| | <<<Back 1 day (to 2013/04/24) | 2013/04/25 |
| ray_laptop | bmpcmp didn't help me on the cups differences. At least for the psdcmyk I am able to see what's going on -- the 'equivalent_cmyk_colors.color[sep_num].color_info_valid is false in the device I clone for the bg printing :-( | 04:18.00 |
| | I have to figure out how to copy that data across into the cloned bg_print device (safely) | 04:19.11 |
| | I don't know if there is a way to do this. This is a private element of some devices (such as psdcmyk). I probably have to do something similar to what we do for devn_params (which is the same kind of thing) | 04:32.10 |
| | i.e., add a "ret_equiv_colors" device proc that will give me a pointer then make an equivalent to devn_copy_params that does the copy_equiv_colors action. I want to consult with mvrhel to see if there is any shortcut, or if this will even work. | 04:34.51 |
| | It will probably work if the device uses the standard update_spot_equivalent_cmyk_colors with the standard equivalent_cmyk_color_params struct in the device that the device will return a pointer to from its ret_equiv_colors proc | 04:39.43 |
| | work on it (and consult with mvrhel) in the AM when I'm fresh and coffeed up | 04:40.58 |
| mvrhel_laptop | ok. so I am calling my mupdf with C# through my c++ winrt component | 05:20.46 |
| | I think I see what we will need to do to make this work with ghostscript | 05:21.09 |
| | we will need to create it in in visual studio though as a separate project outside the normal build process. So you will want to take and create a windows runtime c++ component and create an interface to the gs library. then you can create apps in c++ javascript c# etc that can call into gs | 05:24.08 |
| | henrys: no sure if you want paul to get bogged down with that or not. I think I could do it relatively quick (a day) | 05:25.02 |
| | s/no/not/ | 05:25.11 |
| | the problem is though that we can only do this in windows 8 and with visual studio 2012 | 05:25.44 |
| | so it would be a separate little solution project | 05:25.52 |
| | but that is fine | 05:26.10 |
| | at least it gives our customers an interface to gs for windows store apps | 05:27.07 |
| | in what ever language they want to use | 05:27.19 |
| natronite | good morning | 05:28.26 |
| | we use mupdf on android and sometimes we have the issues that it does not render the page correctly. What happens is that the text is in a light gray color instead of black and the pictures (which I assume usually have a transparent background) have a gray background | 05:29.41 |
| | does anyone know what could be the issue here? | 05:29.48 |
| mvrhel_laptop | uhoh I see ray_laptop had some issues that he wants to chat about in the morning. done for the night so I can get some sleep | 05:30.03 |
| | natronite: mupdf folks will be on in a few hours | 05:30.22 |
| natronite | no hurry :p | 05:35.15 |
| | .. | 06:47.07 |
| | we use mupdf on android and sometimes we have the issues that it does not render the page correctly. What happens is that the text is in a light gray color instead of black and the pictures (which I assume usually have a transparent background) have a gray background | 08:31.25 |
| | does anyone know what could be the issue here? | 08:31.31 |
| kens | natronite : when you say 'sometimes' do you mean the problem is inconsistent (sometimes it happen on a particular file, sometimes it doesn't) or that the problem only occurs with some files, but those files are always incorrect ? | 08:32.45 |
| natronite | i mean it does not happen every time | 08:33.11 |
| | if I render the same file again a little later it renders fine | 08:33.42 |
| kens | Well that sounds like some kind of display list problem, and I'll have to refer you to the real MuPDF developers. tor8 ping ? | 08:34.10 |
| natronite | what also happens on other files is that several characters are renderes as rectangles | 08:34.38 |
| kens | That sounds like the well known 'TrueType /.notdef' problem. I would imagine the fonts are not embedded in the document. | 08:35.12 |
| tor8 | natronite: I've never seen that before. could you open a bug on bugs.ghostscript.com and attach the file? | 08:35.14 |
| | fonts not rendering properly is almost always because the font isn't embedded and there isn't enough reliable information in the pdf file to create a proper substitute | 08:36.14 |
| natronite | so why does it render correctly when I render it other times? | 08:36.37 |
| kens | natronite : are we talking about the fonts or the gray ? | 08:36.55 |
| tor8 | natronite: that would probably be a bug. we are supposed to be deterministic :) | 08:37.01 |
| natronite | the gray and the fonts are two different things | 08:37.16 |
| | some files get gray text and images | 08:37.25 |
| kens | natronite : yes, which is why I was asking. | 08:37.28 |
| natronite | and other files have rectangles while the text color is fine | 08:37.38 |
| kens | I'm assuming the 'rectangles' is deterministic, that is it is always the same | 08:37.46 |
| natronite | both aren't. Sometimes the file with the rectangles renders fine | 08:38.35 |
| kens | OK so that is interesting indeed. | 08:38.45 |
| natronite | I'll make screenshots to add to the bug I'll report | 08:38.54 |
| kens | More important is the relevant file. THis sounds like it will be difficutl to pinpoint though, unless you can provide a means to reproduce the problem reliably | 08:39.27 |
| natronite | true. Let me explain what I do | 08:39.53 |
| kens | That sounds like a good idea, yes | 08:40.11 |
| tor8 | natronite: if you're on a linux machine, I would suggest running mupdf in valgrind and see if that spots anything | 08:40.12 |
| kens | tor8 he did say Android | 08:40.23 |
| natronite | Using minizip I copy the memory of the needed file into memory to create a stream which I can use to open a document | 08:40.26 |
| | because the files we display are within a minizip | 08:41.11 |
| | I mean within a zip file | 08:41.19 |
| | and it's faster to extract the needed files on the run than to extract the whole archive beforehand | 08:41.53 |
| tor8 | natronite: that's a supported workflow, so nothing strange there | 08:44.29 |
| natronite | sadly I'm on os x | 08:45.14 |
| | but I can extract the zip on the phone | 08:45.48 |
| | and open it with the mupdf app | 08:45.53 |
| | the unzipped version opens fine every time in mupdf | 09:03.06 |
| kens2 | chrisl ping | 11:42.22 |
| pgogna | Robin: I tried the idea you have given and found the bounding box at the position where i double tapped on the pdf | 12:36.01 |
| | But I doubt whether i am getting correct bounding box or not? | 12:37.02 |
