| | <<<Back 1 day (to 2013/06/12) | 2013/06/13 |
| mvrhel_laptop | off to sleep. Robin_Watts for the logs, I wrote a simple locking cache for my display list (lust) objects in the winrt project. This should fix the issue that you pointed out. I need to beat on it a bit to make sure all is well. It may be a day or so before I commit it for you to take a look at | 06:51.55 |
| kens | chrisl ping | 08:47.25 |
| chrisl | kens: poing | 08:58.42 |
| kens | chrisl, got a build problem but I need a few minutes.brb | 09:03.55 |
| chrisl | Okay | 09:04.01 |
| | kens: going to grab a drink, back in a sec | 09:05.38 |
| kens | no problem | 09:05.44 |
| chrisl | back | 09:10.42 |
| kens | right, sorry, was in middle of build, still failed. | 09:15.01 |
| | I'm implementing the ramfs files, to test with | 09:15.13 |
| | I have 3 files, 2 C, on .h | 09:15.24 |
| | I've modified lib.mak : | 09:15.47 |
| | # ---------------- Support for %ram% IODevice ----------------- # | 09:15.48 |
| | gsioram_h=$(GLSRC)gsioram.h $(GLSRC)ramfs.h | 09:15.48 |
| | ramfs_=$(GLOBJ)gsioram.$(OBJ) $(GLOBJ)ramfs.$(OBJ) | 09:15.48 |
| | $(GLD)ramfs.dev : $(LIB_MAK) $(ECHOGS_XE) $(ramfs_) | 09:15.48 |
| | $(SETMOD) $(GLD)ramfs $(ramfs_) | 09:15.48 |
| | $(ADDMOD) $(GLD)ramfs -iodev ram | 09:15.48 |
| | $(ADDMOD) $(GLD)ramfs -obj $(GLOBJ)ramfs.$(OBJ) | 09:15.48 |
| | $(GLOBJ)ramfs.$(OBJ) : $(GLSRC)ramfs.c | 09:15.49 |
| | $(GLCC) $(GLO_)ramfs.$(OBJ) $(C_) $(GLSRC)ramfs.c | 09:15.49 |
| | and makefile.in I've added ramfs.dev to FEATURE_DEVS and FEATRUE_DEVS_ALL | 09:16.12 |
| | But its not building the c files the buidl finishes though, without errors | 09:16.35 |
| | So clearly I've not put some missing link in place | 09:16.46 |
| chrisl | What platform? | 09:16.58 |
| kens | Windows | 09:17.01 |
| chrisl | makefile.in is used in the configure buld | 09:17.15 |
| | bulid | 09:17.19 |
| kens | Duh | 09:17.20 |
| chrisl | build | 09:17.24 |
| | (having a bad day!) | 09:17.34 |
| kens | Me too I think | 09:17.43 |
| chrisl | You need to modify psi/msvc.mak | 09:17.48 |
| kens | Hmm, any idea where ? | 09:18.10 |
| chrisl | line ~937 | 09:18.43 |
| kens | AH got it, thanks | 09:18.58 |
| kens | tries another build | 09:19.22 |
| chrisl | msvc.mak doesn't have a FEATURE_DEVS_ALL variable | 09:19.25 |
| kens | Yeah, I see that | 09:19.33 |
| | Did a 'clean', now the build is off, lets see what happens.... | 09:19.56 |
| | Oh well, I guess makefil.in woudl have needed to be modified anyway | 09:20.49 |
| | OK build failed, that progress | 09:21.12 |
| | Forgot to add a build directive for one of the C files.... | 09:24.16 |
| Robin_Watts | kens: Urm... it sounds wrong to me that you're doing platform inspecific things in the platform specific build files. | 09:27.28 |
| | Surely you should be altering gs/base/lib.mak (or similar) that is used both for ms and non ms builds? | 09:28.02 |
| kens | Robin_Watts : as far as I know, I need to modify msvc.mak and makefile.in to alter FEATURE_DEVS | 09:28.06 |
| Robin_Watts | oh, sorry! You have changed lib.mak. | 09:28.27 |
| kens | Yep :-) | 09:28.33 |
| Robin_Watts | Yes, you need to change the device lists too, and those are in the platform specific stuff. | 09:28.45 |
| | I'll get me coat. | 09:28.49 |
| kens | Not a problem, I'm ahead of you out the door already | 09:29.01 |
| | Now I just need to bring this code up to the current GS, which may take me a while..... | 09:29.29 |
| kens | needs more caffeine | 09:30.51 |
