | I'd skip it for drawing annotations; it might cause blending to go a bit wrong on the annotations if they rely on the background being set etc | 12:54.11 |
paulgardiner | Ok great. I'll give it a go | 12:55.01 |
tor8 | Robin_Watts: it's the page level transparency group, we had some bug. IIRC it had something to do with the background being white in some blend modes, but the group should've been rendered as if it was transparent | 12:55.15 |
| paulgardiner: but in general, I think we'll be safe enough to draw annotations after the fz_end_group in pdf_run_page | 12:56.04 |
| I'm assuming you want to split the rendering up into two passes/functions, one for the page one for annotations | 12:56.25 |
Robin_Watts | I would have thought that annotations SHOULDN'T be in the page transparency group. | 12:56.36 |
| i.e. we should begin, render page, end, render annotations. | 12:56.57 |
tor8 | Robin_Watts: that's a reasonable guess, but I frankly don't know | 12:56.59 |
| it certainly makes most sense :) | 12:57.07 |
| but that's rarely what the pdf spec does... | 12:57.12 |
paulgardiner | :-) | 12:57.21 |
tor8 | Robin_Watts: I got the gtk viewer to build on windows. really annoying how many -I paths you have to bash in there... | 12:57.58 |
Robin_Watts | In the absence of evidence to the contrary, if it makes paulgardiner's life simpler (and I suspect it will) he should feel free to make it work in the "begin, render page, end, render annotations" way. | 12:58.08 |
tor8 | then the whole thing depends on 25 (count 'em ... 25!!!) dlls | 12:58.12 |
Robin_Watts | tor8: Ouch! | 12:58.20 |
tor8 | so I think we should do a native win32 viewer as well, and just use gtk for linux | 12:58.46 |
Robin_Watts | tor8: Have fun with that. | 12:58.59 |
chrisl | tor8: I did warn you that gtk+ on Windows was a PITA - why do you think I gave up on it?!?! | 12:59.03 |
tor8 | Robin_Watts: yeah... well, an Edit control should give me most of what I want. and I'm *not* stooping so low as to start using a C++ crossplatform toolkit like wxWhateveritisthesedays | 13:00.01 |
| chrisl: building is fine, but distributing that? no way I'm stuffing 25 dll:s in there... | 13:00.32 |
paulgardiner | Robin_Watts: yes, I was just going to say my change will be simpler if I take the annotations out of the group in all cases. | 13:00.57 |
tor8 | chrisl: at least on osx you can hide them all away in a .app bundle | 13:01.06 |
Robin_Watts | C:\Program Files (x86)\Artifex\MuPDF\DLLs\... :) | 13:01.55 |
chrisl | tor8: the reason I was looking at it was for a GUI for Jaws, so I was going to have to document *all* that so a customer could build it themselves (Jaws not being GPL) - complete dead end! | 13:02.14 |
tor8 | Robin_Watts: I like the one exe to rule them all approach. | 13:02.15 |
Robin_Watts | tor8: I agree. | 13:02.28 |
| Or at least 'everything bundled portably under 1 thing' approach (like .app) that Apple stole from RISC OS :) | 13:03.09 |
tor8 | chrisl: oh boy. not unless you send them a tarball of all the gtk+ binary development libraries as well... | 13:03.21 |
| building gtk+ statically seems like a no go these days as well. | 13:04.04 |
| so, scratch gtk for windows :( | 13:04.13 |
chrisl | tor8: even that was too close to legal problems - what a horror :-( | 13:04.52 |
Robin_Watts | paulgardiner: Looking at your commit now. | 13:07.30 |
| I dislike magic numbers; rather than passing 1 and 2 to pdfloadpage can we pass PDFAPP_LOADPAGE, PDFAPP_REFRESHPAGE ? | 13:08.26 |
| and is it feasible to have pdfapp_loadpage call pdfapp_recreate_displaylist ? | 13:10.54 |
| to avoid duplication of code? | 13:11.03 |
| (can't easily see that from the diff, it may not be) | 13:11.15 |
paulgardiner | Robin_Watts: getting rid of the magic numbers might be nice, but I was trying to avoid changing the large number of calls to pdfapp_showpage. | 13:13.35 |
