| <<<Back 1 day (to 2015/08/04) | 20150805 |
rayjj | tor and chrisl (for the logs): The commit http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=ab674af32798e8b7ce46bb093acfe756d226cdf6 was to the ghostpdl.git repo, but didn't update the jbig2dec repo http://git.ghostscript.com/?p=jbig2dec.git;a=summary (the last commit shown there is from April) | 00:10.34 |
| I'll go ahead and push it, but we probably need to look into why it wasn't pulled in automagically (as chrisl indicated it maybe should have been) | 00:11.22 |
| grr... jbig2dec is locked down to tor (it's not like he wrote it) :-( I'm going to fix permissions to g+w and the group to gs-pub | 00:38.10 |
| (we can always undo it) | 00:38.23 |
| I have to shut down peeves and peeved for a while. Note that they will come up with different IP addresses (may take a while) -- Changing over to fiber optic at 25Mb symmetric (25/25) | 00:41.59 |
| wow! on bug 695968 I can get it down from 121 seconds to 4.2 seconds (and the output is identical) Now to try a regression run... | 03:18.51 |
sebras | tor8: morning! | 09:05.40 |
tor8 | sebras: morning. | 09:08.30 |
| rayjj: (for the logs) the jbig2dec.git repo is locked down because the only way it's supposed to be updated is via the crontab script that pulls from the ghostpdl.git | 09:09.59 |
| pushing manually is going to confuse the scripts | 09:10.05 |
| the reason it hasn't been updating recently is because of the directory restructuring ... the script is still looking for files in the old location | 09:12.02 |
kens | Heading out, back later | 09:12.44 |
tor8 | chrisl: argh. this jbig2dec crap is a mess | 09:52.12 |
| and given that ray has now changed the repository permissions and pushed manually (exactly the reason that I had the permissions set as they were, so this wouldn't happen) I'm inclined to let him do that for all future jbig2dec commits as well | 09:53.07 |
| the "git subtree" command isn't happy about directories moving :( | 09:53.43 |
| rayjj (for the logs): okay, I've updated the scripts to copy commits from ghostpdl to jbig2dec to cope with the directory restructure | 11:26.24 |
| rayjj: I've also undone the damage you did to the repository when you changed permissions and ownership and pushed a manual commit to it | 11:26.56 |
| rayjj: A git repository needs to be set up with certain configuration flags to be a 'shared' repository; just changing the permissions won't work | 11:28.00 |
chrisl | tor8: I can't think what else I could do other than say "ask Tor"! | 11:28.22 |
tor8 | pushing a commit without setting the config has created files that are owned by ray/ray | 11:28.24 |
| so I've reset everything back to tor/tor and made it non-writable and please be a bit more patient next time :) | 11:28.49 |
| chrisl: the directory structure caught me off guard, the git-subtree script kept looking for changes in gs/jbig2dec and got very confused when it couldn't find the directory | 11:30.24 |
chrisl | tor8: tbh, I completely forgot, otherwise I've have poked you about it (or hacked on it myself, given you were off at the time) | 11:31.07 |
tor8 | git-subtree is pretty stupid and can't cope with renames, so I've had to resort to git grafts and filter-branch to extract a new jbig2dec subtree for the new location and graft it on top of the old history | 11:31.19 |
| the upside is ... it now runs much faster! it no longer has to dig through 16k commits since I can limit it to just the history past the directory rename :) | 11:32.03 |
| sebras-outside: the sun! it burns? | 11:32.28 |
sebras-outside | tor8: no kidding! | 11:32.49 |
chrisl | tor8: it seems git-subtree is one area where git needs attention - I guess it doesn't get used much :-( | 11:33.01 |
sebras-outside | tor8: even if I'm sitting in the shade my legs are sticking out into the sun... I guess i should have brought sunscreen.... | 11:33.33 |
tor8 | chrisl: git subtree is a hack ... git submodules is the real solution | 11:33.36 |
chrisl | tor8: I don't like git submodules either - I *really* think stuff like that should be totally transparent to the "normal" user...... | 11:34.17 |
tor8 | git subtree is a wrapper around git filter-branch like functionality | 11:34.19 |
| git submodules are not ideally integrated, I will admit that | 11:34.43 |
| but like most things in git, once you grok how they actually work under the hood, everything makes sense | 11:35.02 |
| git is definitely not newbie friendly | 11:35.16 |
paulgardiner | tor8: you expected git-subtree to handle the a change in location of the subtree? | 12:07.32 |
tor8 | paulgardiner: no, but I expected maybe some command line flag to do the history grafting | 13:07.23 |