| Robin_Watts | reboot | 09:51.47 |
| | tor8: ping | 09:57.00 |
| | Did you have an opinion on the lexing fix? | 09:57.28 |
| tor8 | Robin_Watts: I'll check soon. buy components to make a new computer so I can run an nvidia card on linux. | 09:59.44 |
| | s/buy/i'm shopping for/ | 09:59.56 |
| Robin_Watts | Which NVidia card? | 10:02.07 |
| | I have a GTX 560 Ti, but that's got to be at least a year old now. | 10:02.51 |
| tor8 | I have not decided. preferably not a super-noisy one... | 10:04.45 |
| Robin_Watts | http://www.quietpc.com/gv-n66toc-2gd | 10:09.52 |
| | Morning paulgardiner | 10:10.33 |
| paulgardiner | hi | 10:10.47 |
| | That cluster test seemed okay. Did you suss out what was happening? | 10:13.04 |
| Robin_Watts | paulgardiner: Yeah, the thirdparty dir for my user tests was screwed. | 10:15.48 |
| | I think everything should be working now. | 10:15.53 |
| paulgardiner | oh right. good | 10:16.09 |
| Robin_Watts | tor8, paulgardiner: And a fix from zeniko for a lock problem. | 10:22.35 |
| | (on robin/master that is) | 10:22.48 |
| paulgardiner | Presumably fz_hash_find and fz_hash_insert can't throw? | 10:32.43 |
| | Possibly could be more obviously correct if more of the code was within the try/catch clause. | 10:34.14 |
| Robin_Watts | fz_hash_insert can throw. | 10:39.09 |
| | new version up with that in mind, thanks. | 10:41.56 |
| | oh, wait. | 10:43.33 |
| | damn. should have left it as it was. | 10:43.40 |
| | The fz_hash_insert is in a try/catch of its own. | 10:44.34 |
| | old version back in place. | 10:45.15 |
| paulgardiner | Ah. Sorry, so it is, with the throw turned into a warn | 10:49.46 |
| | Looks fine then, although I think stuffing the lot into the central try/catch wouldn't have hurt. | 10:50.51 |
| tor8 | Robin_Watts: the gt640 is the highest powered passively cooled nvidia card I can find | 11:01.45 |
| | ... and a computer to go with it :) | 11:01.52 |
| | alternative fix looks okay | 11:04.31 |
| | no objections to lock fix | 11:04.50 |
| | though I can't help but wonder why we don't just use fz_always for fz_unlock? is it because the fz_lock is taken inside the try block and putting it outside would be more awkward than keeping track of the lock with a variable? | 11:06.16 |
| | Robin_Watts: what does printf format %Ls do? | 12:07.25 |
| | trying a mingw build, and I'm getting warnings/errors about these MSVC specific things | 12:07.43 |
| Robin_Watts | tor8: ah, yes. | 12:09.34 |
| tor8 | %ls compiles fine with mingw | 12:09.48 |
| | but %Ls warns about unknown format specifier | 12:09.57 |
| Robin_Watts | If I move the val && do_cache case into the try/catch, then I can use fz_always. That's nicer. | 12:10.17 |
| | Looking at that particular piece of code... if the fz_hash_insert fails, it looks to me like we still increase the cache total. | 12:11.11 |
| | That seems wrong to me. | 12:11.18 |
| tor8 | yeah. that cache->total looks out of place. | 12:17.50 |
| | we never decrease the total when we evict a pixmap | 12:18.06 |
| | and if it throws due to a failed resize_hash we also bump the size | 12:18.30 |
| | it ought to be total -= pix->w * pix->h in the if(pix) case | 12:18.54 |
| | and total += after the keep_pixmap | 12:19.05 |
| | Robin_Watts: build fix for mingw (only command line tools, not the viewer) using make on tor/master | 12:24.22 |
| | biggest change is moving the main_utf8 stuff into a common file in the library and #define main in fitz.h | 12:24.43 |