| Robin_Watts | pgogna: hi. | 12:49.58 |
| | Well, if it was me, I'd zoom the page in until a piece of text just filled the screen widthwise, then I'd click it and look at what value I got back. | 12:50.46 |
| pgogna | There are two different steps. So lets goto step by step | 13:00.48 |
| | step 1. Please ensure if i am getting correct bounding box? | 13:01.26 |
| | Step 2. Calculate scaling factor to zoom to that bounding box | 13:02.18 |
| Robin_Watts | pgogna: Right, so step 1. | 13:02.33 |
| | zoom the screen using the ordinary zoom functions in mupdf so that a line of text just about fills the screen widthways. | 13:02.57 |
| | Then double click that text, and look at the bbox you get back. | 13:03.10 |
| pgogna | ok | 13:03.33 |
| | I was trying to print the text inside the bbox so that i will get the idea of what that bbox contains | 13:06.36 |
| | And Step 2 | 13:06.51 |
| Robin_Watts | that seems reasonable too. | 13:06.53 |
| | We'll talk about step 2 after you have step 1 working. | 13:07.16 |
| chrisl | kens2: (belated) pong | 13:31.51 |
| kens2 | chrisl no worries, the problem went away, I got a omilefail on feet, but a retry and it was OK | 13:32.13 |
| | compilefail* | 13:32.23 |
| chrisl | Hmm, I've been seeing a few spurious build failures - if it carries on, I'll maybe ask marcosw to see if we can gather data to work out why..... | 13:33.20 |
| kens2 | The error was a memory error, just aa mo | 13:33.36 |
| | It was building gsromfs1.c and said: | 13:34.26 |
| | cc1: out of memory allocating 67108864 bytes after a total of 18354176 bytes | 13:34.26 |
| | make: *** [obj/gsromfs1.o] Error 1 | 13:34.26 |
| Robin_Watts | it got 18meg, and balked at a further 67 ? | 13:35.09 |
| kens2 | I' also seeing "TERM environment variable not set." here and when doing bmpcmp and compiles | 13:35.18 |
| chrisl | Yeh, I don't think the TERM variable one is anything to worry about | 13:35.38 |
| kens2 | No, just mentioning it in passing, it seems not to cause problems | 13:35.52 |
| | If you want the full log for the fail I still haveit for now | 13:36.04 |
| chrisl | No, not just now. But if it comes up more times, let me know | 13:36.29 |
| kens2 | OK | 13:36.35 |
| mvrhel_laptop | good morning | 13:53.00 |
| kens2 | Robin_Watts : ping | 13:54.32 |
| mvrhel_laptop | paulgardiner: did you make any progress on the winrt stuff? I dont know if you saw my earlier comment, but I believe I know what needs to be done | 13:56.01 |
| | or at least on approach | 13:56.17 |
| | s/on/one/ | 13:56.21 |
| paulgardiner | No. Haven't seen your comment. I've been experimenting trying to link a simple win32 dll into a winrt dll. I've read on the net that that can be done, which may mean no special build needed for gs, just removal of calls to no-longer-existing APIS | 13:58.43 |
| | Don't seem to have received your comment. I thought I was on the lists now. | 13:59.20 |
| mvrhel_laptop | paulgardiner: oh it was from yesterday perhaps | 13:59.37 |
| kens2 | Possibly it was on irc ? | 13:59.42 |
| mvrhel_laptop | yes on irc | 13:59.50 |
| | paulgardiner: in the irc logs | 13:59.59 |
| kens2 | paulgardiner : you need to look at the irc logs | 14:00.00 |
| mvrhel_laptop | from today | 14:00.01 |
| paulgardiner | ah right | 14:00.08 |
| mvrhel_laptop | to me this approach I described sounds a little easier than hacking away at our build scripts | 14:01.31 |
| paulgardiner | Ok. That sounds good, but does he need to call directly to gs from C#. Is his winrt dll written in C# or something? (Excuse me if that makes no sense. I may be misunderstanding something completely). | 14:02.42 |
| mvrhel_laptop | paulgardiner: he could write an application in c#, c++, javascript etc and create the winrt object in his app and make calls into gs | 14:04.03 |
| paulgardiner | I thought he'd already split off a dll and wanted gs in a separate dll, which he called from the first | 14:04.58 |
| kens2 | I understadn his problem to be that he can't call the GS API (from a managed code application) because its has no winmd file. | 14:05.41 |
| | THe GS DLL that is. | 14:05.56 |
| mvrhel_laptop | ok. I guess I misunderstood what he wanted. if he is not in a windows store app the that is fine. if he is in a windows store app then I suspect we need to do things the way that I described | 14:06.45 |
| paulgardiner | mvrhel_laptop: I wouldn't say that. I suspect you were probably right. I'll just read the emails again a minute. | 14:07.29 |
| mvrhel_laptop | to be able to call from managed code into gs will require the creation of a windows runtime component | 14:07.47 |
| kens2 | I must admit its somewhat opaque to me. My understanding has been that he has written an application using managed code (c# for example) and wishes to call the GS DLL from that application, in the wya he is used to with a C app. | 14:07.49 |
| mvrhel_laptop | at least for a windows store app | 14:07.59 |
| pgogna | Am i doing correct to get the instance of current loaded page? | 14:08.00 |
| | this is what i am doing pc = &glo->pages[glo->current]; page = pc->page; | 14:08.11 |
| mvrhel_laptop | kens2: if he is working in a windows store app, he is not going to be able to do that without having gs wrapped up in some way as a windows runtime object | 14:08.51 |
| | have to help get kids out door. bbiab | 14:09.13 |
| kens2 | mvrhel_laptop : I was under the impression it was a (coding) language problem, but I am not at all certain | 14:09.27 |
| paulgardiner | Yes. I think he needs to create a windows runtime component, but it may be possible to have gs in a seperate win32 dll, to which the wr component makes calls | 14:09.57 |
| pgogna | can anybody help me | 14:10.01 |
| kens2 | But my feeling was that we would require to wrap the DLL in different code to get it to work though | 14:10.02 |
| paulgardiner | ... or at least there seems to be claims of that set up being possible on the net. | 14:10.22 |
| kens2 | Well, that would be nice if true, since we wouldn't then have to set up a new component | 14:10.45 |
| | just for 'windows store apps' | 14:10.53 |
| | But if we have to do so I guess we can, its not really any worse than the hoops required to set up a DLL. | 14:11.21 |
| paulgardiner | Some one posted that they'd tried using their 2008 built win32 dll directly from their wr component, and it didn't work, but then did when they rebuilt the win32 dll in 2012 | 14:12.15 |
| | The wr compoent linked with .lib file from the win32 dll build. | 14:12.46 |
| kens2 | Hmm. I guess that would meanus building with VS 2012, that won't beentirely popular no doubt | 14:12.59 |
