| | <<<Back 1 day (to 2013/11/21) | 2013/11/22 |
| tor7 | oh, that's rich. sourceforge is spamming google code projects that have downloads to move to sourceforge... | 10:56.36 |
| chrisl | Yeh, I saw that..... | 10:59.57 |
| paulgardiner | tor7: Hi. A couple more iOS commits on paul/master | 12:13.16 |
| | tor7: and one more little commit | 12:47.09 |
| tor7 | paulgardiner: there's some messed up whitespace in one of them | 12:50.23 |
| | those switchCase methods are getting pretty long by now... | 12:50.32 |
| paulgardiner | tor7: oh okay. I'll take a look | 12:50.37 |
| tor7 | annotation deletion has one line that's indented with all spaces | 12:50.50 |
| paulgardiner | tor7: *spaces*?!?! Shocking. I'll get rid of those straight away. | 12:51.48 |
| tor7 | paulgardiner: if you run the scripts/gitsetup.sh script it will set up a git alias "git wsfix" and "git wsfixi" (i for 'index') to clean up white space errors in the workspace and index | 12:52.28 |
| paulgardiner | The switch thing does have the nasty property that I have to change n lines and add 1 to update to n+1 cases | 12:52.28 |
| | I could do it with target and selector which would avoid that, but then I wouldn't be able to handle all the cases in line so neatly. Not sure whether it is overall a win or not | 12:53.32 |
| | tor7: re wsfix. right ta | 12:54.10 |
| tor7 | paulgardiner: it only works on the diffs though, so not as useful to fix things after the fact | 12:55.00 |
| | paulgardiner: #define the switchCase prototype? (and Robin will shout at us for using macros) | 12:57.38 |
| | and I believe it's spelled "deprecate" :) | 12:58.26 |
| paulgardiner | Updating the switches each time isn't much of a pain in Xcode fills them in automatically after typing just the first few chars, but still I'd like to find another way, with the same advantages, but less wordy. Anonymous classes in Java does it really neatly | 13:04.58 |
| | tor7: So do the white-space problems seem to be in just the one file. | 13:06.31 |
| tor7 | grepping through the source, pdf-jsimp-v8.cpp has a couple of odd whitespace problems too | 13:15.38 |
| paulgardiner | tor7: Ok. White space and spelling fixed, I believe | 13:16.05 |
| | tor7: okay. I'll fix that too. | 13:16.34 |
| tor7 | paulgardiner: actually, there's a rather lot of whitespace errors in the .m files ... I suspect Xcode is to blame | 13:17.31 |
| | paulgardiner: there is a way to set up git hooks to disallow or automatically clean up whitespace problems | 13:18.02 |
| paulgardiner | Damn! I thought I'd kept and eye on that. It's a pain: I don't like fixing a load of them after the fact becaue that puts a barrier in the git history for doing diffs across several commits. | 13:18.59 |
| tor7 | paulgardiner: git diff -w and git blame -w allows you to ignore whitespace changes when doing diffs and logging | 13:20.03 |
| paulgardiner | Can I get gitk and gitgui to do that as default? | 13:20.50 |
| tor7 | paulgardiner: "git config apply.whitespace fix" | 13:21.58 |
| | the .gitattributes setting should affect source files, but it looks like we don't have *.m in the patterns | 13:22.42 |
| | paulgardiner: there's a commit on tor/master to fix the .gitattributes to include objc files | 13:25.21 |
| | paulgardiner: to do what as default (diff -w or fix ws errors?) | 13:26.09 |
| paulgardiner | tor7: I'm easy. Whatever you think best. | 13:27.59 |
| | tor7: I've stuck one more on paul/master fixing the other .m file white space. Are these good to go? | 13:35.10 |
| tor7 | paulgardiner: yes, all LGTM now. | 14:02.02 |
| Robin_Watts | So... in the psdcmykok device I write planes 1-5 out to temp files. | 14:03.39 |
| | and at the end of the process I seek them all back to 0, and copy them onto the end of plane 1, then delete the temp files. | 14:03.59 |
