| <<<Back 1 day (to 2014/05/01) | 2014/05/02 |
Guest96917 | hi I need some help on printing PCL file to printer | 07:03.32 |
| Can not print barcode | 07:03.46 |
kens | How does this relate to Ghostscript ? | 07:04.13 |
Guest96917 | I am using pcl6 with command line to convert pcl to pdf | 07:04.39 |
| Is this right mailing list to as the question for ghostpdl which privides pcl6 command line | 07:05.08 |
kens | OK you need to be aware that PCL can contain 'rasterops' which cannot be represented in the PDF graphics model. Any such rasterops will be ignored, which can lead to incorrect results. | 07:05.33 |
| If you use Ghostpcl to the display device, does your barcode display properly ? Does the barcode display properly when you open the PDF file in a PDF viewer ? | 07:06.30 |
Guest96917 | Even I do not want it to convert pdf, but no success for barcode to print directly on printer with pcl6 command | 07:06.41 |
| no | 07:06.57 |
kens | there were 2 questions, which one is 'no' ? | 07:07.17 |
Guest96917 | pcl6 -dNOPAUSE -sDEVICE=pdfwrite -sOutputFile=/tmp/123.pdf Desktop/tmplpt/25121930.LPT | 07:07.23 |
| both no | 07:07.44 |
kens | Which version of GhostPCL are you using ? | 07:08.00 |
Guest96917 | the command I used to convert pcl to pdf for testing | 07:08.10 |
| ghostscript-8.70-19.el6.i686 | 07:08.33 |
kens | That's very old, the current version is 9.14. I would suggest you upgrade and try again.If the problem persists when rendering to the display device, you should open a bug report. Don't forget to attach teh example file and command line to the report. | 07:09.23 |
Guest96917 | ok will try and check | 07:09.43 |
kens | chrisl do you know any way to get the cluster to report a scan-build without doing a commit ? | 07:24.23 |
chrisl | No, I don't think we can | 07:27.19 |
| kens: ^ | 07:27.24 |
kens | Oh well I guess I'll just have to do a commit and wait then :-( | 07:27.34 |
| Its irritating because I spent quite a bit of time sorting out the cland and gcc compiler warnings, and now I'm faced with a bunch of scan-build warnings I wasn't expecting..... | 07:28.14 |
| s/cland/clang/ | 07:28.23 |
chrisl | kens: IIRC, the reason is that the scan-build thing isn't run as a normal cluster task, it's run on a difference machine.... and may be done overnight, rather than actually with the commit | 07:33.24 |
| kens: I *could* install scan-build here, and run it for you | 07:33.49 |
kens | No don't worry, I think (looking again at the code) I need to rework it anyway. Probably a closer steal of Michael's original code is required. | 07:42.42 |
| chrisl that's an impressively incorrect PDF file for Bug #695194, a /Separation space with a 3 component alternate :-) | 08:00.36 |
| Anyway, the fix looks fine to me, please go ahead and commit it. | 08:01.11 |
chrisl | kens: there's a lot more wrong with it than that! | 08:06.25 |
kens | I saw the impressive list of errors that GS ignores when processing the file | 08:06.42 |
| BUt it does eventually produce the same results as Acrobat with your fix | 08:07.00 |
chrisl | The really impressive thing is that Acrobat throws errors, too..... and the pre-flight shows a raft of errors, as well! | 08:07.21 |
kens | I never even tried the preflight..... | 08:07.32 |
| I saw the Acrobat error though | 08:07.42 |
| You'll have to redo the fix with the modified pdf_main.ps, or merge the diffs I'm afraid | 08:08.08 |
chrisl | There are a load of shading dicts which are "null" | 08:08.12 |
kens | Great..... Kwality product. | 08:08.31 |