| paulgardiner | All this subject to my not having completely misunderstood something. | 14:13.00 |
| mvrhel_laptop | right. it really would not be too hard to do. can make a new solution for vs2012 that uses the gs library | 14:13.04 |
| | we def. have to use vs 2012 | 14:13.09 |
| kens2 | I'm sure chrisl will be delighted | 14:13.41 |
| | as it happens I have a VS 2012 project for GS | 14:14.03 |
| | had to make one for that customer the other day with the thread local storage problem | 14:14.26 |
| mvrhel_laptop | well kens2 we dont need all of gs wrapped up in vs2012 | 14:14.27 |
| | we use the current build to create the library | 14:14.39 |
| | and have a very small vs 2012 solution that creates the windows runtime componenet | 14:14.57 |
| kens2 | Hmm, ok | 14:15.05 |
| mvrhel_laptop | and the gs library | 14:15.06 |
| | it is a separate thing | 14:15.15 |
| | excuse me | 14:15.24 |
| | it uses the gs library | 14:15.30 |
| kens2 | 'gs library' is a DLL or something else ? | 14:15.31 |
| mvrhel_laptop | well I would prefer a static lib | 14:15.45 |
| | I think that you may have to do that | 14:15.54 |
| kens2 | Right, so its pretty much what I expected | 14:16.08 |
| mvrhel_laptop | so we use the existing make stuff to make the static lib | 14:16.21 |
| | and have a VS2012 solution that make the windows runtime component | 14:16.44 |
| | and raed uses that to make his calls from c# etc | 14:16.54 |
| kens2 | Actually Raed could do this himself,it seems to me | 14:17.08 |
| mvrhel_laptop | ye | 14:17.12 |
| | s | 14:17.13 |
| | Like I said earlier it can probably be done in a day | 14:17.38 |
| kens2 | weshoudl probably do it and give him instructions, then discuss it at the staff meeting to see if we think we should support this ourselves | 14:17.47 |
| mvrhel_laptop | with a coffee break and lunch thrown in | 14:17.48 |
| pgogna | Robin: pc = &glo->pages[glo->current]; page = pc->page; Is this correct to get the current loaded page instance? | 14:17.49 |
| kens2 | mvrhel_laptop : yes its pretty much what I expected, just like a different kind of DLL | 14:18.16 |
| Robin_Watts | pgogna: That's under android? | 14:18.24 |
| mvrhel_laptop | kens2: right | 14:18.36 |
| pgogna | yes | 14:18.41 |
| Robin_Watts | sounds plausible, yes. | 14:19.36 |
| pgogna | But i am getting the text of second page in the bounding box | 14:20.21 |
| Robin_Watts | pgogna: The first page in the document is page 0, right ? | 14:20.53 |
| pgogna | If i double tap on first page then i am getting the text of the bounding box of second page | 14:21.08 |
| | yes | 14:21.57 |
| | Its 0 | 14:22.12 |
| Robin_Watts | pgogna: are you call run_list or run_page ? | 14:23.07 |
| pgogna | I called fz_run_display_list | 14:23.41 |
| Robin_Watts | Right, so where did you get the display list you are passing to it ? | 14:24.00 |
| paulgardiner | Ah here it is "To make more clear, if you have X.dll which calls GS.dll (both are WinRT dlls), it will not be possible for X.dll to load GS.dll unless you have winmd file added to the winRT APP." | 14:26.09 |
| pgogna | I am passing the page_list from page_cache returned by &glo->pages[glo->current] | 14:26.57 |
| paulgardiner | So we might suggest that, rather than trying to build both X.dll and GS.dll as WinRT dlls, he might build GS.dll as a win32 dll and link from X.dll using the .lib file. | 14:27.36 |
| | Ah! mvrhel_laptop went to lunch. I forgot. | 14:28.08 |
| kens2 | making a 'proper' winrt component just seems the right thing to do | 14:28.40 |
| Robin_Watts | sorry, I missed all that while I was out running, then went for lunch without seeing it. | 14:28.55 |
| | AIUI, Raed is NOT trying to call from C# or .net or whatever. | 14:29.19 |
| pgogna | Robin_Watts: I get the page_list from page_cache that is returned by &glo->pages[glo->current] | 14:29.21 |
| kens2 | Robin_Watts : so why does he need a winmd file then ? | 14:29.35 |
| Robin_Watts | kens: He says that he wants to have his own dll, X.dll, make calls to gs.dll. | 14:30.02 |
| | but in order for him to put both X.dll and gs.dll into a "windows store app", he needs winmd files for BOTH dlls. | 14:30.21 |
| kens2 | but from what paulgardiner syas, he can do that | 14:30.29 |
| Robin_Watts | kens2: It's quite possible that Raed is wrong. | 14:30.50 |
| kens2 | I am puzzled, but then what do I know ?..... | 14:31.04 |
| | Of course, he can't do it using our .dll anyway because that's not built with VS 2012 | 14:31.25 |
| Robin_Watts | What we need to be able to do is to authoratatively show that he is wrong. | 14:31.31 |
| paulgardiner | I'm working on it, and perhaps close | 14:32.05 |
| Robin_Watts | kens2: Our existing build scripts support us adding "METRO=1" to the nmake calls, to make the required build changes. | 14:32.18 |
| kens2 | I'm still inclined to feel that making a winrt component is the right thing to do anyway | 14:32.18 |
| Robin_Watts | and I have a VS2012 project, that I've shared with paulgardiner | 14:32.33 |
| paulgardiner | On msdn: "You can use Visual Studio 2012 to create either a standard Win32 DLL or a Windows Runtime component DLL that can be consumed by Windows Store apps" | 14:32.40 |
| Robin_Watts | (just a copy of the VS2005 one with METRO=1 added in the appropriate places) | 14:33.04 |
| paulgardiner | He may not be able to call it, but he will be able to "consume" it! | 14:33.13 |
| Robin_Watts | paulgardiner: I'm not sure I agree with your reading of that. | 14:33.30 |
| | I'm still catching up with the logs. | 14:34.09 |
| paulgardiner | Ah right. What might it mean. | 14:34.09 |
| kens2 | I think we need to consider that MS are definitely pushing people to using managed code, and our 'wrapping' GS up in a winrt component makes that easier, we should think about doing it | 14:34.41 |
| | Provided its not a lot of effort, and michael seems to think it isn't | 14:34.54 |
| paulgardiner | "A standard DLL that was created by using an earlier version of Visual Studio or an earlier version of the Visual C++ compiler may not load correctly in a Windows Store app and may not pass the app verification test in the Windows Store." | 14:34.54 |
| Robin_Watts | Right. What mvrhel_laptop suggests sounds like a good plan. It's something worth having and it expands the uses for gs. But I'm not sure it helps with Raeds problem. | 14:35.27 |
| kens2 | Ah, well that's a differnet issue :-) | 14:35.43 |
| mvrhel_laptop | I think that is what Raed is going to have to do also and live with | 14:36.02 |
| kens2 | He will need to make his 'x.dll' a winrt component as well ? | 14:36.29 |
| paulgardiner | I thought his x.dll was a winrt component. | 14:37.01 |