| | that could be dangerous to expose publicly though, but we do the same with fopen... | 12:24.58 |
| Robin_Watts | tor8: 3 minutes to meeting | 12:57.58 |
| | how do we do this? | 12:58.02 |
| tor8 | Robin_Watts: check your skype | 12:58.16 |
| Robin_Watts | tor8: That seemed to go well. | 13:21.17 |
| | Time to buy paulgardiner more caffiene. | 13:21.26 |
| tor8 | yes. light on the technical side, but goals and targets seem to align very nicely. | 13:21.40 |
| Robin_Watts | tor8: I'll send an email out to scott ? | 13:21.59 |
| tor8 | doing their own interpreter is definitely not going to fly, not without significant investment of time. | 13:22.15 |
| | please do. | 13:22.17 |
| | and CC me | 13:22.26 |
| Robin_Watts | tor8: Sending mail to scott/miles/tech, so everyone has visibility of it. | 13:35.24 |
| tor8 | Robin_Watts: I think he said preloads, rather than freeloads | 13:38.04 |
| Robin_Watts | tor8: Ah! Do you have his email address? | 13:38.19 |
| | Scott will need that :) | 13:38.31 |
| tor8 | I do, and so do you (I CC:ed you when setting up todays meet) | 13:38.49 |
| Robin_Watts | d'oh. | 13:39.11 |
| | got it, thanks. | 13:39.16 |
| henrys | did paulgardiner nod out at the meeting ? ;-) | 13:40.05 |
| Robin_Watts | henrys: paulgardiner wasn't in the meeting. | 13:40.20 |
| henrys | oh you said time to get paulgardiner caffeine | 13:40.43 |
| Robin_Watts | but their prime list of "wants" are forms, annotations and digital signatures. | 13:40.48 |
| | Meth might work too. | 13:40.59 |
| paulgardiner | Yeah Meth's good, thanks | 13:41.11 |
| | Must be blue | 13:41.28 |
| Robin_Watts | Did you spot that the bag of meth in Season 1 of The Walking Dead is blue? :) | 13:41.54 |
| paulgardiner | Robin_Watts: yes, I saw that. Nice touch | 13:42.44 |
| Robin_Watts | henrys: http://www.kickstarter.com/projects/minimonkey/monkey-light-pro-bicycle-wheel-display-system?ref=category | 13:43.01 |
| | It'll make you race faster, honest. | 13:43.07 |
| | or for the people burdened with a smaller wallet: http://www.monkeylectric.com/m210_bike_light/ | 13:43.25 |
| kens | Think Geek were selling one of those ages ago. | 13:52.59 |
| | Or was it firebox.... | 13:53.05 |
| henrys | I wonder if we could license adobe's DRM stuff. | 13:53.34 |
| Robin_Watts | kens: That may have been the revolights.com | 13:53.42 |
| tor8 | henrys: DRM is stupid and evil, we shouldn't be helping our customers shoot themselves... | 13:54.21 |
| Robin_Watts | kens: ah. http://www.thinkgeek.com/product/eb7e/?srp=1 | 13:54.38 |
| henrys | tor8:I know but selling mupdf is not evil. | 13:54.59 |
| | I don't see how we could possibly get into the digital book space without drm. | 13:55.30 |
| Robin_Watts | henrys: Yes, licensing that would make us more compelling. | 13:55.53 |
| tor8 | henrys: we could start by having the core tech to do epub... | 13:55.56 |
| henrys | I "dedrm" all my ebooks I buy - can't stand it. But we should have it. | 13:56.40 |
| tor8 | henrys: shhh! the NSA are listening. | 13:57.12 |
| Robin_Watts | henrys: how? | 13:57.33 |
| henrys | calibre and the dedrm plugin | 13:57.53 |
| tor8 | calibre has dedrm plugins | 13:57.54 |
| Robin_Watts | oh, right. I buy in kindle format. | 13:58.22 |
| henrys | Robin_Watts:It read my entire kindle library | 13:58.42 |
| tor8 | Robin_Watts: calibre can convert to and from kindle format, so thankfully you're not as locked in as amazon wants you to be | 13:59.13 |
| Robin_Watts | nice! | 13:59.22 |
