| <<<Back 1 day (to 2012/10/25) | 2012/10/26 |
Etale | Hi all! Is there a way for me to reduce a pdf's size, while controlling the quality of that resulting pdf numerically? (As opposed to using something like "-dPDFSETTINGS=/ebook") | 00:34.42 |
mvrhel | weird my cluster push I would have expected to be done by now | 00:39.46 |
| Robin_Watts: what is going on with the cluster? | 00:52.02 |
| Regresion terminated due to missing clustermaster.pl process? | 00:52.22 |
| this explains why my clusterpush is not working | 00:52.40 |
| henrys: do you have any ideas? | 00:53.02 |
| looks like marosw is not here right now | 00:53.29 |
| cluster machines claim they could not connect to casper.ghostscript.com | 00:56.06 |
| and so... the process starts again | 00:56.22 |
| Robin_Watts: maybe you know what is going on? | 00:56.39 |
| I can get to casper just fine | 00:56.46 |
| dinner time | 00:57.09 |
| bbiab | 00:57.11 |
Etale | Hi all! Is there a way for me to reduce a pdf's size, while controlling the quality of that resulting pdf numerically? (As opposed to using something like "-dPDFSETTINGS=/ebook") | 02:33.06 |
| Hi all! Is there a way for me to reduce a pdf's size, while controlling the quality of that resulting pdf numerically? (As opposed to using something like "-dPDFSETTINGS=/ebook") | 03:09.15 |
henrys | mvrhel:so it magically cured itself? | 03:32.59 |
mvrhel | apparently | 03:33.11 |
| like a clogged toilet | 03:33.15 |
| henrys: closed my last customer bug in bugzilla | 03:36.08 |
| of course I have 60 bugs still | 03:37.46 |
| and likely there will be work for customer 330 once he starts integrating | 03:39.01 |
henrys | congrats mvrhel | 04:23.29 |
kens | tor8 I have an XPS file which has 2 pages, I'd really like to only process teh second paghe, but -dFirstPage doesn't seem to work. Is there a way to do this easily or do I have to go and rewrite the XPS contents ? | 09:17.56 |
tor8 | kens: sorry, I don't think you can | 09:18.56 |
| you *could* add -dFirstPage support though :) | 09:19.05 |
kens | Ah, well I guess I need to go and try to remember XPS :-( | 09:19.18 |
tor8 | kens: you can run gxps on an unzipped xps if you point it to the _rels/.rels file | 09:19.50 |
kens | Yeah, I'm doing that now | 09:20.02 |
| THough it seems a biut unhappy about it | 09:20.20 |
| Iy gives me the seond page instead of the first (possibly a result ;-) and complains that it can't find 2.fpage | 09:21.27 |
tor8 | :/ | 09:22.27 |
kens | Well, I'll go and hack the contents | 09:22.54 |
| AH, it looks like I already have hacked this around, hence why it only runs page 2 | 09:33.30 |
| Obviously I haven't doneitright. | 09:34.09 |
| Fixed it :-) | 09:36.08 |
Robin_Watts | pops out for haircut. | 09:43.01 |
| back in 5 mins :) | 09:43.05 |
Nr18 | If i try to generate jpg's from a pdf i receive the following error: Unknown trailer dictionary key: 'DALiMInfo'. | 12:50.07 |
kens | Using what application ? | 12:50.23 |
Nr18 | gs 8.70 | 12:50.36 |
kens | Well the first thing to try is a version of Ghostscript that's less than 4 yearsz old.... | 12:50.59 |
| The current version is 9.06 | 12:51.06 |
Nr18 | actually a php script that does a exec | 12:51.13 |
| thanks will try to update and reproduce | 12:52.28 |
kens | Hmm "Regression terminated due to missing clustermaster.pl process" | 13:10.20 |
| Sounds kind of bad.... | 13:10.34 |
Robin_Watts | kens: It'll self heal in a few minutes. | 13:12.28 |
kens | I don't have to resumbit ? | 13:12.39 |
Robin_Watts | I don't think so. | 13:12.45 |
kens | Well,we'll see.... | 13:12.51 |
| Yes, seems to be restatring. Pity it was nearly done.... | 13:14.53 |
chrisl | I had that happen this morning, and run aborted, and I think mvrhel saw it last night, too | 13:18.29 |
kens | I got that impression form last nights logs | 13:18.40 |
| Ohterwise I might have been more alarmed... | 13:18.56 |
Robin_Watts | I haven't changed anything on the cluster for a few days at least, so I don't believe it's me :) | 13:18.56 |
kens | Looks more like comms problems | 13:19.05 |
| the clustermaster is on AWS ? | 13:19.18 |
Robin_Watts | on casper, yes. | 13:19.25 |
| and no, I don't think it's comms. | 13:19.34 |
kens | Yeah, I wonder if their East Coast Data Centre is having a problem | 13:19.45 |
| So what do you think it is ? | 13:19.51 |