| Robin_Watts | paulgardiner: Right. My reading of the problem was that x.dll was a winrt component. | 14:37.38 |
| kens2 | Possibly at this point we should get Raed to send a concrete (skeleton) exmaple of what he needs so we cna run some tests | 14:37.42 |
| paulgardiner | msdn has instructions for migrating existing dlls. We could just point him at those instructions. | 14:38.35 |
| kens2 | Presumably that incolves rebuilding the DLL ? | 14:39.02 |
| pgogna | Robin_Watts: Could you please help me to achieve what i am trying to | 14:39.21 |
| paulgardiner | kens2: Yes, but not adding explicit interfaces or putting on a c++ wrapper | 14:39.40 |
| | Or needing to generate a winmd file | 14:39.55 |
| kens2 | I get the feeling he wants to void rebuilding GS | 14:40.00 |
| Robin_Watts | pgogna: I am confused as to why what you are doing is not working. | 14:40.11 |
| kens2 | paulgardiner : I'm guessing he thinks that if he uses the standard DLL he cna give us bugs, and that if he rebuilds it we'll shrug and say 'show it on standard GS' | 14:41.26 |
| pgogna | ok that means i am doing it correctly | 14:41.27 |
| Robin_Watts | paulgardiner: "msdn has instructions for migrating existing dlls" - so those instructions would be for what he needs to do to alter the way X.dll is built ? | 14:42.15 |
| paulgardiner | Alter the way gs is built, I thought. | 14:42.35 |
| | gs.dll | 14:42.41 |
| kens2 | If x.dll is a winrt component, I'm assumign he would nbeed to rebuild GS | 14:42.46 |
| Robin_Watts | paulgardiner: Then why would we point him at those instructions? Surely *we* should be following those instructions? | 14:43.03 |
| kens2 | Robin_Watts : if we are supporting winrt yes, hence my poitn that we should offer a winrt component..... | 14:43.31 |
| Robin_Watts | kens2: Offering a winrt component is one thing (and I agree that it would be a good thing to offer). | 14:43.54 |
| kens2 | Its probably as much effort as migrating (and maintaining) a DLL | 14:44.26 |
| paulgardiner | Robin_Watts: yes, I see what you mean; I was thinking of the continuation of the process where you were suggesting build options to him. | 14:44.42 |
| Robin_Watts | If we need to take steps to ensure that the existing gs.dll can be called *in the same way as it currently can* from windows rt components, then we should do that too. | 14:44.47 |
| | Palming that off to a customer feels like poor service to me. | 14:45.01 |
| kens2 | Robin_Watts : but can that same DLL then be used by non-winrt components ? | 14:45.09 |
| paulgardiner | No fine. That's what I've been working towards. | 14:45.23 |
| kens2 | especially applications built with earlier versions of VS.... | 14:45.35 |
| Robin_Watts | kens2: I don't know for sure, but they sound like different jobs to me. | 14:45.36 |
| kens2 | Robin_Watts : if the DLL after being migrated can't be used by older applications, tehn I don;t really see the point | 14:46.00 |
| | better to do it as a component I'd have thought (assuming that the component can be uzsed the way the DLL woujld have bneen) | 14:46.35 |
| paulgardiner | AFAIK the msdn suggesting builds an ordinarey win32 dll usable from anything. | 14:46.37 |
| kens2 | Well, if that works, it makes sense | 14:46.50 |
| Robin_Watts | kens2: I think our reading of the situation is diffrent here. | 14:46.59 |
| kens2 | Robin_Watts : yes, which is why I thought getting a skeleton applicatin from Raed might be helpful.... | 14:47.28 |
| Robin_Watts | I believe the situation is that Raed has an app that wraps ghostscript. And he's integrated the clever 'guts' of his app into a DLL. | 14:47.33 |
| | The idea is that he can then have a thin application that runs on Win32 to call the DLL which calls through to GS and everything is happy. | 14:48.03 |
| | Now, he wants to get that working on Windows 8. | 14:48.13 |
| kens2 | wRight, but we seem to be confused over whether that (x.dll) is a winrt component or just another dll | 14:48.21 |
| Robin_Watts | He'd expect to have to rewrite the thin application for windows 8 (possibly by using C# or something). | 14:48.48 |
| paulgardiner | I believe he has a winrt version of x.dll | 14:48.56 |
| Robin_Watts | That will mean that he needs to tweak X.dll to be a winrt thing. All makes sense so far. | 14:49.10 |
| | Why should he have to rework the X.dll -> GS.dll code though? | 14:49.22 |
| kens2 | but we don't know for sure, I still thnk asking him for a skeleton probject would show us for *certain* what he is doing | 14:49.30 |
| Robin_Watts | kens2: Indeed, a skeleton project would help clarify this. | 14:49.48 |
| | but if what I have just described is correct, then: | 14:49.56 |
| | 1) a winrt component for gs.dll, while desirable in it's own right, will not help raed avoid work. | 14:50.16 |
| pgogna | Robin_Watts: can you look at my function http://pastebin.com/abnzZU5G to get the bounding box of enclosed point? | 14:50.17 |
| Robin_Watts | 2) we should take steps to ensure that gs.dll can be called in the old way (assuming they aren't massively onerous). | 14:50.56 |
| kens2 | Robin_Watts : yes, I agree with both, but I think we need to know which one Raed actually needs, and if *either* will solve his problem | 14:51.23 |
| Robin_Watts | kens2: Right. | 14:51.34 |
| | So paulgardiner could write a mail to Raed introducing himself, describing what he's discovered, the possible routes forward and asking for clarification on exactly what Raed needs? | 14:53.28 |
| | paulgardiner: Words of 2 syllables though. | 14:53.40 |
| kens2 | Yes, but if it was me I would explicitly ask for a simple sample project | 14:53.53 |
| Robin_Watts | right. That too. | 14:54.07 |
| kens2 | the 'x.dll' need be nothing more than a simple 'passthrough' to GS, as long as it replicates what he is actually doing | 14:54.19 |
| paulgardiner | Sounds like a plan | 14:56.00 |
| Robin_Watts | pgogna: That code looks plausible enough. So you're clicking on page 1 of the document, and you're seeing it report "glo->current : 0" and yet the text it reports is from page 2? | 15:09.04 |
| pgogna | I click on page 1 and it reports "glo->current : 1" and reports text of 2nd page | 15:10.59 |
| Robin_Watts | Right. The pages are numbered from 0. | 15:11.47 |
| | so if you click on page 1, then you should hope for glo->current = 0 | 15:12.02 |
| pgogna | And next time when i click then it reports "glo->current : 2" and reports the correct text.. | 15:12.43 |
| Robin_Watts | Eh? | 15:13.04 |
| | When glo->current = 2 it reports the text from page 1? | 15:13.13 |
| pgogna | yes! | 15:13.23 |
| Robin_Watts | I don't understand that :( | 15:14.18 |