| tor8 | Robin_Watts: did you see the iain m banks sale on amazon? | 13:59.22 |
| Robin_Watts | tor8: I own them all. | 13:59.30 |
| henrys | it converts to pdf also, I like to make notes in my books. | 13:59.40 |
| tor8 | I don't, but ick, kindle and drm :( | 13:59.51 |
| henrys | I'm putting DRM on the workflowy | 14:00.12 |
| tor8 | henrys: are there actually pdf files with adobe drm out there? | 14:00.45 |
| henrys | yes nook licenses it. | 14:01.15 |
| tor8 | crikey. we should not support it. if enough people refuse drm, just maybe it'll go away... | 14:02.58 |
| paulgardiner | The risk of not supporting it is that it might be us that go away | 14:04.31 |
| | Can't see a Iain M Banks sale | 14:04.55 |
| tor8 | paulgardiner: all of his books are reduced price on amazon.co.uk | 14:05.44 |
| paulgardiner | Oh okay. I thought you might have meant a buy-the-lot deal | 14:06.48 |
| | I have just two books rattling about in my newly bought Nook. | 14:07.58 |
| Robin_Watts | paulgardiner: Did you get it rooted OK? | 14:09.03 |
| | tor8: OK, back to the glyphcache thing. | 14:09.36 |
| | Our current approach is to start with cache->total, and increase it til it's full. | 14:09.57 |
| paulgardiner | Yes, thanks. | 14:10.15 |
| Robin_Watts | When we get too big, we throw away ALL the contents of the cache. | 14:10.16 |
| tor8 | Robin_Watts: right. I tried an LRU eviction scheme way back when, but that ended up being slower due to shuffling all the data around for book-keeping | 14:12.19 |
| | I think I was trying to keep the glyph memory contiguous back then, but we don't do that any more so I think it may be worth revisiting | 14:12.35 |
| Robin_Watts | So, all we need to to to make it correct is to move the cache->total += ... into the try/catch. | 14:12.52 |
| tor8 | yes, to make it correct at the moment is to move the total update into the try/catch as I mentioned earlier | 14:13.25 |
| Robin_Watts | But we can't just do the unlock in an always block. | 14:14.07 |
| | the problem is that we unlock for a while during the operation in the t3 case. | 14:14.30 |
| | hence we still need the 'locked' flag. | 14:14.38 |
| | but I think the code can be simplified a bit. | 14:15.09 |
| | tor8: Yet another version of the locking fix online. | 14:24.46 |
| tor8 | Robin_Watts: um, did you forget to squish in the fix for the pdf parsing commit? | 14:26.32 |
| Robin_Watts | tor8: did I / | 14:27.03 |
| | ? | 14:27.05 |
| tor8 | the pdf_lex stuff with the 'repair' argument is still there | 14:27.19 |
| Robin_Watts | Sheesh. I forgot to strip that out, even though it's not used. | 14:27.46 |
| | will fix. | 14:27.54 |
| | commit to strip that out on robin/master now, sorry. | 14:30.34 |
| | Hmm. Calibre is extremely confused by the presence of a floppy drive :) | 14:35.30 |
| tor8 | Robin_Watts: fz_warn(ctx, "cannot cache glyph; continuing") is more in line with our other warnings in terms of language | 14:57.48 |
| Robin_Watts | ok. | 14:58.27 |
| tor8 | question is, do we want to swallow other errors as well? | 14:58.47 |
| Robin_Watts | tor8: We should only swallow errors that occur during the caching process, IMHO. | 15:00.02 |
| tor8 | if fz_hash_insert returns an old pixmap we shouldn't increase the cache total | 15:01.20 |
| Robin_Watts | New thigns on robin/master then. brb. | 15:03.15 |
| tor8 | Robin_Watts: looks good. | 15:03.54 |
| Robin_Watts | There is one there to make display lists refcounted. | 15:06.50 |
