| <<<Back 1 day (to 2015/11/05) | 20151106 |
stevenm | is it possible to edit a PDF... in a way that only the chunks of the original that *need* changing - gets changed | 10:47.23 |
| i.e. not a full re-compile of the original pdf | 10:47.29 |
| kind of like 'Direct stream copy' in VirtualDub :) Not that Video/Audio is comparable to PDF :P | 10:48.05 |
chrisl | stevenm: basically, no. | 10:52.46 |
| stevenm: *sometimes* Acrobat can manage that for very small kinds of edit. And you can decompress the PDF and edit it by hand *if* you are very familiar with the internals of PDF. But in general, it's not really viable | 10:55.00 |
Robin_Watts | stevenm: Well, there is some code in MuPDF that lays the foundations for being able to do that. | 10:55.07 |
| But it would require C coding skills to take it further. | 10:55.26 |
stevenm | chrisl, is that what that ghostscript studio is capable of doing? | 10:55.45 |
Robin_Watts | And it would rely, as chrisl says, on you being familiar with the internals of PDF. | 10:55.55 |
chrisl | stevenm: I have no idea what ghostscript studio is - it's nothing to do with us | 10:56.26 |
stevenm | ok | 10:56.41 |
| thanks chrisl & Robin_Watts | 10:56.48 |
Robin_Watts | https://ghostscriptstudio.codeplex.com/ | 10:56.52 |
chrisl | So, really just a Windows GUI for gs, so no, not really | 10:57.34 |
stevenm | and I'd want a Linux UI anyway | 10:57.46 |
chrisl | You could do something along those lines with Ghostscript, but you'd need a *lot* of Postscript programming knowledge..... | 10:58.12 |
| Anyway, I have to head out for a while..... | 10:58.20 |
Robin_Watts | Is Ken off today? | 15:01.41 |
chrisl | Robin_Watts: yes | 15:08.54 |
Robin_Watts | back monday? | 15:09.03 |
chrisl | Yep | 15:09.14 |
Robin_Watts | I'll drop Phil a mail saying that Ken is away, hence explaining the crickets. | 15:09.21 |
chrisl | I think the answer will be "no" anyway..... | 15:09.55 |
tor8 | mvrhel: mvrhel_laptop: Robin_Watts: I wrote down some ideas and thoughts about the 'mutool create' project. Hopefully that will set you in the right direction. | 16:16.13 |
| It's in the mail. | 16:16.20 |
henrys | Robin_Watts: he probably wants -Z: so he can see the time and memory for each page, feel free to give it to marcosw or ray to followup | 16:16.41 |
mvrhel_laptop | hi tor8: is that on a wiki someplace? | 16:16.41 |
| ah ok | 16:16.45 |
| got it | 16:17.00 |
Robin_Watts | mvrhel_laptop, tor8: Should we put it on the twiki? | 16:17.19 |
mvrhel_laptop | this is great. thanks | 16:17.52 |
tor8 | Robin_Watts: I dislike the wiki... formatting lists and code is a PITA there... | 16:18.00 |
| but yes, this should probably live in the wiki. some of these ramblings could be useful to other people as well. | 16:18.39 |
Robin_Watts | But for collaboration, it's far better than throwing files back and forth via email. | 16:19.02 |
tor8 | mvrhel_laptop: forgive me for asking, but how familiar are you with the pdf file format? If you've never like Robin and I, written a PDF parser, then I may assume things that you just have no clue about yet :) | 16:20.10 |
mvrhel_laptop | tor8: familar enough to get myself in trouble | 16:20.59 |
tor8 | If there are things that are unclear, just say "stop" and we'll get you caught up. | 16:22.00 |
mvrhel_laptop | I have read the spec and I am familiar with a fair amount but I suspect it will take me a bit to get up to speed on a few things | 16:22.10 |
Robin_Watts | tor8: You mean, he might not have had his expectations of sanity dashed on the Adobe rocks yet ? | 16:22.11 |
tor8 | mvrhel_laptop: if you want to keep your sanity, and get a basic grasp of things, start with PDF reference 1.3 :) | 16:22.49 |
| most things learned there are still true; you just skip all the worst complications and nightmares. For a brief orientation I find it a better place to start than with the current specs. | 16:23.37 |
| Of course, you'll still need the current specs when doing the actual work :( | 16:23.51 |
mvrhel_laptop | so my hard copy book is actually for version 1.0 | 16:28.07 |
| I suppose I do need to reread the the doc and file structure sections | 16:30.08 |
Robin_Watts | mvrhel_laptop: I wouldn't worry too much. | 16:33.10 |
mvrhel_laptop | still reading tor's email.... | 16:33.25 |
| ;) | 16:33.27 |
Robin_Watts | If you understand 1.0, then the structure is broadly similar until you get to 1.5 when they do compressed object encodings. | 16:33.42 |
chrisl | When did compressed xref streams and incremental xrefs come in? Was that 1.5? | 16:34.54 |
Robin_Watts | chrisl; incremental xrefs were there from the start. | 16:37.29 |
| compressed xref streams were 1.5, with compressed object encodings, I think. | 16:37.58 |
mvrhel_laptop | tor8: ok thank you for the details. I think I am on the right path with the create tool | 16:38.53 |