Robin_Watts | I add a new param to pdfapp_showpage for the presentation stuff, so I've just him them all :) | 13:14.12 |
paulgardiner | shim? | 13:14.37 |
Robin_Watts | s/him/hit/ sorry. | 13:15.28 |
paulgardiner | Oh I see. You already have. | 13:16.05 |
Robin_Watts | The fact my commit is going to change them doesn't really have much bearing I guess. | 13:16.41 |
paulgardiner | Might be best to leave it how it is for now, because my next commit will change it again, when I make it updating only one of two display lists | 13:16.51 |
Robin_Watts | Personally, my dislike for magic numbers outweighs such things, but... | 13:17.11 |
| Ah, you're going to change it again... | 13:17.18 |
| Will it become a bitfield, perchance? | 13:17.35 |
paulgardiner | I don't know. Good time to discuss it. | 13:17.53 |
Robin_Watts | 1 = load, 2 = refresh main page list, 4 = refresh annot list ? | 13:18.01 |
paulgardiner | The current version was because I wasn't sure what was best, so just looked for the smallest change. | 13:18.16 |
Robin_Watts | Then PDFAPP_LOAD_PAGE = 7, etc... | 13:18.32 |
paulgardiner | I don't believe we ever need to refresh the main page list other than when we are changing page | 13:19.02 |
Robin_Watts | Possibly we could move the whole long list of pdfapp bool args into bits of a flagword. | 13:19.12 |
paulgardiner | ... I guess we might if ever we supported editing | 13:20.19 |
Robin_Watts | As I say, my transitions commit will add a new bool. | 13:20.45 |
| And it would all be much nicer if it was a flags word, IMHO. tor8, thoughts? | 13:21.05 |
| Otherwise the commit looks good. | 13:21.52 |
paulgardiner | Another possibility is to split into multiple calls, then each place will explicitly list the necessary actions. | 13:22.08 |
| ... but then I guess a flags word does that too in a way | 13:22.22 |
| Another style is to explictly destroy the cached stuff that is made dirty, but to make the recreation of it automatic. | 13:23.00 |
Robin_Watts | I find pdf_loadpage(app, 1,0,1,1,1); unclear. | 13:23.37 |
| pdf_loadpage(app, PDFAPP_LOADPAGE | PDFAPP_TRANSITION | PDFAPP_LOADANNOTATIONS); etc much clearer. | 13:24.09 |
paulgardiner | Or pdf_loadpage();pdf_transitions();pdf_loadannotations(); | 13:24.56 |
Robin_Watts | yes, but sometimes stuff doesn't break down into that linear order. | 13:25.18 |
| transitions, for example, need to do stuff before loading the page (to avoid the destruction of the old page bitmap), and afterwards. | 13:25.55 |
paulgardiner | Or pdfapp_destroypage();pdfapp_destroyTransitions();pdf_showpage() (where showpage recreates all that has been destroyed) | 13:26.17 |
Robin_Watts | That pulls order and logic out from under the interface. | 13:26.42 |
| I'd rather state the list of things I want done, and have the mechanics of how it does it hidden from the caller. | 13:27.03 |
paulgardiner | I was thinking the destroy;destroy;destroy;paint being less showing the mechanics because you destroy each thing you know to have become dirty and then trust paint to create what it needs | 13:28.28 |
Robin_Watts | right ,except that doesn't really work for transitions. | 13:28.59 |
paulgardiner | Yeah, I need to understand how transitions work. | 13:29.20 |
| Is this like annimation on things on the page? | 13:29.38 |
| Or animations between pages? | 13:29.56 |
Robin_Watts | You'd end up doing: destroypage, prepareTransition, showpage, continuepreparetransition or something. | 13:30.14 |
| animations between pages. | 13:30.18 |
| commit pushed. | 13:31.53 |
paulgardiner | Hmmm. Thinking of it from the destroy cache perspective. It would make sense for prepareTransition to do a call to destroyPage | 13:32.10 |
Robin_Watts | But the fact we need to make 2 transition calls, one before and one afterwards that's nasty. | 13:32.42 |
| I'd far rather just give the list of tasks and have it handled for me. | 13:32.57 |
paulgardiner | Yeah true. What do they do? | 13:32.58 |
