| <<<Back 1 day (to 2012/06/13) | 2012/06/14 |
kens | chrisl I see that everything on the cluster seems to be fine since you fixed the ref ** last night | 08:47.06 |
chrisl | kens: yeh, I think the pcl/pxl errors were "allowed" errors, but the cluster had lost its record of them - that made things seem worse than they were | 08:48.50 |
kens | Well, beats me... | 08:51.08 |
| At least my runs are making sense agin :-) | 08:51.19 |
chrisl | Well, there's no way a change to the PS interpreter could cause failures in the pcl/pxl interpreters, so...... | 08:52.15 |
kens | Hmm, I see customer #670 is 'low support'... | 08:52.16 |
| chrisl I was seeing that with a change in ps2write ! | 08:52.30 |
chrisl | kens: yes, but I saw only a few seg faults apparently caused by my mistake, whilst more than half of all the PCL/PXL tests were reported as throwing new errors | 08:55.35 |
kens | Yes, that's what was puzzling me | 08:55.50 |
| The seg faults were few enough to be worth investigating, the hundreds of PCL errors were a worry | 08:56.07 |
| And tis simply not possible that a minimal change in ps2write could break teh interpreter that badly | 08:56.28 |
chrisl | But it does seem that those are expected errors from PCL, so...... | 08:57.03 |
kens | Yes, not going to worry about it if Henry doesn't :-) | 08:57.33 |
chrisl | Exactly. | 08:58.29 |
| kens: Strange problem from Phil, that font thingy - do they produce the PS? | 08:59.43 |
kens | Yes they do :-) | 08:59.52 |
| I believe its their driver at fault, its certainly not us | 09:00.04 |
| As shown by the very simple modification to the file | 09:00.23 |
chrisl | I just wondered - the PS still seems needlessly convoluted, but much, much less so than the usual Adobe driver PS | 09:01.32 |
kens | Yep, its similar to the Adobe PostScript, but not the same. | 09:01.50 |
| My guess is that they are constructing the font from some larger (probably TrueType) font | 09:02.03 |
| Given that the copyright notice is Microsoft, bu thte font name is garbage | 09:02.26 |
chrisl | Yep, I don't much like the randomly dotted around eexec encoded data...... | 09:02.52 |
kens | They add glyphs to /CharStrings on the fly | 09:03.07 |
| The font is defined initially with one glyph /G0 then as glyphs are required they get added G0, G1, G2 etc | 09:03.35 |
| When a font is full, they switch to a new one | 09:03.48 |
| But it seems like they expect 'definefont' to make the font current, which it doesn't | 09:04.09 |
chrisl | Sure, but usually when I've noticed that happening, the "clear" PS takes specific steps to make sure that the correct font dict is on the stack, where that one didn't seem to - from what I saw | 09:04.46 |
kens | Of course the fonts are technically invalid since they don't include a /.notdef | 09:04.53 |
| chrisl when it defines the first font it does, but when it defines the second, it doesn't. | 09:05.18 |
| I'm as sure as I can be that its a driver problem | 09:05.44 |
| Especially since Distiller produces exactly the same. | 09:05.54 |
| As does Jaws | 09:06.05 |
chrisl | And Jaws and HQN, so if it's not a driver problem, it's a Postscript problem :-) | 09:06.25 |
kens | :) | 09:06.35 |
| Its certainly not our problem... | 09:06.42 |
kens | polishes off second customer bug report | 09:15.23 |
| back to XPS :-( | 09:15.32 |
chrisl | I wonder about our customers sometimes..... :-( | 09:21.18 |
kens | What, that last one ? | 09:21.37 |
chrisl | Yes: "9.05 works, 9.04 doesn't.... what should I do??" | 09:22.08 |
kens | :-) | 09:22.13 |
Robin_Watts | "Downgrade to 8.71" | 09:26.00 |
kens | Well in fact 9.04 will work too, if he actually gets the font name right in cidfmap | 09:26.42 |
| On the plsu side chrisl, it shows how valuable that change will be, when we can finally get customers to use a recent version.... | 09:27.12 |
chrisl | I guess so - although we will then have a tougher time explaining the few cases where it doesn't "work" - despite the clear disclaimer in the documentation | 09:28.23 |
kens | "Always look on the bright side of life" | 09:28.52 |
chrisl | I'm decidedly "glass half empty" today...... | 09:29.36 |
| kens: I just couldn't leave that issue with the vertical writing font alone - I'm pretty sure I know where pdfwrite's problem is, but I can't see a pdfwrite based solution..... | 14:17.07 |