Robin_Watts | the clustermaster process is run every few minutes. | 13:20.02 |
chrisl | Well, clustermaster.pl does seem to be being run on casper..... | 13:20.04 |
Robin_Watts | I think that error message is caused by a watchdog spotting that the current clustermaster has died. | 13:20.27 |
| so I suspect that (for some reason) the clustermaster process has exited unexpectedly. | 13:20.48 |
kens | Hmmmaybe, I notice feet and i7 are down now | 13:21.05 |
| and macpro is idle | 13:21.12 |
| Oh it all went funny again | 13:21.31 |
| Could be just dashboard confusion | 13:21.51 |
chrisl | We seem to have two instances of clustermaster running on casper - that probably isn't good..... | 13:24.00 |
kens | Sounds like a bad idea to me | 13:24.13 |
| Maybe it'll get better if I just leave it | 13:24.28 |
chrisl | Well, like I say, my run just auto-terminated | 13:25.03 |
Robin_Watts | chrisl: The clustermaster.pl is started repeatedly on a timer. | 13:59.27 |
| The first thing it does is look to see if there is an existing one running, and if there is, it exits. | 13:59.43 |
chrisl | yes, it's called from a crobtab entry. | 14:00.03 |
| When I first checked there were two instances running at the same time, but not now | 14:01.09 |
kens | So... | 14:12.32 |
| My job has disappeared from the cluster, its in a bad-looking state and I didn't get an email | 14:12.54 |
| Should I restart it do you think ? | 14:13.04 |
chrisl | So far 3 out the last 3 jobs aborted - I think it's buggered..... | 14:13.37 |
kens | Hmm, well I guess I'll have to wait for marcosw to turn up and make some mustical passes. | 14:14.17 |
| mystical | 14:14.23 |
| THough musical might work too | 14:14.31 |
kens | coffees | 14:14.49 |
chrisl | I assumed it was a combination of mystical and musical - "mustical" | 14:14.58 |
kens | AFK for a short while | 15:16.12 |
henrys | sigh more cluster problems? | 15:26.27 |
chrisl | It seems so, yes | 15:26.57 |
henrys | might be best to send marcos mail and not wait. | 15:33.14 |
Robin_Watts | I find that sending him a text works well, as he gets that even if he's away from his computer. | 15:35.33 |
henrys | I can do it if you like let me know. | 15:39.44 |
bz | is there a zoom-to-window-width in mupdf? as in the opposite of the shrinkwrap command | 15:41.41 |
henrys | right phone voice, sms, mail, irc - decreasing order of us needing the cluster to run. | 15:42.06 |
| have some errands to do this morning bbiaw | 16:19.00 |
sebras | bz: as far as I know -- no. however a new sample viewer is in development which aims to have this feature. the release date for this viewer not set at the moment. | 16:23.52 |
bz | sebras: cool, can i submit a patch for it? | 16:24.27 |
sebras | bz: oh, and I assume we are talking about the X11-based linux viewer and not android/ios/whatever. :) | 16:24.34 |
bz | sebras: yep | 16:24.40 |
sebras | bz: patches are always welcome of course. | 16:24.49 |
bz | sebras: well, i don't want to reinvent work that already has been done, so is this new viewer published anywhere? | 16:25.17 |
Robin_Watts | bz: You certainly can, and submissions are always nice. But there is no guarantee we will take it on as we are concentrating on the new viewer ATM. | 16:25.20 |
bz | git branch | 16:25.25 |
| hm, can't find a relevant remote branch on mupdf.git | 16:27.28 |
tor8 | sebras: bz: try key 'W' | 16:27.33 |
sebras | tor8: wait? what? | 16:28.02 |
bz | meh, just found it in pdfapp.c | 16:28.06 |
sebras | installes mupdf on work computer. | 16:28.14 |
bz | but debian has an older version, guess i'll have to compile | 16:28.39 |
| cool, capital W works | 16:34.48 |
| Robin_Watts: any chance i could compile that fancy new viewer on my own and try it out? | 16:35.00 |
Robin_Watts | bz: Sure. it's on tor's mupdf repo on git.ghostscript.com | 16:35.34 |
bz | Robin_Watts: thanks. | 16:36.39 |
mvrhel_laptop | hi kens | 17:05.08 |
| I hope it is not too late | 17:05.13 |
kens | about to go for dinner. | 17:05.23 |
mvrhel_laptop | hehe | 17:05.27 |
kens | Will try to be on later | 17:05.28 |
mvrhel_laptop | tell you what. I will try to get up early either Monday or Tuesday | 17:05.54 |
kens | I'll be back at some point this evening. I'm leaving this logged in. | 17:06.16 |
mvrhel_laptop | Monday works out pretty good | 17:06.16 |