Robin_Watts | See the commit on robin/master :) | 13:33.12 |
paulgardiner | :-) | 13:33.18 |
Robin_Watts | There are 4 commits on robin/master. Ignore the top 2. the 2 after that are ready to go, I think. | 13:33.33 |
| Gotta go make lunch. bbs. | 13:33.48 |
tor8 | Robin_Watts: just refactor the whole mess instead... | 14:00.17 |
paulgardiner | I was wondering about introducing, in addition to the pdfapp object, a pdfpage object that would encapsulate the bitmap, display-list, pdf_page etc, along with the number of the page the parts are derived from. Then for transitions, we can have two such objects, and the pdfapp's image can be the transition-combined images from the pdfpage objects. | 14:10.56 |
| At least that way it's clear which display-list pdf_page and image go together, and it should be clear what to throw away and regenerate when something changes. | 14:13.01 |
| We may need to add partial upate to this too, which could be a bit scary. | 14:13.47 |
mvrhel_laptop | chrisl: when I do the Debug-cups configuration build in the visual studio solution and try to use -sDEVICE=cups I get unknown device error | 14:18.55 |
| is that what you did? | 14:19.03 |
chrisl | mvrhel_laptop: I just did gswin32c --help and looked at the list of devices - "cups" was listed | 14:19.51 |
kens | is building it now | 14:20.06 |
mvrhel_laptop | not listed for me | 14:21.26 |
| I have to help get the kids breakfast and out the door. back in about 1 hour | 14:22.27 |
kens | When I try it I get an uknown device: cups error | 14:22.56 |
| cups is not listed in the available devices | 14:23.25 |
mvrhel_laptop | ok same as me | 14:23.35 |
kens | THis is a 64-bit build, I'll try 32-bit | 14:23.36 |
mvrhel_laptop | I am doing 32 bit | 14:23.51 |
chrisl | I just did 32 bit | 14:25.18 |
| kens, mvrhel_laptop: did you both do "clean" first? | 14:27.04 |
kens | I did, but I'm doing it again for 32-bit, might take a little while | 14:28.32 |
| finished, just a moment | 14:28.44 |
| 32-bit verison is OK | 14:29.23 |
chrisl | Works fine for me | 14:29.26 |
kens | let me trbuild 6t4-bit again | 14:29.30 |
| Huh, 64-bit is OK too, must have left something lying around before | 14:33.08 |
| ANyway, seems OK to me. | 14:33.37 |
Robin_Watts | I made changes to this the other day. | 14:33.38 |
| previously debug-cups wasn't available as a solution property, only a project property. | 14:33.52 |
kens | My source is up to date from that change | 14:33.54 |
| Can't speak for Michael | 14:34.00 |
Robin_Watts | But both debug and debug-cups (and debug-contrib) all build into the same debugbin/debugobj directories, so a clean is required. | 14:34.25 |
kens | Indeed, and doing so works for me | 14:35.21 |
chrisl | Yep, 64 bit works for me, too | 14:36.14 |
| So, I can close Visual Studio now? It offends me just knowing it's open.... ;-) | 14:37.02 |
Robin_Watts | paulgardiner: For transitions, I don't need a displaylist or a page object. I just need the bitmaps. | 14:38.46 |
| -s. | 14:38.54 |
paulgardiner | Robin_Watts: yes, but I thought it would make the logic more explicit and simpler. | 14:39.23 |
Robin_Watts | It would also mean we needlessly kept a (potentially large) display list around for longer than we needed. | 14:39.46 |
paulgardiner | At the moment the bitmap and the display list and the pdf_page are all for the same page number (thought the page number isn't there too) | 14:39.58 |
| You don't have to keep them. There could be a pdfpage_clean method | 14:40.24 |
Robin_Watts | And that stays the case under my patch too; 'old_bitmap' holds the previous bitmap for the duration of the transition. | 14:40.37 |
| new_bitmap holds the current one, and bitmap holds the combined one. At the end of the transition, new_bitmap goes into bitmap and old and new are freed. | 14:41.13 |
| but anyway, transitions are a lower priority than your stuff. | 14:41.40 |
| If there is a natural way for your stuff to work, then transitions can work around it. | 14:42.01 |