tkamppeter | kens, chrisl, Robin_Watts, I mailed to TurboPrint and they answered me. They want to cooperate with OpenPrinting and they have put up the source of there GS patch. The supply of the old GS was a temporary workaround for a LibreOffice bug and will be removed soon. Intention is to turn TurboPrint in a pure CUPS Raster driver. | 14:43.41 |
Robin_Watts | tkamppeter: Regardless, to comply with the GPL, they must release the source code for their device. | 14:44.35 |
| This can be as simple as putting an archive of their source on their website. | 14:45.10 |
kens | chrisl, sorry been looking at CIDFont substitution, I missed you rcomment | 14:45.11 |
chrisl | kens: CIDFont substitution? I thought you'd done with Dan's problem? | 14:45.45 |
kens | Just checking it with 9.04, where it works just fone of course. | 14:45.59 |
| fine* | 14:46.04 |
chrisl | They may have random patches applied to theirs...... | 14:46.33 |
kens | Indeed, but that's not *my* problem :) | 14:46.44 |
| THey are a 'low support' customer so I'm not going any further for them | 14:46.55 |
chrisl | I did wonder if that was why he'd suddenly gone straight to bugzilla rather than mailing support | 14:47.26 |
kens | Yeah, I wondered that too :) | 14:47.38 |
Robin_Watts | Wow. This test file jumps between pages as I move the mouse pointer around on it in mupdf. | 14:48.37 |
kens | Intriguing bug | 14:48.53 |
| Oh god, now Phil is back again. I can't see me getting anything constructive done today | 14:49.27 |
henrys | kens:sorry I guess I missed that omtool bugzilla thing, can you remind him to send stuff to support? | 14:51.04 |
kens | I can add it to the bug thread if you like, or would you rather I mail him direct ? | 14:51.40 |
| OMG If I decompress Phils' 11MB file it comes out at 420MB | 14:52.14 |
| It must be all images | 14:52.25 |
henrys | I'd just add something on the bug | 14:54.01 |
kens | OK will do | 14:54.08 |
Robin_Watts | My God, it's full of images. | 14:55.05 |
kens | :) | 14:55.11 |
| Its also full of AcroForms | 14:55.21 |
| Great, the original file is a PDF 1.6 so all the useful stuff is buried in streams, and the 420 Mb file is too big for anything to open it. | 14:57.51 |
Robin_Watts | kens: mupdfclean -d -i -f in.pdf out.pdf | 14:58.22 |
| That leaves images and fonts compressed. | 14:58.30 |
kens | Aha, I'll d that thanks | 14:58.37 |
Robin_Watts | If you can isolate a particular page as being the problem, then you can extract that page and use -ggg to garbage collect unused objects out. | 14:59.28 |
kens | Looks like I need to get MuPDF 1.0 :) | 14:59.33 |
Robin_Watts | :) | 15:00.16 |
kens | I've been using a variety of older versions for quite a while, but those options aren't in 0.9 | 15:00.43 |
| Hmm, illegal option -- d | 15:02.32 |
| And it doesn't give a version number any more :( | 15:03.01 |
| Seems it shuold be -x I thought that was for old versions of pdfclean | 15:03.27 |
| Ah, name change mupdfclean, oops | 15:04.23 |
Robin_Watts | ah, sorry, yes. | 15:04.32 |
kens | THat seems to ork better :) | 15:04.49 |
| A mere 31MB file now | 15:04.59 |
| Ah, I see the file has a UserUnit of 5.5, that'll be why then. | 15:06.26 |
| henrys do we have a support policy on pdf_info.ps ? | 15:08.35 |
| Come to that, where is pdf_info.ps ?.... | 15:09.58 |
chrisl | toolbin? | 15:10.24 |
kens | Of course, where else, given that its not a binary file :-) | 15:10.39 |
henrys | kens:not really but the customer you are talking to matters. | 15:14.40 |
kens | henrys I know.... | 15:15.01 |
| I can extract the UserUnit easily, shoudl I display it separately, or multiply it into all of the boxes ? | 15:15.26 |
| separate display is easy, multiplication is not. | 15:15.37 |
chrisl | What units do we report the boxes in? | 15:16.10 |
kens | Whatever is in the PDF file | 15:16.16 |
| We just pull the array otu directly | 15:16.23 |
| That's the problem | 15:16.31 |
chrisl | So, it's in points IIRC? | 15:16.42 |
kens | The file has a UserUnit of 5.5 so the boxes we display ar 5 times too small | 15:16.44 |
| chrisl its in 'user units' | 15:16.58 |
| deaftul is 72/inch | 15:17.03 |
| But if UserUnit is 5.5 then teh unit is 72/5.5 per ich | 15:17.24 |
chrisl | Yeh, my feeling is we should print UserUnit, and report the boxes as they are | 15:17.35 |
kens | THqat was my feeling too, its more faithful to what's in the PDF file. | 15:17.54 |
chrisl | But I suspect I may be in the minority...... | 15:17.55 |
kens | I expect the customer will want it multiplied in. | 15:18.15 |
| But.... | 15:18.18 |