marcosw | henrys: your macpro seems to be having trouble on the cluster, it times outs trying to download source files. I tried logging in but it appears your ip address changed or your router isn't forwarded or something else broke. I was going to suggest just taking it off the cluster but it's the only non-Linux cluster node. | 16:39.02 |
mvrhel_laptop | I have more or less been doing what you have stated in the MUTOOL CREATE SKELETON | 16:39.13 |
| I am at the point now where I am going to start creating the objects for the fonts and images | 16:39.47 |
henrys | marcosw: I do have it on power line networking and the electrical in my garage is not "top notch", I wonder if that could be the issue | 16:44.43 |
| marcosw: when/how long? | 16:45.16 |
marcosw | when did it start failing? The last time it completed a cluster run was 3 days ago,. but it may have been unreliable before then. | 16:54.14 |
henrys | marcosw: not seeing anything particularly interesting in the logs, can you make if fail now while I'm looking? | 16:57.44 |
marcosw | i can try, things never fail when someone is looking at them. | 16:58.03 |
henrys | marcosw: or I can just set you up with a login again. | 16:58.12 |
marcosw | i've started a clusterpush and your macpro is stuck on "Fetching source from users/marcos" while all the other nodes have moved on to building ghostscript | 16:59.51 |
henrys | marcosw: let me try resetting the network in the basement. | 17:06.36 |
| marcosw: the network reset did make my remote login more lively but I don't know if that was the actual issue. | 17:10.38 |
marcosw | I'll try again. | 17:10.47 |
henrys | marcosw: I feared that would happen I may have to relocate the machine. | 17:15.46 |
marcosw | henrys: no, still not working. If you rsync a large file from your mac to your i7 what kind of speeds do you get? | 17:15.55 |
henrys | I was just doing scp and seeing 5 MB/s, which is pokey for ethernet but it shuld work, let me try rsync | 17:17.18 |
marcosw | As a point of reference I'm seeing 12.42 MB/s over my MoCA adapter, but you are right, 5 MB/s is plenty fast enough. | 17:18.21 |
henrys | yeah rsync seems about the same, let me try rsyncing from casper. | 17:19.34 |
marcosw | my problem with powerline adapters was consistency, presumably related to whatever appliance was on or off (probably the refrigerator). | 17:21.21 |
henrys | marcosw: rsync from casper did fine similar speed. | 17:21.54 |
| marcosw: if you can think of something else for me to check I will but maybe best if I just get you a login. | 17:22.58 |
marcosw | can you look at ~marcosw/cluster/users/marcos/rsync.log? Also what is datestamp of that file? | 17:24.18 |
| As a test I'm trying a clusterpush as henrys, let's see if that works. | 17:25.24 |
henrys | marcosw: it was stamped 10 minutes ago | 17:26.29 |
Robin_Watts | henrys: Have you been in contact with Mike Detnya ? | 17:26.50 |
| I haven't seen any emails to him since Miles said he'd leave it to us. | 17:27.14 |
| I don't need to be involved of course, just want to know that it hasn't been dropped. | 17:27.28 |
marcosw | henrys: what were the last 4 or 5 lines of the file? | 17:27.38 |
henrys | Robin_Watts: no Miles took over, Mike initially avoided our "buyout" question and I asked Miles to follow up directly and ask again. It's on my todo list to check with Miles again | 17:28.33 |
Robin_Watts | henrys: Ah, OK. I didn't see that question go past. I must have been dropped by that stage. As I say, not objecting to being dropped, just didn't want to be the one that was doing the dropping :) | 17:29.27 |
henrys | total: matches=0 hash_hits=0 false_alarms=0 data=0 | 17:30.10 |
| | 17:30.10 |
| sent 35002 bytes received 226287 bytes 16857.35 bytes/sec | 17:30.10 |
| total size is 124604971 speedup is 476.89 | 17:30.10 |
| | 17:30.13 |
| marcosw: lgtm | 17:30.17 |
marcosw | The transfer speeds some reasonable, it must be something else. Can you put the file ~marcosw/cluster/macpro.dbg onto casper where I can get to it? | 17:31.25 |
henrys | Robin_Watts: no thanks for checking in about it. I should add it to the agenda. so it isn't lost | 17:31.37 |
| marcosw: home directory | 17:33.32 |
| on casper | 17:33.42 |
marcosw | thx | 17:34.07 |
| oops | 17:35.16 |
| henrys: I had a "$$=" instead of "$t=" in the mac only version of the cluster code. Setting the pid confused the code. | 17:37.53 |
| all good now. SOrry to have distracted you from actual work. | 17:38.14 |
henrys | marcosw: testing is actual work ;-) | 17:38.30 |
Robin_Watts | And people say perl is unreadable. | 17:38.49 |
marcosw | Perl is a write only language. | 17:39.06 |
henrys | writable by owner - 0300 | 17:41.20 |
| I wonder if lisp will have it's revenge. Clojure is a really nice language and it's getting some traction. | 17:43.13 |