| ok | 17:06.20 |
kens | Will be 90 minutes or so | 17:06.27 |
sebras | tor8: oh, W and H does work well, but I'm seeing a strange black column to the right of the page... hm... | 17:06.34 |
mvrhel_laptop | ok. I have to run to the kids school this morning for a bit so I will be in and out | 17:06.49 |
| Robin_Watts: was the cluster issue figured out | 17:07.14 |
Robin_Watts | mvrhel_laptop: Not by me. | 17:08.17 |
| but then I just ran a test and a bmpcmp and it was fine. | 17:08.28 |
mvrhel_laptop | indeterminism in the cluster | 17:08.46 |
Robin_Watts | It is possible that it's a communication thing; maybe if some communication hangs in a wierd way the clustermaster gets stuck. | 17:09.53 |
| which means it could be down to AWS. | 17:10.00 |
mvrhel_laptop | isn't ray_laptop working on some speed up with respect to interpolation of images when we have the pdf14 device? | 17:16.27 |
| Robin_Watts: do you know? | 17:16.32 |
| I am wondering if I should drop bug 693134 on him | 17:16.49 |
Robin_Watts | urm... | 17:19.52 |
| I don't know, sorry. | 17:20.44 |
| I'm trying to get landscape interpolation working at the moment. | 17:21.01 |
| and I believe that interpolation with banding should now be faster. | 17:21.32 |
| but I have nothing up my sleeve for interpolation + transparency. | 17:21.51 |
| though the presence of pdf14 does force banding in most cases. | 17:22.03 |
| kens job has killed the cluster again. | 17:22.34 |
| I do wonder if there is something specific with kens job that is causing this. | 17:22.45 |
mvrhel_laptop | yesterday it was you and me that were killing it | 17:23.28 |
| need to go | 17:23.29 |
| bbiaw | 17:23.31 |
Robin_Watts | mvrhel_laptop: Famous last words, but I don't think I've killed it recently. | 17:25.38 |
| Ok, I've just hopped my test in front of kens in the queue. This way, when the cluster restarts, we'll see if it manages to run mine. | 17:42.25 |
| If it then dies running kens, it will add credence to the idea that kens job is killing it. | 17:42.40 |
| I've split kens job up into 4 smaller jobs. | 18:05.43 |
| Even when split up kens job kills the cluster. | 18:19.19 |
| when both my test and the bmpcmp with it ran fine. | 18:19.34 |
| My test was only running a subset of jobs though, so lets try a full one. | 18:20.31 |
kens | Aggh, mvrhel ran away again | 18:30.43 |
| Robin_Watts : I can't think what about my test could be killing the cluster, but I can easily revert it and try a clean set of code | 18:31.32 |
| Robin_Watts : ping | 18:32.56 |
Robin_Watts | pong | 18:45.30 |
kens | Aha | 18:45.35 |
| Can you kill the pending jobs for me if they are just going to crash the cluster ? | 18:45.50 |
Robin_Watts | I can. | 18:46.07 |
kens | OK can you do that please, then I'll try a clean test, no changes to amster | 18:46.23 |
Robin_Watts | I wonder if my job is dying too. | 18:46.24 |
kens | Oh, it does look sick | 18:46.38 |
| THat's actually something of a relief | 18:46.51 |
Robin_Watts | Let me text marcos. | 18:46.56 |
kens | Yours looked OK a minute ago when I last checked it | 18:47.21 |
Robin_Watts | yeah. My gs highres -filter=ppmraw jobs all completed fine. | 18:48.49 |
| so either I was just lucky, or there is something in the lowres or non ppmraw stuff that's killing it. | 18:49.11 |
| I could try and bisect it a bit, but hopefully marcosw will have more of a clue. | 18:49.30 |
kens | Probably best | 18:49.40 |
Robin_Watts | I'll try running a full job with -filter=ppmraw cos that's enough for what I need. | 18:50.38 |
kens | OK, I inten to hang around, Micahel isn't getting away that easily :-) | 18:51.18 |
| Marcos is obviously about, he's sending email | 18:54.01 |
marcosw | Robin_Watts: the log messages for the cluster are very odd. I'm going to move casper to a new instance. | 19:06.24 |
Robin_Watts | henrys: http://ghostscript.com/~regression/robin/compare12.html | 20:39.15 |
| First one on that page. | 20:39.25 |
| So enabling interpolation gives a MUCH nicer picture. Makes you wonder if the non interpolated one is broken in that case... | 20:39.50 |
| Same with the top one here: http://ghostscript.com/~regression/robin/compare14.html | 20:42.26 |
| Oh, I get it I think. The images are 1bpp black and white predithered at high res. When displayed at 75dpi with no interpolation, everything posterises to black or white. | 20:45.56 |
