| <<<Back 1 day (to 2013/04/14) | 2013/04/15 |
curoi | Hi everyone. I'm new to Ghostscript so I was wondering if there is a way I can download and install it on my Mac so I can use it in conjunction with Matlab. | 07:21.37 |
chrisl | curoi: Sorry, but I've no idea how Matlab uses Ghostscript, or what the prerequisites might be...... We don't supply a Mac binary. | 07:28.03 |
kens | Ghostscript is available for download, either as source or as pre-compiled binaries for certain operating systems | 07:28.50 |
chrisl | curoi: FWIW, I think you can get Ghostscript binaries from the macports project | 07:29.20 |
kens | http://www.ghostscript.com/download/ | 07:29.21 |
curoi | It's just some user submitted Matlab .m file called export_fig requires I have Ghostscipt in order to save some figures to a vector format like .pdf | 07:31.09 |
| I found something from: http://pages.uoregon.edu/koch/ | 07:31.37 |
chrisl | curoi: from what I can tell, that should work - it seems matlab wants the ghostscript executable in either /usr/bin or /usr/local/bin | 07:43.24 |
| See: http://www.mathworks.co.uk/matlabcentral/fileexchange/23629-exportfig/content/ghostscript.m | 07:43.40 |
curoi | chris1: Great, thanks! | 07:50.05 |
kens | OK flight booked, 57D out 61D back (so I can kick Robin_Watts's seat in both directions ;-) | 11:39.23 |
Robin_Watts | I shall eat beans the night before :) | 11:40.05 |
kens | Gas warfare ? That's banned you know.... | 11:40.40 |
chrisl_r61 | Is there a way to see the available seats before you actually book a ticket?? | 12:07.54 |
kens | I don't think so no | 12:08.07 |
| but there's quite a few free on the way out, not so many coming back | 12:08.24 |
chrisl_r61 | Hmm, that's a bit crap - I'd prefer an aisle seat in economy than a blocked-in seat in PE, but if I can't check.... otherwise, I'd spend some miles. | 12:10.10 |
kens | You cna always use the phone | 12:10.30 |
chrisl_r61 | I could - maybe next time | 12:11.52 |
Robin_Watts | PE was too expensive. I looked on saturday. | 12:16.25 |
chrisl_r61 | Robin_Watts: I was going to do it entirely with miles. | 12:16.57 |
Robin_Watts | Ah. | 12:17.37 |
chrisl_r61 | (given my complete lack of success using the air miles for holidays!) | 12:17.45 |
| But I was not keen to book it and find I couldn't get an aisle seat | 12:18.41 |
Robin_Watts | IME you're not even guaranteed if you book on the phone. | 12:19.02 |
| I've upgraded before on the promise of an aisle seat, and by the time it had all gone through, they'd not been able to get me the aisle seat they thought was free. | 12:19.46 |
chrisl_r61 | Still, it's probably better to save the miles for the next SF meeting | 12:20.08 |
Robin_Watts | So they "put a note on another seat", but I then found I couldn't check in online. When I got to the airport, they said that they needed to process the payment then, and that no note had been left. | 12:20.33 |
| In the end I got an aisle seat, but it was more by luck than by anything the phone people did. | 12:20.58 |
chrisl_r61 | That's pretty normal customer service crapulance - at least, so I've found...... | 12:21.47 |
Robin_Watts | Yeah, but virgin are supposed to be good. | 12:22.20 |
| Going to Boston, we have to fly BA as flight times back are better for Helen. | 12:22.45 |
| BA want 28 quid each to prebook seats. | 12:22.54 |
chrisl_r61 | Ugh, that's just a con :-( | 12:23.14 |
| I was rather hoping we'd have done the meeting in Boston - I'd like to go there | 12:25.57 |
Robin_Watts | yeah, I was kinda hoping that when I volunteered to do the show :) | 12:27.00 |
chrisl_r61 | Ah well, maybe some other time | 12:27.22 |
kens2 | Good grief, gs_client_color->pattern is cast to (gs_pattern1_instance)pinst. gs_client_color->pattern->pattern_id is not equal to pinst->id..... | 12:34.25 |