| pgogna | And even more when i move to page 2 and hit there then it reports "glo->current = 0" and reports text from 3rd page | 15:14.54 |
| Robin_Watts | Try calling gotoPageInternal(desired page num) before processing the click ? | 15:15.42 |
| | Ah, no. core.gotoPage(page) | 15:16.25 |
| pgogna | ok let me try | 15:16.51 |
| Robin_Watts | If you look in MuPDFCore.java, a lot of functions always call gotoPage(pageNum) before then doing stuff. | 15:17.12 |
| | searchPage being one of them. | 15:17.24 |
| mvrhel_laptop | brb | 15:21.31 |
| pgogna | Yes i got it and It worked | 15:21.36 |
| Robin_Watts | pgogna: Fab. | 15:22.03 |
| pgogna | but still glo->current giving strange numbers | 15:22.05 |
| Robin_Watts | pgogna: Right. Don't worry about glo->current if it's consistently giving the right results. | 15:22.26 |
| pgogna | Now lets come to step 2 | 15:23.11 |
| | Now i found the correct rect on which user double tapped | 15:23.33 |
| Robin_Watts | pgogna: I have an email to write to a customer which has to take priority. Let me do this first. | 15:23.42 |
| | In the meantime, find me an example bbox that you get returned, and the current page transformation that goes with it. | 15:24.13 |
| pgogna | But how do i calculate the scaling factor as i am scaling it in android and then rendering the page | 15:24.14 |
| | ok no problem | 15:24.39 |
| | catch you later with the info you want | 15:25.20 |
| Robin_Watts | pgogna: ok. | 15:27.46 |
| | paulgardiner: Muahahaha! | 15:51.24 |
| ray_laptop | I guess mvrhel is still on kid delivery | 15:51.27 |
| Robin_Watts | I was cluster testing the old version. New version looks fixed. Thank heavens! | 15:51.53 |
| Robin_Watts | slows the fans on peeves a bit by killing his tests. | 15:52.54 |
| paulgardiner | Robin_Watts: Phew! | 15:53.52 |
| ray_laptop | chrisl: I wonder if we should default to GS_USE_MEMORY_HEADER_ID=0 in release builds. The information is not really very useful in non-DEBUG builds is it ? | 15:55.27 |
| | and it does is take up space, in release builds | 15:55.53 |
| Robin_Watts | ray_laptop: I suspect that it may be useful in non-DEBUG builds, but that's not to say that it should default to be on. | 15:59.51 |
| henrys | made it to st louis | 16:02.51 |
| ray_laptop | Robin_Watts: it might be possible to glean some information out of the header if we get a SEGV from a release build, but I'm sure our first approach to a reported problem is to try a DEBUG build, and only if we can't reproduce the problem, try a release build, but since the object sizes are larger with the header, the allocation pattern will change | 16:03.29 |
| Robin_Watts | ray_laptop: So it sounds like you're arguing for having it default to being on in release builds :) | 16:04.05 |
| ray_laptop | and the allocation pattern with release and debug builds is already slightly different due to things that are added for debug builds | 16:04.05 |
| henrys | bbiaw | 16:04.05 |
| ray_laptop | Robin_Watts: no, I'm trying to say (and not doing it well) that we aren't likely to try and dig into a problem on a release build, so why have the HEADER_ID | 16:05.14 |
| chrisl | ray_laptop: I figured a memory diagnostic tool which changed the memory layout was of very limited use, and most workstations are hardly short of memory these days | 16:05.23 |
| ray_laptop | IMHO, it would be quite difficult to trace around using GDB or something on a release build. | 16:06.14 |
| | chrisl: but DEBUG and release builds alreaddy change the memory layout | 16:06.47 |
| chrisl | ray_laptop: why? The pattern of allocations should be largely the same | 16:07.19 |
| Robin_Watts | ray_laptop: But we've all been in the position of having to in the past, right? It's specifically for such things that the ID is useful. | 16:07.34 |
| ray_laptop | chrisl: and we do still have embedded customers that are memory sensitive | 16:07.37 |
| chrisl | ray_laptop: which is why it's easy to disable - for those cases when memory is tight | 16:08.00 |
| ray_laptop | Robin_Watts: right, so being ABLE to build it for a release build may be something that we resort to iff we can't get it to fail on a debug build | 16:08.32 |
| pgogna | Robin_Watts: Do you have time to help me now? | 16:08.40 |
| Robin_Watts | pgogna: 2 minutes, then sure. | 16:08.49 |
| pgogna | Ok as i have got the info that you want. | 16:09.08 |
| chrisl | ray_laptop: but then we get a memory bug report, oh I'll enable the header ID - problem disappears...... | 16:09.22 |
| kens2 | I'm inclined to agree | 16:10.23 |
| Robin_Watts | what chrisl said. | 16:10.37 |
| ray_laptop | chrisl: marcosw usually tries to reproduce on a release build. Then when (me, and probably most of us) start to look into it we generally use a DEBUG build | 16:10.39 |
| Robin_Watts | ray_laptop: And if we can make it fail under the DEBUG build, then great. | 16:10.57 |
| ray_laptop | I've rarely seen problems that are not reproducible on a DEBUG build. | 16:11.16 |
| Robin_Watts | but if we can't, I'd *way* rather be working with a release build with the header ID in than without. | 16:11.17 |
| kens2 | nods | 16:11.26 |
| | and I have seen many cases that will not fila on a debug build | 16:11.43 |
| Robin_Watts | ray_laptop: Most problems I see that happen in the release build, but not in the debug build are indeterminisms from the cluster. | 16:11.53 |
| kens2 | I';vew debugged several while doing this PCL work | 16:11.54 |
| ray_laptop | poor kens2 (I hate those) | 16:12.08 |
| kens2 | ray_laptop : they were segv so not a huge problem, I built a profile version and reproduced them there | 16:12.30 |
| chrisl | ray_laptop: The last few gc problems I worked on went wrong in the same way in a release and debug build, but *any* other change made the problem disappear | 16:12.52 |
| ray_laptop | kens2: the profile build had the symbols, so you could get more info, right. You are lucky that the problem didn't go away then (I've seen that, but also rarely) | 16:13.32 |
| kens2 | yes, but since it was a segv it was sever enough not to go vanish (these *were* real problems) | 16:13.58 |
| Robin_Watts | pgogna: Go for it. | 16:14.05 |
| ray_laptop | well, the UK contingent seems in consensus. Just wanted to have it discussed. | 16:15.43 |
| pgogna | Robin_Watts: Page resolution: PageWidth=1224 and PageHeight=1584 | 16:16.46 |
| ray_laptop | chrisl: BTW, did you see my email about the other devices that use fseek and either check for seekable in a (potentially) non portable way, or don't even check. I was going to mention it here, but you weren't on IRC anymore | 16:16.50 |