| | I *could* hide fz_free_display_list, but that would break all our existing callers code. | 15:07.27 |
| tor8 | Robin_Watts: yeah. sticking to free for non-refcounted objects was a mistake. | 15:09.01 |
| | all in favour of a global s/free/drop/ pass raise their hands! | 15:09.57 |
| Robin_Watts | not convinced. | 15:10.18 |
| tor8 | well, if the api changes to refcounted, it'll have to break backwards compatibility. | 15:12.50 |
| | fz_free_display_list should be hidden, and break existing callers, IMO | 15:14.12 |
| Robin_Watts | or8: ok. | 15:14.21 |
| tor8 | if we add caching of display lists somewhere, then existing callers could break things very much by calling the free function directly | 15:14.45 |
| tor8 | dinners. | 15:15.57 |
| mvrhel_laptop | oh so fz_free_display_list will go away and I will do a ref count inc or dec? | 15:16.04 |
| tor8 | mvrhel_laptop: git.ghostscript.com/?p=mupdf.git;a=blob_plain;f=doc/refcount.txt | 15:19.20 |
| | http://git.ghostscript.com/?p=mupdf.git;a=blob_plain;f=doc/refcount.txt | 15:19.28 |
| Robin_Watts | mvrhel_laptop: You'll call fz_drop_display_list instead. | 15:20.32 |
| mvrhel_laptop | ok. | 15:21.49 |
| Robin_Watts | and you can take extra refs by calling fz_keep_display_list | 15:22.26 |
| mvrhel_laptop | ah ok | 15:23.06 |
| | the drop will decrement and the keep will increment. | 15:23.45 |
| Robin_Watts | yeah. We already have loads of stuff that works using fz_keep_... and fz_drop_... | 15:24.18 |
| | This just moves display_lists into that class of things. | 15:24.30 |
| mvrhel_laptop | Robin_Watts: sounds good. how long until this is commited? | 15:26.20 |
| Robin_Watts | mvrhel_laptop: depends how long until tor8 finishes his dinner :) | 15:26.43 |
| henrys | chrisl: for the logs LDFLAGS="-L/usr/lib/" ./autogen.sh --with-drivers=HP fails with Undefined symbols for architecture x86_64: | 15:33.37 |
| | "_gdev_prn_set_procs_planar", referenced from: | 15:33.38 |
| | _setup_device_and_mem_for_thread in gxclthrd.o | 15:33.39 |
| Robin_Watts | tor8: ref counting commit on robin/master good to go - passes cluster tests. | 15:45.52 |
| tor8 | Robin_Watts: LGTM. | 16:17.16 |
| Robin_Watts | tor8: 2 more small patches. | 16:18.25 |
| | Did you have ones for me to look at? | 16:18.32 |
| | tor8: sorry, I hadn't pushed them when I said that. I have now. | 16:19.53 |
| | mvrhel_laptop: reference counting is in golden now. | 16:20.04 |
| tor8 | Robin_Watts: 2 small ones LGTM | 16:22.18 |
| | I have a build fix on tor/master | 16:22.36 |
| Robin_Watts | Does msys enter wmain too ? | 16:22.37 |
| | I'm confused. | 16:23.42 |
| | In mudraw we currently have a function that's always called with utf8 arguments. On linux that's called main, on windows, it's called main_utf8. | 16:24.30 |
| | You've changed it so it's always called main. | 16:24.54 |
| | But you have some hacky stuff in fitz.h to #define main to be main_utf8. | 16:25.39 |
| | I don't see how that can ever be safe. | 16:25.48 |
| | suppose I'm writing my own app, and I include fitz.h - suddenly my main function gets renamed. That can't be right. | 16:26.18 |
| | Likewise, I suspect the fopen hack that's there at the moment (which is probably my fault!) needs to be fixed. | 16:27.11 |
| | We should use fz_fopen throughout the code and redefine that as required. | 16:27.33 |
| | And we should never have wmain defined anywhere in our libs. | 16:28.43 |
| | wmain is an entrypoint function - we have no business having that in our libs. In our apps, yes, but not in our libs. | 16:29.10 |