paulgardiner | Good idea to get something that fits with transitions though | 14:42.05 |
Robin_Watts | Anyway, there are those 2 reviews on robin/master. I'm going to think about progressive rendering now for a bit, | 14:42.36 |
henrys | looks my hpgl/2 hell is going to continue | 14:43.00 |
| looks like | 14:43.08 |
paulgardiner | So should we go with my current commit, given that I have to revisit that file in my next one? | 14:43.57 |
Robin_Watts | paulgardiner: I've already pushed it. | 14:44.18 |
paulgardiner | Brill. Thanks | 14:44.26 |
Robin_Watts | np. | 14:44.32 |
| henrys: Yeah, I read the mail this morning and my heart ached for you :( | 14:44.47 |
| "All those things you said couldn't be fixed? They are high priority. k, thx, bai" | 14:45.22 |
kens | THe perils of dealing with Japanese customers.... | 14:45.32 |
paulgardiner | Robin_Watts: did you want those first two commits reviewed? Obviously the first one is fine. | 14:50.36 |
Robin_Watts | Yes please. | 14:50.53 |
| (of the 4, the trivial one, plus the transitions one. Not the functions or the pdfwrite) | 14:51.12 |
| tor8 mostly looked over the transitions one the other day, and i've addressed the stuff we discussed. | 14:51.43 |
henrys | tor8:meant to ask you about the viewer yesterday. How's that exercise? | 14:59.55 |
tor8 | henrys: read my outburst in the irc logs first... :/ | 15:00.17 |
henrys | what about qt for windows? | 15:05.12 |
| maybe it is just as bad. | 15:05.30 |
tor8 | henrys: yeah, and I expect anything c++ on windows will be plagued by the "find the right c++ runtime to install" | 15:07.28 |
Robin_Watts | Plus the fact that they'd have to find a development team prepared to maintain a viewer written in C++ :) | 15:08.10 |
henrys | tor8:do you have the latest release of virtual box? looks like someone is going to do native windows programming ;-) | 15:13.20 |
Robin_Watts | thought that tor8 was actually a secret windows user. | 15:14.00 |
henrys | a possibly wild idea is to buy summatra make it portable and call that gsview. | 15:14.23 |
Robin_Watts | henrys: I suspect it's an impossible idea. | 15:14.41 |
tor8 | you'd have to hire zeniko to maintain it though. | 15:14.50 |
| and making it portable, hahaha! | 15:15.01 |
| well, we *could* wrap it in wine. | 15:15.25 |
Robin_Watts | tor8: The new linux viewer you've got... how much of that is UI, and how much is reusable code? | 15:16.11 |
| The existing linux/windows apps have a pdfapp layer that's portable, and then specific sections on top (win_main, x11_main). | 15:16.44 |
| Does the new viewer have something similar ? | 15:17.01 |
henrys | tor8:but the original summatra guy is doing other stuff and doesn't seem interested anymore, doesn't he hold the license? | 15:17.21 |
Robin_Watts | All the contributors for sumatra have contributed patches under the understanding that they become GNU GPL. | 15:17.58 |
| They haven't (AIUI) copyright assigned them to the original author. | 15:18.22 |
| So the original author has the right to release under GNU GPL, but he can't relicense it. | 15:18.54 |
henrys | well anyway the woods are full of GUI apps that run on Mac, Windows and Linux, so we don't have an unsolvable problem, it is just an unpleasant problem. It is imaginable we could hire paul to do the windows gui - linux and mac handled by the gtk. On the mac you can put the gtk shared objects in the App directory so stuff doesn't break. | 15:24.06 |
mvrhel_laptop | ok. chrisl, kens, and Robin_Watts sorry to waste your time on that. cleaning fixed the issue | 15:24.34 |
Robin_Watts | mvrhel_laptop: np. | 15:24.50 |
chrisl | NP, glad it's working! | 15:24.54 |
mvrhel_laptop | Robin_Watts: can I drag you in on this issue though | 15:25.11 |
Robin_Watts | mvrhel_laptop: ok... | 15:25.23 |
mvrhel_laptop | I have a patch that fixes the issue with the structure | 15:25.33 |
| and I did a cluster push | 15:25.45 |