chrisl | No conflicts - git merged the fix with your changes without a blink. | 08:10.22 |
kens | That's nice, I wasn't sure it would be happy | 08:10.39 |
chrisl | It's generally pretty good - unless you actually change the same line, or the changes are *so* extreme, it can't identify the line in question | 08:11.32 |
kens | I've had mixed results with it, as you might expect | 08:11.51 |
chrisl | There's no getting away from that, but I've found it makes a much better fist of it than svn used to. | 08:13.27 |
kens | Well that's not too bad | 08:13.37 |
| I think I've got the ICC profile stuff sorted out properly now with the single device parameter (I understand Michael's code there better now , so I just ripped it off verbatim). Hopefully that should shut up scan-build. | 08:14.46 |
kens | coffees | 08:18.44 |
| anyone else having trouble with Bugzilla ? | 09:13.52 |
chrisl | Either not connecting or *really* slow | 09:15.22 |
kens | Well Bugzilla seems to be totally dead for me | 09:17.49 |
chrisl | A bunch of cluster nodes just dropped off, too, so maybe Miles's office has dropped/power cut, or something | 09:18.54 |
| But bugzilla does seem kaput, yes | 09:19.39 |
kens2 | Well my own connection seems to be less than stable today too :-( | 09:22.35 |
| OK Bugzilla is back :-) | 09:55.26 |
norbertj | henrys: did you have a look at the profiling data for the cicero-testfile (that's the only testfile currently not meeting targetperformance: when printing LeadingEdge=3), Others are ok or very close. | 12:27.07 |
kens | its a bit early for henrys yet norbertj | 12:28.31 |
norbertj | kens: offcourse, I forgot. So then it's just for the logs | 12:30.18 |
kens | No problem, I'm sure he'll read it when he comes online, just didn't want to leave you dangling expecting a reply | 12:30.45 |
norbertj | kens: who knows about sse2? | 12:30.57 |
kens | Not me :-) | 12:31.10 |
norbertj | kens: because first glance at the profiling shows that SSE2 using more cycles in 9.14 then in 9.06 (only when printing LeadingEdge=3) | 12:32.13 |
kens | Not my field norbertj sorry | 12:32.29 |
norbertj | kens: no problem.I'm just leaving some crumbs to follow... | 12:32.56 |
kens | :) | 12:33.03 |
sebras | tor8: Robin_Watts: there are a few patches over at sebras/master | 12:53.58 |
| some of them might make sense. some of them probably more illustrates a wish rather than are acceptable solutions. :) | 12:54.30 |
| perhaps this means that http://bugs.ghostscript.com/show_bug.cgi?id=694366 can be closed? | 12:56.55 |
henrys | norbertj: not yet no, what is the slowdown between versions? | 13:21.51 |
| norbertj: sounds like a job for git bisect, let me know if you want us to do it. | 13:27.38 |
Robin_Watts | Woo Hoo! | 13:40.02 |
| I'm debugging a linux binary on a VM with MSVC :) | 13:40.12 |
pedro__ | bluff! :) | 13:40.28 |
kens2 | Hmm very nice | 13:40.37 |
pedro__ | is that using visualgdb? | 13:40.42 |
Robin_Watts | pedro__: Yeah. Using the "QuickDebug" feature. | 13:40.54 |
pedro__ | top stuff | 13:40.59 |
Robin_Watts | It connects using ssh and runs gdb, and sends source files over etc for viewing. | 13:41.09 |
pedro__ | pretty cool | 13:41.20 |
kens2 | Astonishing piece of work | 13:41.25 |
kens2 | is thoroughly impressed | 13:41.34 |
Robin_Watts | It's a tad slow when doing android stuff (but, hey, anything is better than nothing!), but appears seamless on linux. | 13:41.54 |
tor8 | sebras: all except the WIP LGTM. could you take a peek at tor/master as well? | 13:44.40 |