| | This works fine on windows and linux. | 14:04.06 |
| kens2 | delivery brb | 14:04.26 |
| Robin_Watts | on macos, the seek works (ftell reports 0), but all freads return 0, and feof is not set. | 14:04.30 |
| kens2 | Robin_Watts : but pdfwrite uses fseek a lot, and as far as I know it works on macos | 14:11.13 |
| Robin_Watts | kens2: I was opening the scratch file "wb". | 14:11.26 |
| | Everyone else lets you read from that, macos doesn't. | 14:11.36 |
| | "w+b" seems to solve it. | 14:11.40 |
| tor7 | Robin_Watts: yeah, the spec requires you to open with + in order to do both read and write | 14:23.36 |
| | http://pubs.opengroup.org/onlinepubs/009695399/functions/fopen.html | 14:23.38 |
| | the choice of "r" or "w" is whether or not to create a new file when you open it | 14:24.06 |
| Robin_Watts | tor7: yeah, fixed now. | 14:24.30 |
| | I also only just found out about "x" | 14:24.52 |
| tor7 | I'm not surprised linux and/or windows lets you get away with it, but not the *bsd based kernels | 14:24.56 |
| | "x"? | 14:25.18 |
| | oh, a gnu libc extension. I'd advise against that and use open(O_EXCL) follow by fdopen() for better portability | 14:26.18 |
| Robin_Watts | no, it's a C99 thing I think. | 14:26.46 |
| Robin_Watts | lunches | 14:29.26 |
| tor7 | Robin_Watts: no mention of "x" in the C99 draft | 14:29.27 |
| | (which is pretty daft, seeing as it's a really good security feature to have) | 14:30.41 |
| kens | Robin_Watts : your commit seems to introdcue 60 errors, or was that a prioir one ? | 14:33.22 |
| | Seems they are all in the cups device, so I don't see how it could have been you | 14:33.49 |
| kens | notices that my cluster run had 105 errors, all in cups, when my change was for pdfwrite :-( | 14:34.29 |
| Robin_Watts | kens: All the cups failures are on henrysx6 | 15:17.44 |
| | I bet henrys hasn't installed the cups stuff | 15:18.07 |
| kens | aha, didnt' think of that | 15:18.29 |
| henrys | Robin_Watts: ah yes. | 15:18.43 |
| kens | however, it means I can stop looking for my bug and do a commit | 15:18.48 |
| henrys | maybe the cluster should have an install script or something that checks all this. | 15:19.17 |
| Robin_Watts | marcosw.pl | 15:19.26 |
| Robin_Watts | returns to the JNI salt mine then. | 15:19.36 |
| henrys | kens:let me try and fix this cups thing sorry to waste time, silly that we don't have a test script when we change or upgrade machines. | 15:47.15 |
| kens | Not a problem henrys, just had me puzzled for a while | 15:47.30 |
| | Foolishly I nboticed that Robin's test all failed with cups, but didn't notice the same thing about mine..... | 15:48.13 |
| henrys | kens:hmm it's a bunch of machines - fathoms, parsecs etc. I do have cups installed on henrysx6 - I'll ask marcosw1 before I fool with this. | 15:53.33 |
| kens | Makes sense I think | 15:53.45 |
| | henrys the machines on the left are the one used for hte previous run as I understand it, the one opn the right is the machine used in the current run, and that's always henyrsx6 for me | 15:54.39 |
| henrys | kens:ah okay let me try a simple command line with cups | 16:00.24 |
| chrisl | henrys: you need libcups2-dev and libcupsimage2-dev packages - if you are using Ubuntu. | 16:05.14 |
| | henrys: tbh, after arguing so vehemently against the idea, I'm coming round to the option of letting the cluster use the "built-in" libcups sources..... | 16:06.36 |
| henrys | chrisl: yes I'm missing image2 for some reason ... | 16:06.49 |
| | it was there before and I did was upgrade | 16:07.07 |
| chrisl | henrys: for some crazy-assed reason, libcupsimage2-dev isn't a dependecy of libcups2-dev. So it might have been removed as "no longer needed" during the upgrade | 16:07.47 |
| henrys | okay should be fixed, I'd still like to have a test script for the cluster - adding to the agenda. | 16:09.08 |