| pgogna | Rect Corners : A(98.000000; 170.147781) and B(489.823669; 228.707779) | 16:16.56 |
| ray_laptop | chrisl: the one that is probably the only important one to fix is gdevpsu.c | 16:17.48 |
| | since that is a supported and modern device | 16:18.02 |
| chrisl | ray_laptop: I didn't, no. I did originally add code to gdev_prn_open_seekable() but I was concerned about unforeseen side effects | 16:18.16 |
| ray_laptop | I don't feel strongly about digifax and sgi deviecs | 16:18.27 |
| Robin_Watts | pgogna: What device are you on? | 16:18.27 |
| pgogna | Device resolution is width = 540 | 16:19.19 |
| chrisl | ray_laptop: your mail was filtered (wrongly) by one of my mail filters - I read it properly tomorrow, and have a look at the other devices. I have to finish now.... | 16:19.21 |
| pgogna | height = 960 | 16:19.25 |
| chrisl | s/I read it/I will read it.... | 16:19.38 |
| pgogna | Device name sony experia P | 16:19.39 |
| ray_laptop | chrisl: OK. Thanls | 16:19.43 |
| | or Thanks | 16:19.50 |
| kens2 | ray_laptop : see my email I have a fix in place for gdevpsu.c | 16:19.51 |
| | chrisl also see my email | 16:20.04 |
| Robin_Watts | pgogna: OK, so the X extent of that bbox is from 98 to 489. | 16:20.06 |
| ray_laptop | kens2: I didn't see your email, sorry. | 16:20.09 |
| | when was it ? | 16:20.15 |
| chrisl_away | kens2: got it, too, yes - thanks | 16:20.17 |
| kens2 | ray this morning | 16:20.24 |
| pgogna | Yes correct | 16:20.26 |
| kens2 | (my time) | 16:20.28 |
| Robin_Watts | which seems plausible for screen coords of 0 to 540, right? | 16:20.37 |
| ray_laptop | is looking for ken's email ... | 16:20.53 |
| kens2 | ray_laptop : 11:57 UK time, I guess that would be 03:57 your time | 16:21.01 |
| Robin_Watts | (i.e. you've got about 100 pixels to the left of the text, and about 50 to the right) | 16:21.06 |
| kens2 | went to you and cc'ed to tech | 16:21.09 |
| | and chris | 16:21.17 |
| ray_laptop | kens2: Ok, have it now. Thanks. | 16:21.31 |
| kens2 | NP | 16:21.35 |
| Robin_Watts | pgogna: So, what was the page transformation that you fed in for that ? | 16:21.47 |
| | i.e. the "current matrix" that you fed into run_display_list ? | 16:22.22 |
| ray_laptop | kens2: obviously it isn't urgent. It just irks me to have different methods to do the same thing. There's enough of that in GS already. | 16:23.12 |
| kens2 | ray_laptop : agreed, but it was so simple I decided to just do it, also the code is practically identical to chris's | 16:23.35 |
| pgogna | fz_run_display_list(pc->page_list, dev, &fz_identity, &afz_infinite_rect, &cookie); This is what i am doing with the call to run_display_list | 16:24.54 |
| | It doesnot have any matrix? | 16:25.09 |
| Robin_Watts | You're feeding in fz_identity. | 16:25.20 |
| | That means that the bboxes you are getting back for the text are not in screen coords. | 16:25.54 |
| pgogna | yes its an identity matrix | 16:25.55 |
| Robin_Watts | They are in document space. | 16:26.03 |
| | You need to feed in the current page transformation matrix. I did go through this all yesterday. | 16:26.22 |
| pgogna | fz_matrix fz_identity = { 1, 0, 0, 1, 0, 0 }; this is what i am sending | 16:26.48 |
| ray_laptop | kens2: right. I think the only one that worked differently was the vms one (don't ask me why). VMS is so rare, it hardly warrants adding a gp_ function instead of just doing it the psdu way, but this makes porting to strange filesystems on embedded products easier | 16:27.06 |
| kens2 | Yes, completely agree. | 16:27.25 |
| Robin_Watts | pgogna: Look in mupdf.c at "drawPage" | 16:27.46 |
| kens2 | and if we need it for more than pswrite, tehn we should make a proper library call | 16:27.47 |
| | (as chrisl has done) | 16:27.57 |
| Robin_Watts | lines 684 to 692. | 16:28.14 |
| | There we calculate "ctm" (current transformation matrix) | 16:29.07 |
| | And we pass that into display list. | 16:29.18 |
| | That way you'll get rectangles out that are screen coords. | 16:29.35 |
| | OR you can stick with sending fz_identity, but you'll then need to transform the bboxes you get back to allow for the current zoom and the current x/y translation. | 16:31.32 |
| pgogna | Are you talking about running this code? http://pastebin.com/7CSpCfkc\ | 16:36.28 |
| | How do I get bbox before I run this code and pass the fz_identity to run display list?\ | 16:37.11 |
| kens2 | ok goodnight all | 16:38.09 |
| Robin_Watts | pgogna: Right. That code you just pasted will take a rectangle in document space to one in screen coords. | 16:39.28 |
| | So if you feed fz_identity into the display list, you'll get bboxes back in document space. | 16:39.51 |
| | So you'll then need to feed those bboxes through that code to get them converted to screen coords. | 16:40.21 |
| | Then you can test to see if the point you clicked is within them. | 16:40.49 |
| | Does that make sense ? | 16:40.56 |
| ray_laptop | hi, mvrhel_laptop | 16:44.33 |
| | mvrhel_laptop: did you see my problem description in the logs (at the top) ? | 16:44.56 |
| pgogna | I think I understand what you are saying, What you mean is that in my current flow of code, I should take the retuning bbox and run it through the code I just shared with you. After that I will have the new bbox coordinates which are already transformed to the current device resolution and zoom level. | 16:45.07 |
| Robin_Watts | Yes. | 16:45.18 |
| | At the moment, without scaling the returned bboxes, the bbox for a given piece of text won't change depending on how you zoom/pan it. | 16:47.16 |
| pgogna | ok I understand this. I am clear. This is basically a fix to a problem that if we scale, zoom the pdf than our bbox coordinates are not accurate. This is fine. However I still need to understand the next step. Once I have the BBox than what should be my next step to find the scaling factor ?\ | 16:48.47 |
| Robin_Watts | pgogna: Well, look at the piece of code you just gave me. | 16:49.33 |
| | That code shows how to build a transformation matrix that transforms a given 'document space' rectangle into a screen space one. | 16:50.19 |
| | You need to run that process in reverse. | 16:50.46 |
| | To start with you have a bbox (x0, y0) -> *x1,y1) | 16:51.19 |
| | To start with you have a bbox (x0, y0) -> (x1,y1) | 16:51.23 |
| | and you want to set the zoom in such a way that that bbox fills the screen widthways, yes? | 16:51.46 |
| | actually, where did you get that code from? | 16:53.11 |
| | It looks like you're scaling twice. | 16:53.29 |
| | oh. I see. | 16:56.30 |
| | The second scale is just to ensure that we get whole pixels out. | 16:56.41 |
| pgogna | Which code you are talking about? | 16:56.44 |