| We are actually maintaining 2 separate ids <boggles> | 12:34.46 |
chrisl_r61 | Neither of which can be relied upon to indicate a pattern space is unique :-( | 12:35.17 |
kens2 | Looks like i'mgoing to have to do a little morehackery in teh PCL interpreter than I wanted | 12:36.01 |
chrisl_r61 | Like fixing it's use of color spaces? | 12:36.32 |
kens2 | changing rasterops or transparency seems to be changing the pattern ID, so we don't recognise it as the already loaded pattern | 12:36.40 |
| THIs is for a fill followed by a stroke, I'm assuming the same pattern is used, I may be wrong | 12:37.08 |
chrisl_r61 | Hmm, strictly speaking it may not be the same pattern if the rop changes..... | 12:37.26 |
kens2 | For high level patterns I htink it is, but what do I know ? | 12:37.49 |
chrisl_r61 | It may be for high level patterns - and, of course, pdfwrite ignores rops, so.... | 12:38.33 |
kens2 | high level patterns were only written for PS/PDF input.... | 12:38.57 |
| Its not a huge deal, it just means doing stuff in a place I was hoping to avoid. | 12:39.16 |
chrisl_r61 | Bigger mallet required | 12:39.48 |
Robin_Watts | The Drawer of Jeremy | 12:40.45 |
kens2 | I had managed to work in the remap_color stuff at the interpreter level, which is better than I did for XPS, but if some routien uses *2* graphics library operations, it doesn't work | 12:40.49 |
| in this case fill and stroke | 12:41.03 |
| At least, it doesn't work if the pattern ID changes between the 2 | 12:41.22 |
Robin_Watts | (The most recent series of Top Gear had them modifying a car, and one drawer was labelled "The Drawer of Jeremy". When opened it was entirely full of hammers.) | 12:41.28 |
kens2 | Yes, I saw it, seemed like an appropriate label ;-) | 12:41.57 |
chrisl_r61 | My mind boggles at someone who know *so* little about cars actually reviewing them :-( | 12:42.21 |
kens2 | Ah, the PCL interpreter calls gs_setpattern twice, so they *are* different patterns | 12:42.39 |
Robin_Watts | http://blogs.msdn.com/b/eternalcoding/archive/2013/04/15/reading-pdf-and-xps-on-your-windows-8-application-using-winrt.aspx | 12:47.23 |
kens2 | BTW chrisl, PXL has a pen and a brush, (for fill and stroke) and tehy can be different. In this case it seems they are... | 12:50.59 |
| Robin_Watts : that's nice, presumably Michael will be pleased | 12:53.12 |
kens2 | hopes that was Robin aborting the run, not my test causing it | 12:53.56 |
Robin_Watts | I didn't abort the run. | 12:54.23 |
kens2 | lots of abort messages | 12:54.37 |
| also some of the AWS nodes appeared to be running, I didn't think that should have happened | 12:54.58 |
Robin_Watts | The mupdf run seemed to complete OK. It's the bmpcmp run having problems. | 12:54.59 |
kens2 | Yes it was the bmpcmp | 12:55.07 |
Robin_Watts | I think marcos may be active. | 12:55.16 |
kens2 | and lurking :-) | 12:55.27 |
| Not admitting here that he's online | 12:55.40 |
Robin_Watts | but I suspect the bmpcmp will rerun I think. | 12:55.40 |
| tor8: ping | 13:00.50 |
| tor8: Softmask review on robin/master is good to go now. | 13:01.35 |
tor8 | Robin_Watts: looks reasonable | 13:06.09 |
| so, miami again? | 13:06.13 |
Robin_Watts | tor8: indeed. | 13:06.20 |
tor8 | you know why? | 13:06.28 |
kens2 | Miles wants his fishing trip | 13:06.34 |
tor8 | ah. | 13:06.38 |
kens2 | Just guesing | 13:06.46 |
chrisl_r61 | Whose turn is it to bring the poison for Scott? ;-) | 13:07.05 |
tor8 | cause it's barely more convenient than other locations, the time to get to miami is just an hour short of sfo iirc | 13:07.07 |
| what times have you booked? | 13:07.22 |
kens2 | tor8 quite true yes | 13:07.23 |
tor8 | the jet lag is easier though :) | 13:07.39 |