paulgardiner | I'd thought that in general git subtree was favoured over submodule because of the flexibility to work on the library and main source in unison | 13:09.03 |
jogux | I thought git subtree was generally assumed to be a terrible idea apart from some special one off cases like splitting up a repo that had several projects in it :) | 13:13.29 |
tor8 | my thoughts are more in line with jogux than paulgardiner on this issue :) | 13:13.57 |
jogux | though I also agree with tor that submodule isn't hugely better especially for beginners :) | 13:14.06 |
tor8 | but hey, at least it's not clearcase! | 13:15.51 |
paulgardiner | I can see that git submodule is great if it isn't your module. | 13:17.10 |
jogux | tor8: or perforce.... | 13:18.28 |
chrisl | Ah, clearcase - that nearly sent me to therapy! | 13:20.02 |
| back in a couple of hours..... | 13:31.06 |
kens | I see the Greek air traffic controllers are striking. Maybe Miles will need a ferry. | 14:21.47 |
sebras | chrisl_away: the clearcase version tree software... once you have a couple of thousand checkins it always crashed on me. :-P | 14:33.28 |
chrisl | sebras: at my first job (amongst other things) I was a clearcase configuration manager - one of the worst responsibilities I've ever been saddled with.... we use the full configuration management capabilities, which included object file cacheing which *never* worked! | 16:02.49 |
rayjj | strange, I ran a cluster test at (almost) head and got 7300+ files producing errors. My changes were in the pcl/pl directory but the errors were with pdf | 17:30.29 |
chrisl | They are all timeouts | 17:31.41 |
| Oh, not they;re not..... | 17:31.56 |
GregB | Hey guys, I was wondering if anyone knew of a way to use ghostscript to extract the layers as seperate images in a PDF? | 17:38.00 |
| sorry, I just don't believe in wasting my time, or yours. so I just blurted out the question. Hi All. | 17:38.32 |
rayjj | chrisl: I just updated to the latest commits on origin and they went away. The "bad" commit was based on: 6fc5e5a (origin/master, origin/HEAD) pdfwrite - fix some sorting problems in Names trees | 17:51.25 |
chrisl | rayjj: Hmm, strange..... oh well, at least it's cleared up | 17:52.07 |
rayjj | chrisl: my latest re-run that looks OK is b71c1d7 (origin/master, origin/HEAD) Coverity ID #101186 Remove some unused code to silence warnigns | 17:52.09 |
kens | GregB no you can't do that directly with Ghostscript | 17:52.15 |
| rayjj that commit ran fine for me, both on commit and when I cluster pushed it | 17:52.44 |
rayjj | kens: strange. Oh, well. it's gone away, so I'm happy | 17:53.17 |
GregB | kens, do you know of any software that would allow me to do that? | 17:53.56 |
kens | GregB I'm assuming you aretalking about optional content, so no. It doesn't really make much sense to do so. You can turn the optional content opn and off in Acrobat and export to a bitmap format from there. | 17:54.50 |
| ANd now its time for me to go, g'night folks | 17:55.38 |
GregB | ok, thanks kens | 17:55.56 |
chrisl | henrys: bad news, there is a *different* glyph missing in the new fonts - "afii61352", "Numero Sign", Unicode 8470 (decimal). | 18:00.56 |
mvrhel_laptop | bbiab | 18:13.45 |
henrys | chrisl: different from the set this set is meant to correct? or something you didn't notice before? | 19:00.09 |
chrisl | henrys: I can't parse that question..... | 19:01.24 |
mvrhel_laptop | ok I really hate submodules | 19:01.38 |
chrisl | henrys: The "afii61352" glyph was in the previous release from URW, and is not in this release | 19:01.59 |
mvrhel_laptop | I have gs and mupdf set up as submodules for gsview | 19:02.02 |
| for some weird reason they wont update | 19:02.12 |
chrisl | I assume you haven't committed to either of them? | 19:02.52 |
mvrhel_laptop | not on purpose... | 19:03.06 |
chrisl | That's not a categorical "no". | 19:03.27 |
mvrhel_laptop | :) | 19:03.33 |
| let me go through the logs real quick | 19:03.47 |
| no I have not | 19:04.51 |
chrisl | OKay, does it give an error? | 19:05.09 |
mvrhel_laptop | oh not its perfectly happy. I can empty out the entire directory and then I do a update submodule and it rechecks out the same old version | 19:05.45 |
| its like its not getting it from the gs repository | 19:06.25 |
| but from the gsview repos | 19:06.31 |
| perhaps I have things screwed up in some set up | 19:06.51 |
| that is clearly likely | 19:07.00 |
chrisl | it is. The gsview repo contains a pointer to a specific commit in the gs and mupdf repos, and that's the one it's using | 19:07.16 |
mvrhel_laptop | ah | 19:07.29 |
chrisl | I didn't mean it's clear you screwed up..... | 19:07.35 |
mvrhel_laptop | well that is clear to me | 19:07.43 |