| Robin_Watts | pgogna: OK, ignore that. | 16:57.02 |
| | So, let's assume that you have a bbox: (x0, y0) -> (x1, y1) and you want to adjust the zoom so that that fills the screen widthways, right? | 16:57.46 |
| pgogna | Yes | 16:57.56 |
| Robin_Watts | So the width of the bbox is w = x1-x0 | 16:58.07 |
| | Let's 'unzoom' that to get back to document space. | 16:58.46 |
| | document_space_w = w * 72/glo->resolution | 16:59.06 |
| | Now, we want to find what the new resolution should be, such that it would map that width to be the screen width. | 16:59.52 |
| | i.e. if screen_width = document_space_w * new_res / 72, what is new_res ? | 17:00.27 |
| | rearrange to get: new_res = screen_width*72/document_space_w | 17:01.00 |
| | So you need to set glo->resolution = new_res | 17:01.15 |
| | Does that make sense? | 17:01.30 |
| pgogna | So what i understood is that, first i need to find out the bbox width and then applying it in the above equation will give me new resolution and set this new_res to current res will zoom the page | 17:04.42 |
| Robin_Watts | pgogna: Yes. | 17:05.05 |
| | I am sure that there will have to be work done to make sure that the x/y position of the page is correct too. | 17:05.35 |
| | and now I'm confused. Hold on, let me confer with a colleague. | 17:07.54 |
| pgogna | Ok.. meanwhile let me try what you said earlier | 17:08.47 |
| Robin_Watts | ok, resolution is a constant. You don't want to mess with that. | 17:13.27 |
| | ok. Ignore resolution. It's a complete and utter red herring. We should clear that out of the code. | 17:16.47 |
| | The important bit is in the second part of that code. | 17:17.08 |
| | The ctm is calculated so that it scales the pagebbox to be pageW wide. | 17:17.59 |
| pgogna | so how will actual scale happen? | 17:18.59 |
| Robin_Watts | You need to modify the settings so that in future we ask for pageW*screen_width/w | 17:19.03 |
| pgogna | what settings? | 17:19.30 |
| Robin_Watts | The page size/position calculations are all done in the java layer. | 17:19.39 |
| | and when we call drawPage, we pass in the desired size of the complete scaled page (pageW, pageH) and the bbox to redraw. | 17:20.05 |
| | So you need to pass screen_width/w back up to the java layer and apply it there. | 17:20.27 |
| | You need to read and understand the java stuff. It's not in my head at the moment, and I can't spare the time to go trawling through it in detail. | 17:23.26 |
| | essentially, each Page has a PageView, I think, which knows the size/position of itself. | 17:24.02 |
| | zooming/panning works by changing the size/position of each pageview. | 17:24.29 |
| | You need to somehow get the data up from your bbox detection into changing the pageview size/positions. | 17:24.59 |
| pgogna | Ok, So you are trying to say that take the rect to java layer and perform zoom from there | 17:25.22 |
| Robin_Watts | yes. | 17:25.30 |
| pgogna | I tried that and used " scaleFactor = 0.95f/Math.max(rect.width()/metrics.widthPixels, rect.height()/metrics.heightPixels); " to calculate the scaling factor | 17:27.08 |
| Robin_Watts | tor8: ping | 17:29.18 |
| ray_laptop | mvrhel_laptop: Are you still around ? | 17:30.45 |
| Robin_Watts | tor8: Many reviews for you on robin/master (all except the last one good to go, I think) | 17:31.13 |
| mvrhel_laptop | ray_laptop: sorry on phon | 17:31.23 |
| ray_laptop | I think I found a way to handle equiv_colors that doesn't need a new device proc. | 17:31.27 |
| mvrhel_laptop | I will call you back | 17:31.31 |
| | on hold.... | 17:31.35 |
| ray_laptop | mvrhel_laptop: or you can just chat here if you want (when you are off the phone) | 17:31.55 |
| | (that'll make henrys happy, who encourages us to chat here so others can monitor and maybe learn something) | 17:32.30 |
| mvrhel_laptop | since I am on hold I can chat now here if you want. | 17:32.56 |
| Robin_Watts | Your call is important to us. | 17:34.16 |
| Robin_Watts | hums some musak. | 17:34.32 |
| ray_laptop | Robin_Watts: even worse than musak (muzak?) is when they play some top 40 radio station that even has commercials and noisy songs | 17:44.19 |
| Robin_Watts | There was one place in the UK that used to play star trek sound effects. It was great :) | 17:45.01 |
| mvrhel_laptop | ok. finally done with that | 17:45.51 |
| | yes. the commercials are bad | 17:46.10 |
| ray_laptop | mvrhel_laptop: so, the devn_[get put]_params functions have pequiv_colors but don't really ever write the equiv_colors to the param_list, and don't do anything in 'put' except clear the 'valid' flags | 17:47.56 |
| pgogna | Robin_Watts: can you tell me any way to highlight the bounding box for debugging purpose? | 17:48.36 |
| ray_laptop | mvrhel_laptop: what do you think about adding the equiv_color cmyk values and the valid flag to the param list so the put_params will set it ? | 17:48.48 |
| mvrhel_laptop | ray_laptop: I believe that would be a good thing since it keeps it from being this oddball special thing | 17:49.24 |
| ray_laptop | OK. I'll look at doing that | 17:50.06 |
| mvrhel_laptop | where was henry off to these 2 days? | 17:50.28 |
| ray_laptop | It keeps me from having to add a new device proc | 17:50.31 |
| mvrhel_laptop | ray_laptop: yes | 17:50.39 |
| ray_laptop | mvrhel_laptop: he mentioned St. Louis in the logs today, but I don't know if that is where he was stopping, or just a checkpoint | 17:51.12 |
| | mvrhel_laptop: any suggestions on the param name (I assume it will be an arrray). And do you think an array of N*5 numbers (where N is the number of valiid equiv_colors) is OK ? | 17:54.15 |
| Robin_Watts | pgogna: You can invert sections of pixmap I think. | 17:54.39 |
| mvrhel_laptop | ray_laptop: is 5 for CMYK + alpha? | 17:54.58 |
| Robin_Watts | void fz_invert_pixmap_rect(fz_pixmap *image, const fz_irect *rect); | 17:55.21 |
| ray_laptop | mvrhel_laptop: the struct cmyk_color_s doesn't have alpha | 17:55.29 |
| pgogna | where can i find that part of code? | 17:55.42 |
| mvrhel_laptop | oh 5 for the index + CMYK? | 17:55.59 |
| ray_laptop | mvrhel_laptop: actually, the first 'valid' bool is in case a sep gets added and its info is still not valid | 17:57.22 |
| mvrhel_laptop | ray_laptop: oh ok. valid. that is right | 17:57.36 |
| ray_laptop | storing it as a number is just a hack to have the array be homogenous | 17:57.57 |
| mvrhel_laptop | yes | 17:58.03 |
| ray_laptop | (well mixed). oops | 17:58.05 |
| | homogeneous ? | 17:58.17 |
| mvrhel_laptop | something like equiv_cmyk_color would makes sense to me | 17:58.28 |
| | or equiv_color is fine | 17:58.43 |