kens2 | We land at 1bout 16:50 and return at 9pm | 13:07.42 |
chrisl_r61 | Also, Miami international is a very depressing place - even for an airport! | 13:07.56 |
kens2 | 16:05 apparently | 13:08.07 |
tor8 | 16:50 on june 7th and depart 9pm the 9th? | 13:08.42 |
kens2 | 16:05 on the 7th 21:00 on the 9th | 13:08.58 |
tor8 | kens2: thanks. now to see what I can get... | 13:09.25 |
| arrival 15:20, depart at 17:25 | 13:27.53 |
Robin_Watts | tor8: So with luck you should clear immigration as we join the queue. | 14:01.51 |
| How did you all get to the hotel last time? | 14:02.02 |
kens2 | shuttle bus | 14:02.55 |
Robin_Watts | ah, I couldn't see a mention of a shuttle bus on the website. | 14:03.15 |
kens2 | henrys ping | 14:38.05 |
henrys | kens2:happy monday | 14:38.24 |
kens2 | :-( | 14:38.30 |
| Is there any limit on the complexity of a dash fors troking in PXL ? | 14:38.45 |
| for stroking* | 14:38.50 |
| THat's a dash pattern of course | 14:39.02 |
henrys | kens2:hang on I don't recall off hand | 14:40.28 |
kens2 | Actually I think there's an artificial limit going on in pdfwrite, its limiting us to 11 entries (I'm guessing this is because the clist has the same limit). I see no reason why this should be the case in PostScript. | 14:41.20 |
| So I think I'll rewrite pdfwrite. | 14:41.50 |
henrys | XL is limited to 20 segments | 14:42.42 |
kens2 | I'll see if its easy to fix pdfwite, if not I'll push the limit up to 20, thanks henrys | 14:43.15 |
| OK restriction lifted, didn't even know it was there before so another useful fix. | 14:53.02 |
henrys | hmm this address sanitizer business looks like good fun :-^ | 14:55.40 |
kens2 | The fuzzing thing ? | 14:55.55 |
henrys | kens2:yes | 14:56.04 |
kens2 | I think some of them will be the same a shte valgrind reports | 14:56.08 |
henrys | kens2:let's hope | 14:56.49 |
Robin_Watts | read that as "the same shite valgrind reports" :) | 15:00.38 |
kens2 | Oops, yes, my bad spelling | 15:01.15 |
sebras | kens2: keep it up! I liked this one. :) | 15:01.42 |
kens2 | :) | 15:02.10 |
henrys | kens2, tor8:it is 11 hours to miami and a bit more than 8 to Miami (from London), I have to think that is an appreciable difference and preferred. | 15:30.54 |
kens2 | 9.5 hours Miami, 10.5 SFO | 15:31.14 |
henrys | new york would be ideal and fairly equal torture for all. | 15:31.36 |
| kens2:oh google lied. | 15:31.54 |
kens2 | Well, that's the times quoted by Virgin, and they seem to be mostly correct on past experience | 15:32.18 |
Robin_Watts | henrys: By the time you've done the whole "get to international airport/security/immigration" dance the extra hour on the flight is of little consequence. | 15:33.17 |
| though, obviously, we appreciate the thought :) | 15:33.48 |
henrys | oh yes I thought it was more than an hour, I was mistaken. | 15:33.49 |
| Robin_Watts: I had nothing to do with the miami trip this time, likely a fishing trip is the motivation. | 15:34.46 |
Robin_Watts | henrys: Yeah, I just spoke to Scott on skype. I said I'd call the hotel and ask for him to be given the same room as "he had formed a special bond with it last time". | 15:35.27 |
henrys | we'll have the restaurant's pork dish sent to his room. | 15:41.23 |
ray_laptop | henrys: and the ceviche, just in case ;-) | 15:50.10 |
Robin_Watts | mvrhel_laptop: Did you see the article? Looks good. | 15:58.05 |
mvrhel_laptop | Robin_Watts: oh cool. | 16:00.41 |
| I need to restructure my project so that the calls into mupdf are truly winRT to allow the calling from c# , javascript etc | 16:02.11 |
| and then throw in some sample projects in the other languages | 16:02.27 |
| but I need to stop now and do some work for #801 | 16:03.46 |
ray_laptop | Robin_Watts: what article ? | 16:04.06 |