| tor8 | Robin_Watts: yeah, that's unsafe. same as we redefine fopen to fopen_utf8... | 16:44.09 |
| | wmain isn't called by mingw | 16:44.34 |
| Robin_Watts | tor8: Then why move it? | 16:45.11 |
| tor8 | duplicated code... | 16:45.20 |
| | Robin_Watts: we could put the main entry point wrapper to a source file the binaries both depend on | 16:45.48 |
| Robin_Watts | Ok, so move it to a new (app level) file. It shouldn't go into the libs. | 16:46.01 |
| tor8 | and fix the fopen issue | 16:46.16 |
| | one thought I had was having the macros for fopen and main be predicated on "our code" or "user code" | 16:46.48 |
| Robin_Watts | yeah. I think we should change all our code to call fz_fopen, and then #define that to fopen or fopen_utf8 as required. | 16:47.07 |
| | tor8: I would dislike it if fitz.h did different things if it's invoked with different predefines. | 16:47.57 |
| | (I mean, clearly it already does with things like WIN32 etc, but those shouldn't be changed by the user) | 16:48.18 |
| tor8 | yeah. | 16:48.22 |
| | one general theme I ran into was our #ifdef _WIN32 for MSVC specific stuff | 16:48.42 |
| | I think (but correct me if I'm wrong) that #ifdef _MSC_VER is a safer check for MSVC oddities | 16:49.08 |
| Robin_Watts | tor8: _MSC_VER checks for MSVC, where WIN32 checks for windows. It's possible I've used the wrong one at times. | 16:49.43 |
| tor8 | _WIN32, WIN32 is not reliably defined | 16:50.07 |
| | I also meant to ask what the ifdef WIN32 in the scaling debug code does | 16:50.32 |
| Robin_Watts | tor8: It calls OutputDebugString | 16:51.19 |
| | so we see the output in the debugger. | 16:51.46 |
| tor8 | the debugger doesn't show stdout and stderr? | 16:51.59 |
| Robin_Watts | god no. That would be far too sane. | 16:52.12 |
| | vtorri suggested we offer a way of redirecting stdout/stderr yesterday. It's a good idea, IMHO. | 16:52.47 |
| tor8 | ifndef NDEBUG, define fz_debug_printf function that sends to stderr on unix and OutputDebugString on msvc? | 16:53.16 |
| Robin_Watts | tor8: A set of functions in the context. | 16:53.32 |
| tor8 | yeah, redirecting fz_warn and fz_error messages through function pointers in the context sound like a good idea | 16:53.38 |
| | and we can add fz_debug to the set | 16:53.49 |
| Robin_Watts | right. | 16:53.54 |
| | I'll look at that tomorrow. | 16:54.21 |
| tor8 | and I'll fix the mingw build main/wmain stuff tomorrow | 16:54.40 |
| Robin_Watts | paulgardiner: Could you cast your eye over bug 694104? | 16:57.18 |
| | I think it's reasonable. | 17:00.27 |
| | http://developer.android.com/reference/android/view/View.html#isInEditMode() | 17:00.40 |
| vtorri | hey | 17:01.46 |
| ghostbot | hello | 17:01.46 |
| vtorri | who is the maintainer of the ghostscript build system ? | 17:02.03 |
| Robin_Watts | vtorri: We took on your typo fixes, and I'm going to look at stderr/stdout redirection tomorrow. | 17:02.10 |
| | vtorri: chrils. | 17:02.15 |
| | or chrisl. | 17:02.23 |
| | one of the two. | 17:02.32 |
| vtorri | about redirection, just function pointers in error_context + default values for warn and err, + API to set those function pointers | 17:03.12 |
| | Robin_Watts, that's what you'll do ? | 17:03.20 |
| | ok for the maintainer, thanks | 17:03.28 |
| Robin_Watts | yes. | 17:03.28 |
| vtorri | perfect | 17:03.33 |
| | it will solve my problem | 17:03.44 |
| | Robin_Watts, do you plan to add colors ? | 17:03.56 |