| question 1: how do I fix that. question 2: how do I avoid doing that again | 19:08.01 |
chrisl | Basically, in your local repo, you have to update the commit that the submodules point to, and then push that change to the central repo | 19:08.58 |
| But I can't remember how you do that :-( | 19:09.07 |
mvrhel_laptop | chrisl: ok I will do some reading on this | 19:09.48 |
| should be able to figure it out | 19:09.57 |
chrisl | mvrhel_laptop: it looks like (for command line clients) you change into the directory of each submodule, and then do: "git pull --rebase origin master" | 19:13.06 |
| that should update the pointers to "HEAD" for repo, and you can then push that to the central repo | 19:13.28 |
mvrhel_laptop | ok cool . That was just what I was thinking. doing the pull for gs now | 19:13.49 |
chrisl | Okay - I was going to say I can't guess how that would translate into a gui client..... | 19:14.13 |
| Oh, and given you don't commit to submodules, you don't need the "--rebase" | 19:14.41 |
mvrhel_laptop | ok. I can do command line too | 19:15.16 |
| hmm ok so I did the pull and everything looks correct in the ghostpdl directory. Not sure what I need to push to the central repo though. Nothing has really been modified to push. Where are the pointers stored? | 19:25.21 |
| I thought after the pull you do an update, but that just brings me back to the old checkout | 19:25.46 |
| this should not be that difficult | 19:26.43 |
chrisl | Where's Tor when you need him?!? ;-) | 19:27.35 |
mvrhel_laptop | ok reading now. should be able to figure this out | 19:29.28 |
chrisl | So, it seems you need to do the pull in each submodule, then commit then, then push | 19:30.20 |
| mvrhel_laptop: seems like a decent set of hints here: http://blog.jacius.info/git-submodule-cheat-sheet/ | 19:32.37 |
mvrhel_laptop | thanks | 19:34.02 |
| that did it! thanks chrisl | 19:42.12 |
chrisl | Cool! | 19:42.54 |
mvrhel_laptop | off for lunch | 19:59.18 |
rayjj | well, I /think/ I have peeves IP addresses moved over. Now just waiting for DNS to propagate. | 20:08.11 |
| for the logs, peeves is 100.9.62.163 and peeved is 100.9.62.165 | 20:08.37 |
| they're on the faster 25/25 fiber network now | 20:09.13 |
| henrys: can you have a look at my pldict changes for bug 695568. I implemented a hash table approach to speed up searching. That file on linux goes from 375 sec down to 5, (from 115 sec to 5 on Windows). http://git.ghostscript.com/?p=user/ray/ghostpdl.git;a=commit;h=de4d00a4cadddda450497fb17ffb26ece677071f | 20:39.43 |
| henrys: I forgot to put in the log that I made the dict entries non_gc_memory as well (I didn't see any point to keeping it, and enumerating the table of pointers is a pain) | 20:41.02 |
| sort of strange that the profiles didn't hint at this level of performance boost, but it still works | 20:43.42 |
fredross-perry | I feel like an idiot right now. Getting back to gsproof/andriod after a few days away. I could SWEAR I saw some high level functions in ther Andriod build for counting and choosing separations, but I just cannot find them. Was it some strange acid trip? Thanks. | 21:47.43 |
| Here are some checkins to my MuPDF master that folks could look at: | 21:54.54 |
| 9e40dbe490d2d35c9527863c4b72ef45abcc3384 | 21:54.55 |
| adds a build.gradle file for use with Android Studio | 21:54.56 |
| fc0a4c350e7ee9c92a3fb351adf3e1c7f51c0823 | 21:54.58 |
| Changes to improve handling out-of-memory errors. Probably related to bug 695507, | 21:54.59 |
| bit I don't exactly remember now. | 21:55.01 |
| eba2d14a97704a2b84da23e93614271736c457fb | 21:55.02 |
| A first crack at adding iCloud support to the iOS buid. | 21:55.04 |
| More to be done here. | 21:55.05 |
| c495102107c238b4b0c8eb197b48b04d2aef1142 | 21:55.07 |
| Several weeks ago I was working on pushing code in the iOS build into a reusable | 21:55.08 |
| library component. It's in a separate folder. Comes with a test app that does some minimal | 21:55.09 |
| UI just to test the concept. | 21:55.10 |
henrys | rayjj: norbert's already done a fast pldict for pcl we wanted to understand the other performance problem. I'd just close the bug as not fix at this point. A hash table likely won't work, there are some font selections that depend on the order of the table. | 22:02.44 |
| norbert's solution does work because the order is preserved with a binary tree. | 22:06.29 |
| fredross-perry: just send tor an email to review in the morning or ask him tomorrow morning when he's here. | 22:08.09 |
fredross-perry | thanks | 22:08.37 |
rayjj | henrys: sorry. I was working on cust 532's issue (TilingType 2) and didn't see your reply. | 23:55.30 |
| Forward 1 day (to 2015/08/06)>>> | |