Robin_Watts | http://blogs.msdn.com/b/eternalcoding/archive/2013/04/15/reading-pdf-and-xps-on-your-windows-8-application-using-winrt.aspx | 16:04.15 |
ray_laptop | Robin_Watts: tA | 16:04.21 |
mvrhel_laptop | ok onto bug 693583 | 16:08.19 |
| I guess I better find some flights to miami first. | 16:09.59 |
Robin_Watts | tor8: hey | 16:13.43 |
| tor8: On robin/master there are a couple of commits to speed mesh rendering. You may have looked at an earlier version of the first one before. | 16:14.20 |
| The second one was an attempt to avoid copying vertex data when we didn't need to, but my tests don't show it as saving anything. | 16:14.55 |
| So I'm tempted to drop the second one - but thought it was worth showing it to you first, in case you had any opinions. | 16:15.18 |
ray_laptop | Robin_Watts: we _really_ need to get the mupdf shading rendering over into gs. IIRC, mupdf's approach is about 4 times faster than gs's mess. | 16:16.11 |
| I wonder how hard that will be ? | 16:17.05 |
Robin_Watts | ray_laptop: Interesting question. | 16:17.30 |
ray_laptop | I suppose we should benchmark it again just to see if it's still the case. | 16:18.23 |
Robin_Watts | MuPDF decomposes to triangles, and then linearly fills scanlines within those triangles. | 16:18.41 |
ray_laptop | Robin_Watts: that's pretty much what gs wants. We have a device proc "fill_linear_color_triangle" | 16:25.30 |
Robin_Watts | gs decomposes to traps I thought. | 16:25.39 |
ray_laptop | (also fill_linear_color_trapezoid and fill_linear_color_scanline) | 16:25.58 |
| Robin_Watts: triangle or trapezoid is fine. Hardware ASIC's are more common for triangles, however | 16:27.04 |
| Cust 532 has a trapezoid fill, but not yet linear color variant (just solid color) | 16:27.41 |
Robin_Watts | ray_laptop: MuPDFs code treats everything as n color, and linearly interpolates as it decomposes. | 16:28.21 |
ray_laptop | I think their ASIC roadmap calls for linear color triangle next | 16:28.40 |
Robin_Watts | There is a question of where we'd put color management; do we colour manage at the top level (and then linearly interpolate the 'device' colors as we decompose?) or do we colour manage at the bottom (and do colour management on the colors after decomposition). Or something in between? | 16:29.36 |
| The former is likely to be too coarse in some cases, the latter too slow. | 16:33.05 |
ray_laptop | Robin_Watts: mvrhel_laptop would have to answer that, but I think the colors of the vertices are converted to device colorspace | 16:33.18 |
mvrhel_laptop | yes, iirc the vertices are in device color space. | 16:35.10 |
| flights from sea to miami really stink | 16:35.25 |
Robin_Watts | from sea to shining sea. | 16:36.02 |
mvrhel_laptop | henrys: so did we decide that I was going to go to the open printing summit? | 16:49.44 |
henrys | mvrhel_laptop: unfortunately that is what we came up with. I am okay with just marcos going to I understand that is non optimal but so is excessive travel. | 16:51.32 |
mvrhel_laptop | henrys: I am fine with going. I enjoy the meetings and giving the talk | 16:52.10 |
henrys | mvrhel_laptop: okay great | 16:52.26 |
mvrhel_laptop | looks like I am arriving in maimi at 4:45 so perhaps I can have dinner with everyone this time | 16:52.45 |
henrys | my nephew's robot is in the world championships in st louis next week! I'm going to that. | 16:53.46 |
mvrhel_laptop | awesome | 16:54.19 |
| henrys: so we did the tram in palm springs last week. that was pretty cool | 16:54.46 |
henrys | mvrhel_laptop: I saw that on Facebook - amazing isn't it? | 16:55.13 |
mvrhel_laptop | yes. we had a nice day in Joshua tree national park too | 16:55.34 |
Robin_Watts | wants to go to Joshua Tree. | 16:55.55 |
mvrhel_laptop | really nice desert hiking in the area | 16:56.26 |