| kens | SOunds like one for Marcos | 16:09.31 |
| chrisl | henrys: surely the cluster is its own test script - when you setup a machine, do a push, and that will test it | 16:09.47 |
| henrys | chrisl: I suppose that is true I should have just done a dummy push | 16:11.20 |
| marcosw1 | there is a cluster node setup script, it's on casper in ~regression/cluster/setupcluster There is also a readme file in that directory which lists the ubuntu packages that are needed (the setupcluster script doesn't do the apt-get but probably should). | 16:40.26 |
| henrys | marcosw1: well I did a local push and all looks reasonable now. | 16:48.33 |
| Robin_Watts | marcosw1: I hope I've fixed the psdcmykog problems now. Sorry about that. | 16:49.49 |
| ray_laptop | hmm... just figured out what's going on. bmpcmp compares the test code output to the the output of the reference code, BOTH with the extras= option that I gave it. | 16:55.04 |
| | Since the reference is broken, it finds a bunch of spurious differences :-( | 16:55.28 |
| | I guess there is no way to have the cluster compare the the reference code without the option | 16:56.18 |
| | I'll just have to do it on peeved or peeves | 16:56.43 |
| Robin_Watts | ray_laptop: htmldiff.pl might be useful to you. | 16:58.13 |
| marcosw1 | Robin_Watts: running another psdcmykog test as we speak. | 17:01.04 |
| ray_laptop | Robin_Watts: Thanks. I'll have a go with it | 17:03.17 |
| | Robin_Watts: how come it has '\\' path separators instead of '/' ? | 17:06.41 |
| | Robin_Watts: so it looks like I can hack it to add --saved-pages-test to cmd and not cmd2 and compare the two modes | 17:09.26 |
| Robin_Watts | ray_laptop: Cos I wrote it on windows? | 17:10.29 |
| | It was the thing that pngs2html.pl was developed from. pngs2html is the script bit of the clusters bmpcmp support. | 17:12.01 |
| ray_laptop | Robin_Watts: OK. I was just curious. For the most part I use / on windows since it works in shells. I assume that perl on windoze also can open files with / separators (because $fileadjust uses / ) | 17:18.28 |
| Robin_Watts | I've heard you claim that windows supports '/' before, but my experience is that it's best not to rely on it. | 17:24.51 |
| | but I make no claims to this code being 'clean'. | 17:25.03 |
| ray_laptop | Robin_Watts: since I am running msysgit shell, I _had_ to change '\\' to '/' to make it work (and also had to install ImageMagick and build bmpcmp.exe) but it's WORKING !!! Thanks | 17:57.37 |
| Robin_Watts | ray_laptop: Fab. | 17:59.02 |
| ray_laptop | has my poor laptop on its knees :-( | 18:02.46 |
| | oops. bmpcmp: Failed to malloc 1361477592 bytes | 18:04.10 |
| | It died on Bug690395.pdf (I'll start restart the list from after that) | 18:06.10 |
| Robin_Watts | ray_laptop: Yeah, it will tie up whatever machine you use fairly effectively (unless it's a multicore beast, I guess) | 18:11.49 |
| | oh, if you're testing background printing or num rendering threads then it'll even kill beasts :) | 18:12.20 |
| ray_laptop | Robin_Watts: After it died, I had 64 directories in my results directory (out) and 'compare.html'. I only had a total of 18 .png's so there were only 6 actual differences, right ? | 18:19.41 |
| | the display of the compare.html shows most of the files with just the index#: test_description | 18:20.39 |
| | I may change it to just not emit anything of there are no bmpcmp differences from the test | 18:22.38 |
| | Robin_Watts: it seems that htmldiff.pl ignores $band (it parses it, but never emits -dMaxBitmap | 18:30.49 |
| | luckily, that's what I want because I want to compare gs with --saved-pages-test to -dMaxBitmap=0 :-) Clever of you to anticipate me wanting this, Robin_Watts | 18:31.48 |