sebras | tor8: sure. | 13:52.32 |
| tor8: https://software.intel.com/sites/products/documentation/doclib/iss/2013/compiler/cpp-lin/GUID-603F2191-6FC2-4EE6-8B94-55C7CAD72D2E.htm ACK. your patch is benign. | 13:53.34 |
kens2 | Hmm sounds like Apple is going to mac OS/X look as naff as iOS: | 13:53.35 |
| http://www.theregister.co.uk/2014/05/02/apple_os_x_10_10_ios_design/ | 13:53.35 |
tor8 | kens2: looks like a bad clone of win8 ;) | 13:54.21 |
| sebras: thanks. I've pushed the lot. | 13:57.20 |
sebras | tor8: opinions on the wip one? | 13:58.23 |
tor8 | HAVE_CURL should not guard the Makethird build stuff, rather the other way around | 13:59.26 |
| if we have thirdparty/curl, HAVE_CURL should auto to 'yes' | 13:59.42 |
| if we have no thirdparty/curl, pkg-config should be used to detect 'yes' | 14:00.00 |
| and then at the end, if HAVE_CURL is set, use it | 14:00.13 |
| make HAVE_CURL=no on the command line will override any assignments made in makefiles | 14:00.28 |
| (hence why we use XCFLAGS for custom CFLAGS settings on the command line) | 14:00.55 |
sebras | tor8: ok, so it is fine to run pkg-config looking for libcurl even if we are not to use it? | 14:10.00 |
tor8 | sebras: look at how the SYS_xxx stuff that already use pkg-config works | 14:10.28 |
sebras | I did, previously, but didn't want to accidentally run pkg-config when it was not supposed to be. | 14:10.56 |
tor8 | like for OpenSSL | 14:10.59 |
sebras | tor8: do you still want to have one mupdf-x11-curl and one mupdf-x11? | 14:11.33 |
| or it is sufficient to have mupdf-x11 and add curl there if we want it? | 14:11.49 |
| s/we want it/it is enabled/ | 14:11.59 |
tor8 | sebras: for openssl, we only run pkg-config if there is no thirdparty/openssl | 14:12.14 |
sebras | tor8: true, but then again you can't disable openssl. | 14:12.41 |
norbertj | henrys: sorry, was going home early. Slowdown was from 9.06 = 27.6 sec to trunk = 36.4 sec | 14:12.59 |
tor8 | I'm not thrilled about the -curl version altogether and would be happy if we could just scrap it. I think it is mostly a demo and test bed for the progressive loading that customers are supposed to implement themselves | 14:13.10 |
sebras | tor8: alright. then I'll fold it into the regular mupdf-x11. | 14:13.34 |
tor8 | sebras: I'm not happy about openssl either... | 14:14.53 |
| it's an awful lot of bloat just to get digital signatures | 14:15.16 |
henrys_ | this cam is about a half mile from my house, incredible birds: http://www.bouldercounty.org/os/openspace/pages/ospreycamera.aspx | 14:15.16 |
| norbertj: command line? | 14:15.50 |
Ziai | hi all | 14:16.03 |
sebras | tor8: mmm, maybe it is feasible to re-implement the (probably) few pieces we need. | 14:16.08 |
| ghostbot: hi | 14:16.14 |
ghostbot | hey | 14:16.14 |
tor8 | sebras: yes, it might be worth looking into one day | 14:16.25 |
Ziai | jghali, hi, could you confirm that this error has nothing todo with GhostScript.NET http://pastebin.com/uEECk18V | 14:16.30 |
norbertj | henrys: nmake -f pcl6_msvc.mak COMPILE_INITS=1 DEBUG=0 TDEBUG=1 DEBUGSYM=1 GENDIR=".\obj_profile" | 14:18.13 |
| obj_profile/pcl6 -sDEVICE=bit -sOutputFile=nul -r600 -dNOPAUSE -dLeadingEdge=3 cicero.pcl | 14:18.38 |
sebras | tor8: it could be worthwhile to look into libressl once the churn is over. | 14:19.12 |
norbertj | henrys: and pl_print_usage(&inst, "Finished :"); added to plmain.c at line 503 | 14:19.34 |
kens | Ziai : I htink you mean jhabjan, no jghali, jhabjan is not here | 14:19.36 |