tor8 | Robin_Watts: the new gtk+ viewer is mostly UI code, but there are two bits I want to make common code -- the search bits and command keybinding handling | 15:26.01 |
mvrhel_laptop | and if you check you will see that several cups files came back as error reading input file | 15:26.27 |
| and of course in windoze just now the file was created just fine | 15:27.18 |
henrys | is it me? or is the sumatra pdf page ugly? | 15:27.32 |
mvrhel_laptop | Robin_Watts: if I send you the patch can you see if you get any issues with the list of files in my cluster push result? | 15:27.59 |
Robin_Watts | sure. | 15:28.06 |
chrisl | henrys: I think it is intentionally "minimal" | 15:29.15 |
mvrhel_laptop | Robin_Watts: just sent it to you | 15:29.17 |
| I was testing tests_private/pdf/PDF_1.7_FTS/fts_01_0107.pdf.cups.300.1 gs i7 henrysx6 Error_reading_input_file | 15:30.10 |
| chrisl: what magic do I need to do to have cups appear as a device when doing a debug build on linux? | 15:30.45 |
chrisl | mvrhel_laptop: on linux it should just work, assuming the cups libraries are available on the system | 15:31.20 |
mvrhel_laptop | ok. I guess perhaps I dont have this on mine | 15:31.48 |
| no cups device available after building | 15:32.04 |
chrisl | OKay, then try giving configure the option "--with-local-cups" | 15:32.35 |
| Disclaimer - that is *not* well tested | 15:32.54 |
mvrhel_laptop | chrisl: where do I give this option? | 15:33.28 |
henrys | chrisl: last I fooled with this you had to install cups stuff with apt. | 15:33.29 |
| ? | 15:33.32 |
chrisl | mvrhel_laptop: to the configure script | 15:33.57 |
| henrys: that option has been there since we pulled in the cups code | 15:34.26 |
mvrhel_laptop | oh it did something with cups during configure | 15:34.47 |
| let me build and see | 15:34.51 |
chrisl | mvrhel_laptop: you can look in the Makefile to see if the cups device is listed - quicker than building | 15:35.43 |
mvrhel_laptop | and there is now a cups device | 15:35.46 |
| the build was pretty quick | 15:35.59 |
chrisl | Okay great - now <ahem> check it works! | 15:36.11 |
mvrhel_laptop | hehe | 15:36.16 |
| ok | 15:36.18 |
| hold on | 15:36.19 |
| it ran | 15:36.57 |
| and that is with my patch | 15:37.04 |
| so now I am really wondering what the issue is with the cluster push that I did | 15:37.18 |
chrisl | Is your linux 64 bit? | 15:37.41 |
mvrhel_laptop | no | 15:37.58 |
| well hold on | 15:38.05 |
chrisl | "uname -a" will tell you | 15:38.19 |
mvrhel_laptop | it is 64 | 15:38.32 |
chrisl | Okay, and did you fully replicate the cluster's massive command line? | 15:38.53 |
mvrhel_laptop | no | 15:39.00 |
Robin_Watts | mvrhel_laptop: Sorry, was on phone. | 15:39.03 |
mvrhel_laptop | ok. where do I find the full command line again | 15:39.20 |
Robin_Watts | Give me just a mo, then you'll have my full attention and I'll try and reproduce the problem. | 15:39.21 |
| mvrhel_laptop: Look in the logs. | 15:39.28 |
mvrhel_laptop | ok | 15:39.33 |
| good point | 15:39.40 |
Robin_Watts | on the dashboard, click + next to "mvrhel bmpcmp +" | 15:39.53 |
mvrhel_laptop | ha | 15:41.15 |
| here is the error | 15:41.21 |
| Copyright (C) 2012 Artifex Software, Inc. All rights reserved.This software comes with NO WARRANTY: see the file PUBLIC for details.Unknown device: cupsUnrecoverable error: undefined in .uninstallpagedeviceOperand stack: defaultdeviceCommand exited with non-zero status 10.00 0.00 0:00.02 16% | 15:41.26 |
| from the cluster push | 15:41.42 |
| not my issue | 15:41.46 |
| but why would that have occurred only to me | 15:42.26 |
| no one else has seen this? | 15:42.41 |
Robin_Watts | How strange. | 15:43.02 |
mvrhel_laptop | yes... | 15:43.09 |
Robin_Watts | Are all your failures on one particular machine? | 15:43.09 |
mvrhel_laptop | no | 15:43.14 |
| click on my regression results | 15:43.32 |
| oh wait | 15:44.00 |
| yes | 15:44.02 |