| | like, for errors, "error" will be in red | 17:04.07 |
| Robin_Watts | No. | 17:04.14 |
| vtorri | and "warning" in green or yellow :) | 17:04.21 |
| | hehe | 17:04.25 |
| Robin_Watts | Definately not. | 17:04.30 |
| paulgardiner | Robin_Watts: Not something I knew anything about, but as you say, patch looks sensible enough. | 17:04.50 |
| vtorri | just a thought | 17:04.51 |
| | anyway, our log system uses colors, so i do not care, actually | 17:05.08 |
| Robin_Watts | paulgardiner: And there is still the "smooth" annotation on robin/master | 17:05.11 |
| vtorri | Robin_Watts, will mupdf 1.3 be released soon ? | 17:26.32 |
| Robin_Watts | vtorri: August | 17:26.39 |
| vtorri | that is, in the next 2 months ? | 17:26.40 |
| | ok | 17:26.42 |
| | thanks | 17:26.44 |
| mvrhel_laptop | Robin_Watts: ok great. I will update and see if I can get my display list cache working | 17:48.57 |
| | I need to work on my expense reports though. have not done one since december.... | 17:49.43 |
| Robin_Watts | mvrhel_laptop: hehe. I have a hard deadline every 3 months where I need to do company VAT returns, so stuff gets done then. | 17:50.40 |
| mvrhel_laptop | I just hope I can remember everything... | 17:51.09 |
| | and find all the receipts | 17:51.25 |
| Robin_Watts | If you can't remember it, it clearly isn't enough to be worth claiming :) | 17:51.28 |
| mvrhel_laptop | My memory can be pretty bad, but that is probably true | 17:51.43 |
| Robin_Watts | ugh. A bug in the interpolated landscape bitmap code :( | 17:54.25 |
| ray_laptop | mvrhel_laptop: I got a lot of segfaults from my patch -- I wasn't using get_profile :-( An easy fix. Re-testing now | 17:54.37 |
| mvrhel_laptop | ray_laptop: oh you were just grabbing the device profile directly? | 17:55.02 |
| | Sorry I missed that when I looked at your code | 17:55.14 |
| ray_laptop | mvrhel_laptop: yep. In gs_output_page | 17:55.39 |
| | I'm going to look it over again just to make sure. | 17:55.58 |
| mvrhel_laptop | so I lost my wireless mouse on the plane trip back home. either left it in an airport or on a plane :( | 17:56.18 |
| ray_laptop | mvrhel_laptop: I'm not sure why that's needed. The icc_struct is a member of the 'common' device struct | 17:56.48 |
| Robin_Watts | mvrhel_laptop: Ah. kens left his headphones in the hotel. | 17:57.01 |
| ray_laptop | does it move around for some reason ? | 17:57.02 |
| | Robin_Watts: I hope they weren't the expensive noise cancelling type. | 17:57.25 |
| mvrhel_laptop | ray_laptop: you are not sure why what is needed? the procedure? | 17:57.31 |
| ray_laptop | Robin_Watts: The hotel may have them | 17:57.33 |
| Robin_Watts | ray_laptop: Yes, they were exactly the expensive noise cancelling type. | 17:57.41 |
| ray_laptop | mvrhel_laptop: right. Why do we need to access via the proc | 17:57.55 |
| mvrhel_laptop | many devices don't have a profile but forward to their target device | 17:57.57 |
| ray_laptop | mvrhel_laptop: Oh. Thanks | 17:58.11 |
| Robin_Watts | ray_laptop: He called the hotel and told them, but couldn't hang around for them to give him an answer (partly cos we were boarding, and partly cos it was costing 1.20UKP a minute :) ) | 17:58.32 |
| | hopefully he's been in touch since he got back home. | 17:58.42 |
| mvrhel_laptop | ok. I really want to work on my code but I am going to be disciplined and get these expense reports done | 17:59.24 |
| ray_laptop | mvrhel_laptop: yeah. My wife's been reminding me, too. Much more fun to code :-/ | 18:00.55 |