Robin_Watts | I remember an email going around in the 1990s saying that the code for Ronald Reagans Star Wars initiative was written in Lisp. | 17:45.09 |
| Supposedly someone had stolen the code, and to prove it, they posted the last page of it on the net. | 17:45.29 |
| )))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))) | 17:45.41 |
| )))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))) | 17:45.44 |
marcosw | I liked lisp (otoh I liked forth too). I still occasionally use ']' to close multiple '('s when writing emails (you've probably noticed that my emails have lot's of asides (and sometimes the asides have asides]. | 17:46.17 |
Robin_Watts | In a previous life I was responsible for maintaining a hardware compiler that was written in SML. Similar stuff I think. | 17:47.04 |
marcosw | I like the Clojure comment on Object Orientation is overrated: "now used for everything, even when inappropriate" | 17:48.12 |
henrys | Hickey has a lot of interesting things to say that get people going, probably because he TDD (thought driven development). | 17:49.34 |
| s/he/he uses | 17:49.41 |
Robin_Watts | That's a marvelously passive aggressive term isn't it ? | 17:52.09 |
| I use "thought driven development" (unsaid: "unlike the rest of you who clearly don't think at all.") | 17:52.47 |
henrys | he's goofing on test driven development which is all the rage now. And I think that style of work can lead to thoughlessness | 17:53.30 |
marcosw | henrys: so I'm wasting my time working on regression testing code :-) | 17:54.12 |
| I should just be looking at the commits and thinking how they might cause regressions? | 17:54.49 |
henrys | marcosw: the modern tdd discipline is to have every line of code you write be a consequence of a failing test, it is having a falling out now as many are criticizing it. | 17:55.44 |
| hickey objects to that, not testing in general | 17:56.27 |
chrisl | It's a poor extrapolation of the "every line should solve a problem" approach | 17:56.36 |
henrys | I think it's fine for tic tac toe and scoring bowling which are the classic examples folks give, but in a real world I can't see how it would possibly work. | 17:58.26 |
| s/real world/real world program/ | 17:58.44 |
| oh I don't like gs_alloc_bytes definition that just came into support. If there is a performance problem let's have a bug and fix it for all. geez | 18:05.59 |
marcosw | bbiaw | 18:10.07 |
rayjj | henrys: (don't know why my chatzilla didin't auto connect) | 18:49.52 |
| henrys: Part of the reason that MALLOC and FREE are macros in gxclmem.c is to allow for this. gs_alloc_bytes does cause some overhead, but not a lot. | 18:50.37 |
| henrys: I guess we need to see how much, but I know that 'lock' is slower on Windows than on linux | 18:51.24 |
| BTW, I have the monster 2.5Gb test file | 18:51.50 |
henrys | rayjj: I forgot why we are locking in the first place ;-) all modern mallocs have locks or separate arenas. | 18:56.01 |
rayjj | henrys: we lock (as I mentioned in the email) because we are maintaining a linked list of allocated blocks, and a total and max allocation, all of which require a lock for thread safety (gs_alloc_bytes can be called by any thread at any time) | 19:49.41 |
| henrys: if we used TLS to maintain per-thread lists and totals/max usage, then we wouldn't need to lock. IIRC, you and Stefan were really strongly against using TLS | 19:51.03 |
Robin_Watts | TLS is bad, TLS is evil. | 19:58.38 |
rayjj | Robin_Watts: and all OS's have it internally (and many crtl's) | 19:59.07 |
henrys | rayjj: can you respond to ron licht please? Thanks. | 20:40.51 |
rayjj | henrys: oh, sorry. I'll do that (and set up the record he needs) | 20:45.56 |
henrys | oh right we have to march down the linked list and free everything because we can't write code that doesn't leak in ghostscript. It's all coming back to me now, lucky me... | 20:45.59 |
rayjj | henrys: there are only a few things that leak, and since we only do the 'free all" at the heap alloc level when we exit, we really don't have to do the list maintenance | 20:47.38 |
henrys | rayjj: and allocation sizes can be kept per thread and we could get rid of all that crap. that would be really nice | 20:48.35 |
| rayjj: thanks for getting back to ron. | 20:50.05 |
rayjj | henrys: keeping allocation sizes per thread would require either TLS, or some hack equivalent to it | 20:51.13 |
| henrys: we keep allocation size per thread when using multi-threaded rendering since each thread has its own chunk allocator (where the allocation is tracked). | 20:52.50 |
| other threads (the windows "display" thread) aren't really a concern | 20:53.39 |
henrys | rayjj: yeah so what is the problem exactly? | 20:54.05 |
| -K is an issue, but a minor one that I think few would miss. | 21:29.44 |
| Forward 1 day (to 2015/11/07)>>> | |