| henrysx6 | 15:44.05 |
Robin_Watts | Then I blame henrys :) | 15:44.13 |
mvrhel_laptop | well | 15:44.17 |
| I guess I am confused | 15:44.23 |
| there are 2 machines listed | 15:44.28 |
Robin_Watts | They are "new machine" "old machine", I think. | 15:44.43 |
mvrhel_laptop | oh | 15:44.52 |
Robin_Watts | new machine being "the one I just ran the test on" | 15:44.55 |
mvrhel_laptop | new machine is different | 15:45.06 |
| old machine is the same | 15:45.12 |
Robin_Watts | old machine being "the one this file was tested on before and that I am comparing with". | 15:45.13 |
| OK. so it's "old machine" "new machine" then :) | 15:45.26 |
| I bet henrysx6 needs cups installed. | 15:45.38 |
mvrhel_laptop | in any event, the common theme is henrysx6 | 15:45.43 |
| ok. whose task is that? henrys? | 15:46.35 |
Robin_Watts | I don't have a log in there. | 15:46.56 |
henrys | sorry I was away | 15:47.00 |
Robin_Watts | I can remove henrysx6 from the cluster so you can run your test. | 15:47.08 |
henrys | what did I do now? | 15:47.08 |
mvrhel_laptop | Robin_Watts: no I am just going to push | 15:47.21 |
| there were no other issue | 15:47.26 |
| s | 15:47.27 |
Robin_Watts | mvrhel_laptop: Right, but then the cluster test of your push will show failures. | 15:47.49 |
| I've removed henrysx6 from the cluster for now. | 15:48.13 |
mvrhel_laptop | ok. well I will wait until someone fixes the machine | 15:48.14 |
Robin_Watts | Push away. | 15:48.16 |
henrys | I thought chrisl just said it wasn't necessary to install cups - are we not using that option with the regression tests? | 15:48.19 |
mvrhel_laptop | ok | 15:48.20 |
Robin_Watts | henrys: It *is* necessary to install it on regression machines. | 15:48.38 |
| the configure script copes with it not being present though. | 15:48.55 |
mvrhel_laptop | ok. now back to customer 330 | 15:50.07 |
| ok. bbiab | 15:53.15 |
henrys | I am happy to install cups on henrysx6 I'd rather add the option to configure to do everything locally, it is just another dependency that will get broken every upgrade. Is there some reason that is not possible chrisl? | 15:54.53 |
chrisl | henrys: no reason that I know of - except we'll get a *hell* of a lot more compiler warnings noise. | 15:55.44 |
Robin_Watts | Wouldn't that mean us having to ship cups with gs ? | 15:57.18 |
chrisl | We do ship cups with gs | 15:57.28 |
Robin_Watts | oh. so why does henrys need to install it as well ? | 15:57.46 |
henrys | who can make regression script changes in marcos' absence? | 15:58.05 |
Robin_Watts | henrys: theoretically, me. | 15:58.13 |
chrisl | Well, what we have is a hacked up build of cups, that we know (hope!) works *just enough* for Windows users to debug it. We really don't want nor expect anyone to use it "for real" | 15:59.24 |
Robin_Watts | chrisl: Right, so we ship enough of cups for gdevcups to compile. | 16:00.02 |
| Why do we need a local cups installation too? | 16:00.12 |
chrisl | What's the point of regression testing a build no-one should ever be using? | 16:00.37 |
Robin_Watts | Lots of silly colorspaces in cups. | 16:00.55 |
chrisl | Which is why we test cups, built as it is when it gets used for real, and not some hacked up thing we slung together | 16:01.37 |
henrys | chrisl:I thought mvrhel_laptop made substantial changes to the cups code, i.e. we want our cups to be released. What's wrong with making the crap portable? | 16:01.40 |
chrisl | We've made no changes to the cups code, to my knowledge | 16:02.07 |
henrys | the crap == cups | 16:02.09 |
Robin_Watts | chrisl: If we ship the dogfood, we should eat the dogfood. | 16:02.39 |
chrisl | but no one will *ever* use it like that, so it's a crap regression test | 16:03.18 |
Robin_Watts | Ideally I guess the configure script should allow us to disable cups, enable local cups, or enable shared cups. | 16:03.25 |
chrisl | It does | 16:03.37 |