| | I think my UPS is dying that I have peeves on. Every few minutes it starts beeping and red lights come on. | 18:02.17 |
| mvrhel_laptop | maybe it is backing up | 18:02.47 |
| | then it would be white lights though | 18:03.01 |
| ray_laptop | (good thing the english teacher at my kids' school didn't see that last sentence) | 18:03.25 |
| Robin_Watts | This UPS is reversing. This UPS is reversing. | 18:03.36 |
| ray_laptop | it is a backup power supply according to the ads | 18:05.25 |
| Robin_Watts | I'm sure that at some point I understood the PatchWidth calculations in the interpolate code. | 18:05.47 |
| ray_laptop | sympathizes | 18:06.17 |
| Robin_Watts | ha. blame blames me. so I looked for the second most recent person to change it. And it was me too. | 18:08.25 |
| ray_laptop | Robin_Watts: It just dawned on me. Helen is your wife. Was she at AnDevCon using the email address hwatts@artifex.com ??? | 18:09.04 |
| Robin_Watts | ray_laptop: She was at andevcon. She wasn't using an artifex address though. | 18:09.29 |
| | why? | 18:09.39 |
| ray_laptop | somehow some emails are coming through with "Hi Helen, We met at AnDevCon ..." to that address | 18:10.07 |
| Robin_Watts | She registered using @wss.co.uk, I think. | 18:10.16 |
| | ray_laptop: Crumbs. Any worth having ? | 18:10.28 |
| ray_laptop | I'll forward both to you (and any future ones) | 18:10.54 |
| | It took a while for it to sink in past the mental fog that Helen and hwatts was thngs for you. | 18:11.31 |
| | Robin_Watts: OK. Forwarded them. | 18:14.21 |
| Robin_Watts | ta. | 18:14.26 |
| ray_laptop | mvrhel_laptop: that's better. Now just 4 segfaults. (or old friend Bug693711.pdf is 3 of them). Who has that issue ??? | 18:16.24 |
| Robin_Watts | Do we ever cluster test with interpolation enabled? | 18:17.08 |
| ray_laptop | Robin_Watts: if you mean with -dDOINTERPOLATE, then I don't think so. Some of our files have /Interpolate true, however | 18:23.13 |
| | on some or many of the images | 18:23.31 |
| | Robin_Watts: we do test with -dTextAlphaBits=4 -dGraphicsAlphaBits=4 | 18:23.57 |
| | Robin_Watts: and I've heard that mupdf does some kind of attempt at image interpolation ;-) | 18:24.28 |
| Robin_Watts | ray_laptop: Yeah, the problem is not in the interpolate code itself, it's in the setup code. | 18:24.51 |
| | We're miscalculating the size/position of the image. | 18:25.15 |
| ray_laptop | Robin_Watts: that's not good :- | 18:25.26 |
| | :-( | 18:25.30 |
| Robin_Watts | And the only person to have changed that calculation since 1998 is... me. | 18:25.59 |
| | but even then I think the height calculation is bad. | 18:26.17 |
| ray_laptop | My twins get out early today. Last day of school for them :-) Hurray !!! My wife's home with the 7yr old, so I have to pick them up. bbiaw... | 18:27.47 |
| ray_laptop | ducks out before Robin_Watts can ask me to look at his calculations ;-) | 18:28.08 |
| SpNg | is it possible to convert .eps RGB objects to CMYK colorspace? | 19:32.50 |
| Brando753 | Hey guys I am using rghost which is the ruby ghost script library, I have 150 PNG images and I want to convert them into a single multipage pdf, how could I do this? | 23:33.03 |
| | using rghost and not RMagick | 23:33.21 |
| | well if anyone will answer just flag me as I am going to bed and I | 23:39.59 |
| | will see what you said in the morning | 23:40.00 |
| | Forward 1 day (to 2013/06/14)>>> | |
IRC Logs