| with interpolation, we get much nicer results. | 20:46.03 |
| http://ghostscript.com/~regression/robin/compare19.html <- top of that one too. | 20:48.29 |
| I declare this code seaworthy. | 20:53.32 |
| God safe anyone who sails with it. | 20:53.44 |
| Or something like that. | 20:53.47 |
henrys | oh I see you hardwired interpolation on for all files to do the test - we don't want that on in the pcl release by default. | 21:01.58 |
| ow wow Robin_Watts ... nice work | 21:11.09 |
mvrhel | that is nice | 21:38.13 |
Gigs- | is this the fastest way to test for corrupted xrefs? gsc -dNOPAUSE -dBATCH -sDEVICE=nullpage -sOutputFile=/dev/null -f | 22:04.30 |
| I just need gs to get to the point where it will send out the xref warning | 22:04.46 |
| Is there a case where an xref can be valid, but the byte offsets aren't pointing at ... the objects? | 22:47.26 |
| I know about incremental PDFs and all, this apparently isn't one | 22:47.50 |
| It's a 1.3, so it claims | 22:50.41 |
| ghostscript doesn't complain, by my own script I wrote to fix up xrefs is moving all the offsets by about 3000 bytes | 22:58.27 |
| does ghostscript use a search radius around where the xref claims an object is? | 23:00.09 |
Robin_Watts | henrys: Yes, only hardwired on for the test. | 23:16.45 |
| Gigs: In theory you can prepend a certain amount of crap on a PDF file, and it will still count as being valid. | 23:17.52 |
Gigs- | Robin_Watts: it doesn't hurt for me to find the objects and put their real offsets into the xref right | 23:18.16 |
Robin_Watts | (The start of file marker has to be within the first n bytes of the start of the file. Where n is something like 1024 or 4096) | 23:18.31 |
Gigs- | hmmm | 23:18.46 |
Robin_Watts | In the case where there is crap before the marker, I believe the offsets are taken from the marker. | 23:18.56 |
| You could use pdfclean to fix broken pdfs? | 23:19.25 |
| sorry, I mean: mutool clean in.pdf out.pdf | 23:19.38 |
Gigs- | %PDF-1.3^M%âãÏÓ^M | 23:20.01 |
| 1 0 obj^M<< | 23:20.01 |
| there isn't a bunch of stuff before object 1 | 23:20.13 |
Robin_Watts | ok, so no prepended crap. | 23:20.24 |
| Offsets in the xref should be correct then. | 23:20.32 |
Gigs- | that's the thing, gs doesn't warn on the original, or on my corrected version with the offsets moved by nearly 3000 | 23:21.29 |
| actually it's over 4000 | 23:21.44 |
| (energy level) | 23:21.46 |
Robin_Watts | gs may be silently correcting. | 23:21.49 |
Gigs- | the reason I discovered this is because I had a 2000 byte search radius, and then fallthrough to handle if it the object went missing | 23:23.17 |
| that fallthrough code path was never tested because usually it isn't a problem | 23:23.39 |
| then I realized that about 1% of our files have xrefs that far off from reality to start with, it wasn't something I was doing to them | 23:24.24 |
Robin_Watts | Gigs-: Simplest solution is to run mutool clean on the files. | 23:25.12 |
Gigs- | I can't go that path at this point. My script is guaranteed to only make minimal changes | 23:25.48 |
| it's changing the names of the colorants and then fixing the xref, and that's it... that way customers can't say we broke things | 23:26.19 |
Robin_Watts | mutool clean with no options will do no structural changes to the file. | 23:26.24 |
| All it does is read in every object, then write them out again. | 23:26.41 |
| With other options it works more magic. | 23:26.53 |
Gigs- | what about interobject noise | 23:26.54 |
Robin_Watts | random crap between objects? It's ignored. | 23:27.07 |
| removed. | 23:27.09 |
Gigs- | I'll definitely take a look at that, but for this project I need to keep as much of the broken stuff broken as possible :P | 23:27.37 |
| because it might be broken in just the right way heh | 23:27.50 |
| that's one reason I'm concerned about fixing these xrefs | 23:28.24 |
| 50,000 legacy files, some of them dating back to 1999 | 23:29.14 |
| think of it as preserving an antique painting | 23:29.28 |
| you could strip all the varnish off and have at it, but there is something to be said for being gentle | 23:29.44 |
| thanks for the help, I need to get back to it so I'm not here all night | 23:30.48 |
| Forward 1 day (to 2012/10/27)>>> | |