henrys | oh okay, it sounds like I should just install cups, which of 100 or so packages that contain cups in the name or description do I need? | 16:03.44 |
| that is not an exaggeration by the way... apt-cache search cups | wc -l | 16:04.35 |
| yield 92 | 16:04.40 |
| yields | 16:04.51 |
| absurd. | 16:05.14 |
chrisl | libcups2-dev libcupsimage2-dev | 16:05.16 |
henrys | done | 16:06.26 |
Robin_Watts | Hmm. Supposedly "The following 132 regression file(s) have been removed". Anyone know about that? | 16:07.18 |
kens | wasn't me, not guilty | 16:07.35 |
chrisl | I don't think I ever found the logic which caused that | 16:12.25 |
Robin_Watts | http://htwins.net/scale2/ | 16:54.16 |
mvrhel_laptop | Robin_Watts: yes. this is a very cool web site | 16:57.56 |
| kids had quite a bit of fun with it | 16:58.09 |
| and so did I | 16:58.12 |
henrys | I knew we were made of foam | 16:58.24 |
mvrhel_laptop | I like how he has the size of a Minecraft world thrown in there | 17:00.00 |
Robin_Watts | henrys: It's a relief to know that. | 17:00.19 |
| I was convinced that if we got small enough it'd be perl scripts running on DOS 3. | 17:00.37 |
henrys | then we couldn't have free will but with foam the possibilities are endless | 17:01.29 |
| it is a very good website you should post it on facebook. | 17:02.18 |
Robin_Watts | I got it from facebook :) | 17:02.59 |
| And I think the scope for indetermisms is high enough in perl to explain the illusion of free will. | 17:03.57 |
henrys | there is that yes. | 17:04.45 |
Robin_Watts | tor8: ping | 20:25.37 |
| tor8, paulgardiner, sebras: For the logs then... I'd like to update our exception handling code slightly. | 20:26.41 |
| Currently we can throw strings. I'd like to make it so that we throw strings and an error code. | 20:27.09 |
| fz_throw(ctx, "message", ...); will remain unchanged | 20:27.21 |
| but we'll also have a fz_throw_code(ctx, code, "message", ...); | 20:27.37 |
| and that means in fz_catch() we can do: if (fz_caught_code(ctx) == SOME_SPECIFIC_VALUE) then we can handle errors of a particular type. | 20:28.24 |
| (fz_throw() will give it's errors a code of 0). | 20:28.50 |
| This means that for progressive reading, I can throw an error for "Not Enough Data Yet". | 20:29.37 |
| and people can catch that specific error and deal with it differently. | 20:29.58 |
sebras | Robin_Watts: where would do you envision this to be used? | 20:44.49 |
| sounds like it would be image decoding or such..? | 20:45.02 |
Robin_Watts | sebras: I'm thinking of it being used with progressive decode. | 23:21.17 |
| If you try to read a page that the data isn't here for yet, it throws 'NOTYET'. | 23:21.42 |
| (i.e. for reading linearized PDFs as they download) | 23:22.00 |
sebras | Robin_Watts: I'm worried about conflating unrecoverable errors with something which is normal operation. and this ought to be normal operation as I understand it..? | 23:44.00 |
| but I can see that we wouldn't want to treat linearized files differently from the others. hm.. | 23:44.40 |
| btw, gs, made it to ioccc: http://www.ioccc.org/2012/hamano/hint.html | 23:44.48 |
| oh, and mupdf can't decode those files either. the repair logic gets confused by the assumed hex-string after #include and fails soon after this. | 23:45.28 |
Robin_Watts | sebras: Essentially, we want people to be able to say "get page 10" and be informed somehow that page 10 isn't ready yet. | 23:46.46 |
| So I feel it's a reasonable use of the error system to say "failed to render, but it's worth trying again in a bit" | 23:47.40 |
sebras | like EAGAIn. | 23:49.54 |
Robin_Watts | yeah. | 23:51.23 |
| There may be another way of doing it that relies on smartness in the stream reader. I need to think about it some more. | 23:52.22 |
| but bedtime... | 23:53.03 |
sebras | agreed. | 23:56.40 |
| tor8: btw, you asked me to test chinese input on the gtk+ app. works well and searching too. | 23:58.39 |
| Forward 1 day (to 2012/10/18)>>> | |