tor8 | sebras: definitely. maybe even before then, since what we need should be a fairly small subset. | 14:20.12 |
sebras | tor8: IIRC the main problem with openssl is that it doesn't consist of easily separable parts. hopefully they'll address this in libressl. | 14:20.51 |
tor8 | sebras: one other big issue with curl is that it requires pthreads, and we've been threading-library agnostic everywhere else | 14:21.16 |
| sebras: I'm not so sure, maybe a few years down the line | 14:21.40 |
| but if their main task of making the code actually not stink to the high heavens, someone might actually be able to go in and do something useful with it | 14:22.06 |
Ziai | ow. kens, ty u're right | 14:28.58 |
henrys_ | norbertj: and you did a bit compare of each - they are the same? | 14:34.28 |
Robin_Watts | Raising the curl transfer size to 1Meg is a bad idea. | 14:34.55 |
norbertj | henrys: just did a rerun with: 9.06 = 28.7, 907 = 30.5, 9.10 = 30.3, 9.14 = 37.9 secs. | 14:35.52 |
| henrys: bit compare ? you mean comparing the output bitmaps? | 14:36.10 |
| henrys; didn't do that yet. | 14:36.24 |
henrys_ | norbertj: yes I'll do that if you haven't done it already | 14:36.36 |
Robin_Watts | If I'm downloading from a slow server (say one on an ADSL line), I don't want to be waiting multiple seconds for each chunk. | 14:36.53 |
chrisl | Robin_Watts, sebras: could you make the curl transfer size a command line option? | 14:37.58 |
Robin_Watts | chrisl: Not with the current code (it's a #define), but I'm sure it could be done. | 14:38.25 |
chrisl | It's just that would divest you of making the decision - you can blame user :-) | 14:38.50 |
sebras | chrisl: true, but we'd still have to choose a default number. ;) | 14:39.20 |
norbertj | henrys: you have a means to compare the bit-device output ? | 14:40.32 |
henrys_ | diff ;-) | 14:41.50 |
| I assume I can reproduce this with pbmraw | 14:42.16 |
norbertj | henrys: 9.14 = 36.5 sec, 9.06 = 28.8 sec (pbmraw) | 14:44.08 |
henrys_ | norbertj: right comparing bits now | 14:45.28 |
| norbertj: I did notice some "low hanging fruit" in your other files. Increasing the size of the buffer the PCL parser uses might speed things up a bit. That call to gs_alloc_bytes in pcparse.c:444 is too frequent in the 05 file. | 14:50.54 |
| norbertj: I was surprised how high up on the profile "pcl_process" is. That should be something we can fix | 14:51.57 |
chrisl | Crikey, nice small files you get there norbertj!! | 14:54.29 |
henrys_ | chrisl: no kidding I just ran out of disk space doing the bit compare | 14:58.31 |
chrisl | I had the same problem last week! | 14:59.21 |
henrys_ | chrisl: I'd do an md5 but I know that's going to go badly | 15:00.03 |
chrisl | Yes, that wouldn't work too well.... | 15:01.30 |
norbertj | chrisl: I have even bigger ones :) and bitmaps at -r600 is 4.5MB A4 bw | 15:02.35 |
chrisl | norbertj: let's hope you don't find problems only the later pages in these or bigger ones! | 15:04.07 |
henrys_ | norbertj: I'm inclined to wait for michael, he might know off the top of his head why landscape sse would be slower | 15:11.31 |
| marcosw: are you about? | 15:12.03 |
chrisl | One orientation might be slower if we're having to write "up" or "down" (jumping scanlines), rather than across - no idea why we'd see a difference now and not before, though | 15:13.19 |
henrys_ | chrisl: yes, I meant why the regression | 15:14.25 |
norbertj | henrys: _mm_subs_epi8 9.6 = 1.0% 9.14 = 2.4% (both LeadingEdge=3) | 15:16.49 |