| I bet Helen would like the palm springs area | 16:56.45 |
| we should have a staff meeting there some time | 16:56.52 |
Robin_Watts | Vegas, baby! | 16:57.00 |
mvrhel_laptop | Vegas would be fun too | 16:57.10 |
Robin_Watts | I believe they have hotels there, so it's possible :) | 16:57.29 |
mvrhel_laptop | and they are cheap | 16:57.36 |
Robin_Watts | If Miles gambles enough, he can get the rooms for free. | 16:58.26 |
mvrhel_laptop | exactly :) | 16:58.38 |
henrys | I'm ready for europe again | 17:13.25 |
kens2 | Lund ? :-) | 17:13.34 |
Robin_Watts | Copehagen. | 17:13.41 |
| Copenhagen or Lund. | 17:13.51 |
kens2 | Copenhagen would be good, I have a cousin there | 17:13.57 |
| OK off for dinner, night all | 17:22.55 |
chrisl | Robin_Watts: ping | 18:16.21 |
Robin_Watts | pong | 18:19.07 |
chrisl | The cluster seems to have barfed on my commit - do you know how to check if it really was my commit, or whether it's a cluster issue? | 18:19.37 |
Robin_Watts | chrisl: Did you test it before you pushed? | 18:21.03 |
| I think just wait to see whether it dies or not. | 18:21.44 |
| It'll rerun if there is a problem. | 18:21.55 |
chrisl | I think I did cluster test it, but now I'm wondering..... | 18:22.03 |
Robin_Watts | I think if what you've changed was going to fail, it would have failed on all nodes. | 18:22.21 |
chrisl | I think it has already rerun, and it's down to three nodes - could take a while :-( | 18:22.34 |
Robin_Watts | Let me text marcosw. | 18:22.48 |
chrisl | Robin_Watts: I just confirmed I definitely cluster tested the change - at least three times (as a side effect of testing other changes), so it really shouldn't be my fault! | 18:26.38 |
Robin_Watts | It doesn't look like your fault. | 18:27.01 |
| I should find out from Marcosw how to abort/restart a job. | 18:27.20 |
mvrhel_laptop | whats a good name for a device parameter that would invoke the monitoring of pages for gray scale only content? | 18:27.29 |
Robin_Watts | Possibly I could just kill the clustermaster.pl process. | 18:27.39 |
| mvrhel_laptop: -dColorDetection=1 ? | 18:28.07 |
| or maybe -dGreyDetection=1 ? | 18:28.25 |
chrisl | It's such a long time since I looked at the cluster code - thank heavens! | 18:28.27 |
mvrhel_laptop | hmm Gray vs. Grey | 18:29.24 |
| at least you did not say ColourDetection | 18:30.07 |
Robin_Watts | Gray, cos the rest of gs is wrong already :) | 18:30.14 |
| float blah[2][4][8] | 18:32.04 |
| c = blah[2]; | 18:32.10 |
| what type is c ? | 18:32.17 |
| c = blah[0] even. | 18:32.25 |
mvrhel_laptop | I hate multi-dim arrays in | 18:33.14 |
| c | 18:33.16 |
Robin_Watts | me too. | 18:33.21 |
chrisl | float **, I think | 18:33.35 |
mvrhel_laptop | that would be my first guess too | 18:33.52 |
Robin_Watts | no, cos then you'd expect *c to be a pointer. | 18:35.32 |
| and if I take c = blah[1]; there is no pointer stored anywhere in memory that actually points there. | 18:36.01 |
chrisl | I'd expect *c to be a pointer to a pointer | 18:36.20 |
Robin_Watts | It's something like: float (*c)[8] I think. | 18:36.48 |
chrisl | Right, that's a pointer to a pointer | 18:37.24 |
Robin_Watts | chrisl: No, it's just a pointer. | 18:54.44 |
chrisl | *c points to an array of length 8? | 18:55.39 |
Robin_Watts | yeah. | 18:55.54 |
chrisl | So c is a pointer to pointer to a float | 18:56.03 |
Robin_Watts | No, because if it was a pointer to a pointer, you'd expect to be able to write to *c and change a bit of storage somewhere. | 18:56.41 |
chrisl | Robin_Watts: but **c is a float | 18:57.02 |
Robin_Watts | &blah[0][0][0] = 0x100000 | 18:57.09 |
| blah[0] = 0x100000 | 18:57.18 |
| c = blah[0] = 0x100000 | 18:57.32 |
| c[0] = 0x100000 | 18:57.44 |
| c[0][0] = a value. | 18:57.53 |