| | if we dont want to be explicit about a color space | 17:58.55 |
| ray_laptop | mvrhel_laptop: since it is an array of N colors, then equiv_cmyk_colors seems good. Also if we are writing bool(int), C, M, Y, K then beiing explicit about the colorspace seems OK | 17:59.58 |
| mvrhel_laptop | ray_laptop: sounds good to me | 18:00.15 |
| ray_laptop | mvrhel_laptop: would you object to ".EquivCMYKColors" (a few other non-standard params have an initial ".") | 18:02.30 |
| mvrhel_laptop | ray_laptop: I would not object to that | 18:02.51 |
| ray_laptop | anybody else want to weigh in? | 18:02.55 |
| | mvrhel_laptop: thanks. I think I'll try this approach then. | 18:03.16 |
| | probably some CET that dumps all the device params will pop up with a diff, but that's easy. | 18:04.29 |
| | mvrhel_laptop: YUCK. I think I remember tripping over this in the past. The devn_get_params seems like it is just a placeholder dummy function | 18:17.16 |
| mvrhel_laptop | ray_laptop: that whole devn parameters stuff is a big mess and could do with some serious rewrite | 18:18.13 |
| ray_laptop | mvrhel_laptop: looks like I drew the short straw :-( | 18:18.35 |
| | of course, I could just do like I did with PageSpotColors and just add what I need for now (the precedent has been established ;-) ) | 18:20.15 |
| Robin_Watts | ray_laptop: If you leave it just a bit nicer than it was when you arrived, you can consider yourself on the side of sainthood :) | 18:21.00 |
| mvrhel_laptop | yes. sort of like with camping. leave it better looking than when you got there | 18:23.57 |
| chrisl | ray_laptop: I was thinking about the positional file fix for the TIFF devices, and the more I thought it, the more I think the fix *should* go in gdev_prn_open_printer_seekable() - I don't think it should be able to silently ignore the seekable flag. | 18:33.52 |
| ray_laptop | chrisl: that's fine with me. Except we still have to fix up (or get rid of) the dfax and sgi devices since they don't call gdev_prn_open_printer_seekable in their open (as gdevtifs and gdevtfax do). | 18:40.57 |
| | but having the check in one place make sense to me | 18:42.02 |
| chrisl | ray_laptop: cool, I'll do that tomorrow - and fix up the errant devices, too | 18:42.35 |
| | ray_laptop: and thanks! | 18:43.32 |
| ray_laptop | chrisl: of course, ken's change is still valid because he just emits different code (BoundinBox: atend) if the output file is not seekable | 18:51.03 |
| chrisl | ray_laptop: oh, okay, I might need to tweak it to ignore an error message. | 18:52.26 |
| | error code, rather | 18:52.37 |
| Robin_Watts | tee hee. I've got a fix for the vertical text motion bug in mupdf. It's superbly random. Tor will hate it. | 19:02.02 |
| | tor8: Another review on robin/master for you then. | 19:16.22 |
| tor8 | Robin_Watts: I believe I tried that one already. have you localclustered it? | 19:20.38 |
| Robin_Watts | tor8: I have. You can see the bmpcmps currently. | 19:20.52 |
| tor8 | maybe I flipped the same sign elsewhere then... | 19:21.13 |
| Robin_Watts | essentially rounding errors in about 4 files. | 19:21.17 |
| | actually... | 19:22.16 |
| | What I tried initially was leaving it as - but using abs() on the font size. | 19:22.35 |
| | and that gave 24 differences (4 files). | 19:23.07 |
| | but I don't think my latest bmpcmp worked. | 19:23.22 |
| | Let me retest it. | 19:23.25 |
| tor8 | yeah. locally on the test files it has the same problems as my initial attempt... | 19:24.21 |
| Robin_Watts | yeah, I think the version with abs is the right one. | 19:25.09 |
| | Like I say, pretty random. | 19:25.18 |
| | I'll fix my commit. | 19:25.29 |
| tor8 | Robin_Watts: do you recall if you (or paul) have changed the x11 main event loop since 1.1? | 19:26.01 |
| Robin_Watts | I won't have done. | 19:26.15 |
| tor8 | it looks like the dirty flag lazy expose event updating doesn't work as it should anymore :( | 19:26.24 |
| Robin_Watts | paulgardiner might have updated it to cope with text entry for forms? | 19:26.37 |
| | but actually that doesn't work on linux yet, does it? | 19:26.52 |
| tor8 | it can't have been, that's just using gets() I think | 19:27.01 |
| | maybe the slide show transitions changed something? | 19:27.17 |
| Robin_Watts | Possibly :( | 19:27.26 |
| | Company 530 prodded me today to ask about the mesh stuff. | 19:27.48 |
| tor8 | realized the tax declaration deadline was looming so my brain got fried doing that today :( | 19:29.13 |
| Robin_Watts | ah. I lose 1 day a quarter to accountancy crap :( | 19:29.43 |
| | OK, updated patch on my repo. | 19:30.23 |
| | I'll rerun the tests for sanitys sake. | 19:30.31 |
| tor8 | if it really is abs() then it's absurdly random and I'll love it :) | 19:31.05 |
| Robin_Watts | :) | 19:31.22 |
| tor8 | Robin_Watts: yeah, you broke the x11 app with the transitions :( | 19:36.37 |
| | Robin_Watts: while (... XPending()) you changed to while (... XPending() && !dirty) | 19:37.27 |
| | which forces it to drop out of the loop as soon as it's flagged dirty, and forces a new blit | 19:37.44 |
| | where before it'd eat all events and *then* blit once it's idle again | 19:37.57 |
| | Robin_Watts: check tor/master for a potential fix (that might break transitions, I haven't tested those) | 19:41.43 |
| Robin_Watts | tor8: sorry. | 19:45.42 |
| tor8 | I take the blame, I should have caught it in the review | 19:47.36 |
| Robin_Watts | ok, so we should find a way to fix transitions too. | 19:50.27 |
| | probably it should be && (!dirty || !transition_mode) or something. | 19:50.50 |
| | I'll look into that tomorrow. | 19:52.48 |
| | can we hold off on your commit til then? | 19:53.13 |
| | at the moment everything works, just slowly. | 19:53.53 |
| tor8 | yes, we can wait for a proper fix | 19:55.47 |
| ray_laptop | mvrhel_laptop: I have the separation colors coming out of psdcmyk correctly now with BGPrint | 20:04.31 |
| mvrhel_laptop | ray_laptop: great! | 20:04.49 |
| ray_laptop | souping up the devn_get_params and devn_put_params was enough | 20:06.06 |
| | now I have the cups stuff to look at. | 20:07.42 |
| | hmm... looks like all of the cups diffs I have to looked at so far are multi-page files. The honked up cups device probably makes assumptions that aren't valid w/ BGPrint | 20:20.12 |
| | time for a lunch break | 20:20.23 |
| Guest55135 | Hello | 22:19.58 |
| ghostbot | hey | 22:19.58 |
| Guest55135 | Is there anyone here? I'm trying to fix a problem | 22:23.10 |
| | using the tiffpack device is giving me an error | 22:35.57 |
| | other tiff devices work fine | 22:36.13 |
| | I'm using 9.07 | 22:36.16 |
| | Forward 1 day (to 2013/04/26)>>> | |
IRC Logs