| next call from 1.7% to 2.6% | 15:17.26 |
henrys_ | norbertj: I think it's best to wait for michael. I'll change the problem to a p2 customer bug so we don't forget and copy him in. If you can narrow down the time of the regression more that will help him find it more quickly | 15:29.53 |
| mvrhel_laptop: just talking about you | 15:30.31 |
mvrhel_laptop | uh oh | 15:30.38 |
henrys_ | mvrhel_laptop: recall any changes to the sse2 code after 9.06 that would have caused a slowdown. norbertj has found a performance regression | 15:32.34 |
mvrhel_laptop | so the portrait sse can just do the rows of the image directly. due to how the image data is stored the landscape case has to do chunks of rotated data . henrys, but I dont recall any changes that I made to the code other than when I first wrote it. | 15:33.57 |
| I know Robin_Watts and ray_laptop did some fixes though | 15:34.10 |
henrys_ | mvrhel_laptop: okay I'll just bisect it. | 15:34.34 |
tkamppeter | I have found a bug in the build system of GS 9.14 for Linux. | 15:34.54 |
| For the first time a file in devices needs zlib.h, it is devices/gdevfpng.c and on build it errors with "make[3]: *** No rule to make target `src/zlib.h', needed by `soobj/gdevfpng.o'. Stop.". What is wrong here. | 15:36.33 |
| zlib.h included from files in base/ works though. | 15:37.14 |
sebras | tor8: another two patches over at sebras/master | 15:39.07 |
| tor8: I tested to build all permutations of HAVE_X11 and HAVE_CURL with and without thirdparty/curl present | 15:39.33 |
Robin_Watts | mvrhel_laptop: I remember changes being made, but I think it was ray that actually made them, yes. | 15:39.54 |
tor8 | sebras: I don't understand the motivation behind 'Install libmujs alongside libmupdf' | 15:40.09 |
| we treat mujs is just like any other thirdparty library | 15:40.22 |
chrisl | tkamppeter: http://git.ghostscript.com/?p=ghostpdl.git;a=commitdiff;h=0ccf3294 | 15:40.35 |
sebras | tor8: oh!? ok. | 15:40.51 |
| tor8: but at then we should at least add -lmujs to Libs.private, right? | 15:41.36 |
tor8 | I don't know what Libs.private does in the .pc | 15:42.14 |
sebras | tor8: they seem to show up when you ask pkg-config for linkflags for linking statically. | 15:42.44 |
tor8 | right, so they're the full dependency list? | 15:44.20 |
| where shared libraries pull along their own dependencies automatically | 15:44.37 |
sebras | tor8: I guess that's the idea, yes. | 15:44.49 |
| tor8: but I haven't read the manpage, only seen what happens in practice. | 15:45.08 |
| tor8: there. I have reordered the two patches. take the curl one at least, and I'll think more about the libmujs one... | 15:46.00 |
tkamppeter | chrisl, thank you very much. | 15:46.20 |
tor8 | sebras: I think we'll need some facetime before I accept the curl one :) | 15:48.01 |
| and decide on a long term strategy for curl | 15:48.19 |
sebras | tor8: I don't have a mac. | 15:48.51 |
| tor8: and so, no facetime. | 15:49.26 |
ray_laptop | Micro Center has Samsung 250Gb SSD for $129, 500Gb for $249. | 16:09.22 |
jogux_mac | it's rude not to at those prices. | 16:14.32 |
Robin_Watts | That does seem a very good price for a 500Gb one. | 16:15.09 |
jogux_mac | sandisk just announced 4TB SSDs (albeit they're not shipping them yet) | 16:15.41 |
ray_laptop | jogux_mac: I got a 1Tb SSD a couple of months ago. It as $450, iirc | 16:16.31 |
Robin_Watts | Yeah, but they are going to be lots more than 8 times as expensive | 16:16.55 |