Robin_Watts | goes for a walk to clear head. | 18:58.24 |
mvrhel_laptop | sounds like a good idea.... | 19:00.24 |
| so this color detection fix should not be too bad with the introduction of a monitoring cmm which will just pass through to the real cmm | 19:01.17 |
Robin_Watts | marcosw: Hi. | 19:28.49 |
marcosw | Robin_Watts: hello | 19:28.57 |
Robin_Watts | Cluster is very slow with only 3 nodes :) | 19:29.04 |
marcosw | surprised it runs at all | 19:29.56 |
Robin_Watts | marcosw: If this ever happens again, is there something I can do to force it to restart the job? | 19:31.17 |
| I did debate about just killing clustermanager.pl and letting the cluster self-heal. | 19:31.33 |
marcosw | yeah, login in as the regression user and "cd cluster ; ./killClustermaster.sh" | 19:32.00 |
Robin_Watts | ok, thanks. | 19:32.21 |
marcosw | but that just causes it to re-run the last job, which will presumably fail in the same way. | 19:32.39 |
Robin_Watts | marcosw: In this case I don't think the last job was to blame. | 19:32.54 |
| Chrisl has tested in (indirectly) 3 times already on the cluster. | 19:33.03 |
marcosw | it looks like the job you run just before failed and left many of the nodes in a stuck state. By the time the recovered casper had decided they were down and so there were only 3 node left. | 19:35.22 |
| Robin_Watts: your mupdf clusterpush job failed on xeon. I | 19:37.19 |
| I'm not sure I understand why, since the xeon node shouldn't be participating in mupdf cluster runs. | 19:37.42 |
| right, it was the bmpcmp following the mupdf job that was the problem. The code that removes macpro and xeon from mupdf cluster runs does not work if the cluster run was a bmpcmp job. I'll fix that. | 19:41.38 |
Robin_Watts | marcosw: Ah | 19:44.48 |
marcosw | sorted that, conveniently you clusterpushed a mupdf job followed by bmpcmp, so that will verify that my fix works. Now to figure out why the mupdf compiled failed on xeon. It couldn't find the x11 libraries, which doesn't make sense since I compile mupdf on that box all the time. | 19:46.32 |
| Robin_Watts: right, apparently I use "make XCFLAGS=-I/opt/X11/include XLIBS=-L/opt/X11/lib" when I build mupdf on xeon, which is where the X11 includes and libraries live on Mac OS X 10.8. | 19:54.29 |
Robin_Watts | marcosw: Could we have each node set an environment variable with the right paths in for itself? | 19:59.28 |
| or several env variables, rather. | 19:59.49 |
marcosw | I think the traditional way of dealing with this is to run ./configure :-) | 20:00.05 |
Robin_Watts | That way the central server doesn't need to know specifics of individual nodes. | 20:00.11 |
| the makefile calls pkgconfig in some cases. | 20:00.28 |
| but pkgconfig doesn't exist on macos. | 20:00.39 |
marcosw | Robin_Watts: I don't have login access to all of the nodes, so setting an env variable would be difiicult. I think not being able to compile mupdf on mac os x without having to pass options on the command line and/or set env variables might be a bug. | 20:02.26 |
Robin_Watts | marcosw: Does X come as standard on macosx ? | 20:02.59 |
marcosw | nope | 20:03.07 |
| presumably why it lives in /opt/X11 | 20:03.59 |
Robin_Watts | henrys: bomb at finish line on Boston Marathon :( | 20:04.07 |
henrys | yeah I have a lot of friends there but I've heard from everybody they all had finished. | 20:05.08 |
marcosw | Robin_Watts: is there any reason to build mupdf for cluster jobs? isn't mudraw all we need? (or does mujstest use mupdf?) | 20:06.03 |
| looks like adding NOX11=1 to the mupdf build is the best answer. | 20:09.32 |
Robin_Watts | marcosw: We don't need to build mupdf itself. or mupdf-js | 20:09.35 |
marcosw | bbiaw | 20:11.24 |
| Forward 1 day (to 2013/04/16)>>> | |