| Robin_Watts | :) | 18:32.26 |
| ray_laptop | mvrhel just came online (according to skype). It's 4 AM there :-D Maybe their heading for the airport today | 18:52.57 |
| | Robin_Watts: I hacked up the htmldiff.pl to only open the iframe if (stat($outdir."/out.".sprintf("%05d",$images).".bmp")) (before the while) then only close it in an else clause for the if ($diffs == 0) that prints the Failed to find any differences | 19:05.19 |
| | Robin_Watts: so far, comparing --saved-pages-test mode against -dMaxBitmap=0 seems to show no differences ! | 19:06.49 |
| Robin_Watts | ray_laptop: I thought mvrhel was back home. | 19:42.29 |
| | I thought they were flying yesterday. | 19:42.39 |
| mvrhel_laptop | just got home about an hour ago | 19:44.08 |
| ray_laptop | mvrhel_laptop: welcome back ! | 19:46.06 |
| mvrhel_laptop | ray_laptop: thanks | 19:46.38 |
| | flight went pretty fast. only 8 hours and I slept 4 of those | 19:46.54 |
| ray_laptop | Robin_Watts: I mis-read the htmldiff -- I needed to set $outdir and make the directory first, but now it's working the way I want | 19:47.09 |
| | mvrhel_laptop: WOW only 8 hours ? | 19:47.18 |
| | that's the advantage of being so far north | 19:47.52 |
| mvrhel_laptop | Yes. It was pretty fast. I was on the B787 Dreamliner | 19:47.57 |
| | I had a bulkhead seat too which was nice | 19:48.15 |
| ray_laptop | that's handy if you want to use your laptop :-) | 19:48.49 |
| | otherwise you have to tilt the keyboard up (at least I do, unless I have an 'extra legroom' seat) | 19:49.29 |
| mvrhel_laptop | yes | 19:49.35 |
| | it was interesting the seat backs dont tilt | 19:49.57 |
| | but instead the bottom part slides out | 19:50.13 |
| ray_laptop | mvrhel_laptop: we look forward to hearing all about .jp :-) | 19:50.18 |
| mvrhel_laptop | ray_laptop: yes. Thanks for getting back about the job pausing and starting question | 19:50.46 |
| ray_laptop | mvrhel_laptop: is that comfortable ? seems like it wouldn't be | 19:50.52 |
| mvrhel_laptop | I knew you would have a way to do that | 19:50.53 |
| | ray_laptop: well it took some getting used to. but it does keep one from crushing the person behind them | 19:51.34 |
| ray_laptop | mvrhel_laptop: actually kens beat me to the punch on responding (but not about the gp_check_interrupts). I forgot to mention that many printer implementations use gp_check_interrupts to CANCEL jobs as well (not just pause) | 19:52.11 |
| mvrhel_laptop | I think kens response was not for the customer though right? | 19:52.42 |
| ray_laptop | depending on how attractive the woman is behind me, laying back is preferred | 19:52.51 |
| mvrhel_laptop | ha | 19:52.55 |
| ray_laptop | mvrhel_laptop: right, but kens cc'ed the customer accidentally | 19:53.14 |
| | mvrhel_laptop: at least with that type of seat, sitting in the last row before a bulkhead isn't an issue :-/ | 19:54.43 |
| | more than 50% of the way through my 'differences' and none of them are any other than clist/non-clist diffs (comparing --saved-pages-test to -dMaxBitmap=0). That's because using --saved-pages forces clist mode | 19:59.09 |
| | Robin_Watts: thanks again for pointing me to the .pl script. Even though I hate hacking in Perl, it saved me a *lot* of time (probably). At least I now have more than just a hammer :-) | 20:00.17 |
| | hmm... we get a lot of openjpeg warnings from some of the 1.7_FTS files | 20:04.37 |
| | so I've discovered a way to generate all of the differences between banded and page mode :-) I may run that next (just take out the -dMaxBitmap=0 on the reference $cmd2 args) | 20:08.11 |
| | interesting. 1 real difference (the last one on the list: 12-07C.PS). At least it was worth it going through the entire list | 20:15.44 |
| | time for lunch. bbiaw | 20:18.39 |
| | Forward 1 day (to 2013/11/23)>>> | |
IRC Logs