ray_laptop | It cut the "full build" for gs to half the time with a 7200 rpm HD, and the heat load is a LOT lower (on my i7 Dell laptop) | 16:17.39 |
jogux_mac | ray : that's good going. | 16:26.41 |
henrys_ | jogux_mac: have you done most of your sot development in Xcode? | 16:32.29 |
| just curious | 16:32.42 |
jogux_mac | henrys : sort of, but mostly because because the main work I've done that's not ATS has been in the ios alien area. | 16:33.07 |
| it doesn't really work that well in xcode tbh. | 16:33.16 |
tkamppeter | chrisl, GS 9.14 is uploaded to Ubuntu Utopic now. | 16:33.37 |
jogux_mac | the best development setup would appear to be robin's vs solution | 16:33.43 |
chrisl | tkamppeter: cool! | 16:33.56 |
tkamppeter | Thanks for all the great work on 9.14! | 16:45.49 |
pihug12 | Hello guys | 16:49.24 |
| I need some help with pdfwrite | 16:49.37 |
| I want to merge two PDFs with the « gswin64c.exe -dBATCH -dNOPAUSE -q -sDEVICE=pdfwrite -sOutputFile=out.pdf file1.pdf file2.pdf » trick | 16:50.04 |
| I can select the text on the two source files | 16:50.25 |
| But I can't on the output one :-( | 16:50.42 |
Robin_Watts | Ok, to be clear, that is not 'merging' 2 pdfs. | 16:50.44 |
| That is 'completely disassembling 2 pdfs and then creating a 3rd one that hopefully looks identical'. | 16:51.05 |
| kens is the pdfwrite expert, but he's probably just about to disappear for the long weekend. | 16:52.01 |
kens | I am indeed | 16:52.08 |
chrisl | "looks identical" is the important bit - not necessarily "functions identical" | 16:52.18 |
ray_laptop | to both | 16:52.21 |
Robin_Watts | pihug12: What version of gs are you using? | 16:52.28 |
pihug12 | The latest one : 9.14 | 16:53.08 |
Robin_Watts | Your best course of action is to go to bugs.ghostscript.com and open a bug. | 16:53.44 |
ray_laptop | pihug12: since embedded fonts sometimes don't have a ToUnicode we may not be able to properly encode the characters on the output PDF | 16:53.52 |
Robin_Watts | attach the two smallest example files that show the problem you can | 16:54.15 |
| Hmm, sounding like Yoda, I am. | 16:54.31 |
ray_laptop | pihug12: but there are lots of ways PDF's can honk things up so the the text that comes out isn't correctly encoded, even though it looks fine | 16:54.50 |
chrisl | pihug12: can you be more specific about what you mean by "can't select the text"? | 16:55.35 |
ray_laptop | pihug12: yes, please specify (in the bug) what method you are using to select the text (e.g. Acrobat Reader ___) | 16:56.21 |
pihug12 | I'm using SumpatraPDF to read the files | 16:56.38 |
| *SumatraPDF | 16:56.43 |
ray_laptop | probably a mupdf problem then :-) | 16:56.56 |
pihug12 | I can upload the two files, there are 2MB both | 16:57.03 |
| Yeah, but it works with the source files | 16:57.19 |
ray_laptop | pihug12: to bugs.ghostscript.com, please | 16:57.21 |
pihug12 | OK Ray | 16:57.40 |
chrisl | pihug12: "can't select the text" - you can't highlight the text, or something else? | 16:58.14 |
ray_laptop | that way the "answer" (if there is one) gets captured for folks looking for similar issues | 16:58.29 |
pihug12 | Yes, by "select", I mean highlight/copy | 16:58.46 |
mvrhel_laptop | ok I finally have gsview startup time working well with large documents | 16:59.10 |
| not for a bit better performance when scrolling right after scale changes | 16:59.33 |
chrisl | pihug12: Right, but it's not the that copied text is wrong, or that the highlight isn't over the text you meant it to be | 16:59.35 |
pihug12 | Can I embed the files in your bugtracker? Or I have to host them somewhere else? | 17:00.48 |
chrisl | You can upload them direct to bugzilla | 17:01.03 |
kens | goodnight all | 17:02.05 |
chrisl | Robin_Watts: Oh, I meant to mention: I exchanged mails with Miah this morning, and we worked out why we couldn't find that license stuff..... | 17:02.22 |
| It is in a block, and the links on the block admin page are activated by javascript which throws some errors - which seems to stop certain versions of Firefox working with the page (but I found Chromium did work) | 17:03.40 |
Robin_Watts | chrisl: Ah, chrome failed on it too, cos that's what I was using. | 17:04.19 |
chrisl | Hmm, well, Miah is going to have a poke at the Javascript, and hopefully root out the error | 17:05.04 |
henrys_ | chrisl: hmm Miah doesn't work for artifex anymore⦠strange | 17:13.08 |
chrisl | henrys_: Miles just told me that - well, maybe he felt a duty of care..... | 17:13.46 |
| Robin_Watts: do you have details of shell access to the GoDaddy account? | 17:14.12 |
Robin_Watts | chrisl: The only details I have are on the DownloadsInformation.txt file. | 17:14.50 |
| I'm not sure there is shell access. | 17:15.06 |
chrisl | Hmm, I'm not sure how you'd setup a CMS without shell access..... | 17:15.43 |
| henrys_: the javascript in question is generic to the CMS, not specific to the artifex.com site | 17:17.14 |
| Anyways, have to go.... | 17:21.06 |
pihug12 | Bug created : http://bugs.ghostscript.com/show_bug.cgi?id=695204 | 17:31.22 |
| I add some troubles to upload the files. Seems like the bugtracker is slow. :-/ | 17:31.43 |
Robin_Watts | chrisl: OK, apparently we can enable ssh on the webhosting. | 17:32.16 |
| chrisl: Follow the instructions here: http://support.godaddy.com/help/article/4942/enabling-ssh-on-your-linux-shared-hosting-account?pc_split_value=1&countrysite=uk | 17:34.11 |
henrys_ | Robin_Watts: he left | 17:34.41 |
Robin_Watts | henrys: he'll see the logs. | 17:34.52 |
| It's probable that he can do everything he needs from the control panel though. | 17:35.07 |
henrys_ | Robin_Watts: interesting norbert's regression started with http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=7b3a65aab20feac334cac8e5935ba5cbe310ac69 | 17:35.46 |
Robin_Watts | I wonder what backups we have in place for the artifex website. | 17:35.50 |
| henrys: Well, you have to take that together with the one before. | 17:37.44 |
| It means that we are now working in larger stripes than before. | 17:38.22 |
| which may have adverse memory effects. | 17:38.52 |
| If Norberts target can cope with unaligned loads, then he might be able to set it back down to 16 with no ill effects. | 17:39.26 |
henrys_ | Robin_Watts: do you want to update http://bugs.ghostscript.com/show_bug.cgi?id=695195 or should I | 17:43.53 |
| ? | 17:43.55 |
Robin_Watts | I will if you want. | 17:45.09 |
henrys_ | go ahead | 17:45.43 |
| Robin_Watts: I guess we just don't have enough long playing jobs for the regression performance test to notice these. | 17:51.42 |
Robin_Watts | updated. | 17:54.36 |
henrys_ | thanks Robin_Watts | 17:58.07 |
norbertj | henrys_: Robin_Watts: that shaved off a 6-7 seconds, so now 9.14-fixed = 29.9 sec. (from 37.7s) | 18:04.20 |
| still have to see what it does in the integrated code. But I'm optimistic. Thanks. | 18:04.48 |
Robin_Watts | norbertj: No worries. | 18:05.56 |
norbertj | henrys_:Robin_Watts: bye, weekend now. | 18:28.13 |
| Forward 1 day (to 2014/05/03)>>> | |