| I woudl argue this is an *informational* dump | 15:18.27 |
| SO the UserUnit should be broken out separately. | 15:18.40 |
Robin_Watts | kens: Output the user unit separately, and whoever is using the output can deal with it. | 15:19.31 |
| (i.e I'm with you and chris) | 15:19.53 |
kens | Consenses achieved :-) | 15:20.08 |
| I will mail them the file and update it in Git | 15:20.23 |
chrisl | I'd go so far as to say that if we multiplied out the boxes, we'd be wrong.... | 15:22.06 |
ray_laptop | kens: I saw your patch for pdf_info.ps -- is there a reason for not simply using the UserUnit value when reporting the page size ? | 15:40.22 |
| (rather than make the user calculate it) | 15:40.43 |
kens | ray_laptop : see above | 15:40.54 |
ray_laptop | kens: sorry -- I haven't read the logs -- just the email (so far) | 15:42.13 |
kens | :-) | 15:42.19 |
ray_laptop | consults the logs | 15:42.20 |
kens | It wasn't a unilateral decision, I consulted first. But I think its better for an 'info' program to split out the information separately | 15:42.58 |
ray_laptop | kens: I don't quite understand "separate display is easy, multiplication is not." -- seems easy enough | 15:46.38 |
| kens: I read the logs -- consensus of three of you. But the problem is that people using pdf_info.ps tend not to be PDF experts and probably don't know what to do with "UserUnit: 5.5" | 15:50.14 |
kens | I'd have to multiply each element of the array and then disply the modified array. | 15:50.15 |
| ray_laptop : Its an *informational* display. How many of them knwo the difference between MediaBox, CropBox and ArtBox ? | 15:50.48 |
ray_laptop | kens: right. mark exch { UserUnit mul } forall ] | 15:51.19 |
kens | If you want to produce another PostScript program to display simple information, then I'm OK with that, or even modify pdf_info.ps to do so. | 15:51.23 |
| ray_laptop : UserUnit isn't defined at that point | 15:51.38 |
ray_laptop | (above untested) | 15:51.39 |
kens | Right now we make custoemrs intersect CropBox with MediaBox, we don't do that for them | 15:52.17 |
chrisl | ray_laptop: also, they already have to know something about this area to know that user units default to 1/72 inch | 15:52.38 |
ray_laptop | kens: you could have a "/pagesizescale 1 def" at the start and then change it when you see the UserUnit (where you print it now), then use 'pagesizescalie' as the multiplier | 15:53.11 |
kens | ray_laptop : yes, I *could* do that, but see all previsou commetns | 15:53.27 |
| I don't think tis the right thing to do with this program | 15:53.36 |
ray_laptop | kens: I tend to agree with what you said "I expect the customer will want it multiplied in." | 15:54.20 |
kens | I could even have some simple math to turn the various Box values into inches and/or centimetres, but we don't do that right now either | 15:54.28 |
chrisl | ray_laptop: as the various media boxes are specified in "user units" it would not be an accurate relaying of the content of the file if we multiply it out....... | 15:54.48 |
ray_laptop | kens: that wouldn't be a bad idea -- as long as we display both (otherwise half the world wants the opposite of what we choose, or we need YAO) | 15:55.46 |
kens | ray_laptop : the proliferation of options is one problem I wanted to stay awy from | 15:56.12 |
chrisl | kens: actually, in my workflow, I prefer to measure fathoms, could you add fathoms, too? | 15:56.22 |
kens | chrisl how about london buses too ? | 15:56.42 |
| :-) | 15:56.48 |
chrisl | ;-) | 15:56.58 |
kens | I'm not sure what The Register unit of length is.... | 15:57.02 |
| I know area is Wales' | 15:57.13 |
| and volume of liquids is Olympic swimiming pools | 15:57.27 |
Robin_Watts | Well, on the second file I tried (the first with any real form in it), mujstest has found a bug :) | 15:57.27 |
| Hold on, I'll ask... | 15:57.51 |
kens | It appears to be the linguine | 15:59.03 |
Robin_Watts | Alas, drobe isn't logged in at the moment. | 15:59.04 |
kens | http://www.theregister.co.uk/Design/page/reg-standards-converter.html | 15:59.28 |
chrisl | kens: Looks like we have another "don't know what 'crash' means"....... | 15:59.43 |
ray_laptop | Robin_Watts: I suspect that mupdfinfo.c has the same issue -- ignoring the UserUnit setting (but it's hard to trace through the code to see if it's being applied automagically somewhere) | 16:00.03 |
kens | Oh so that's what he meant when he said 9.04 crashed.... | 16:00.09 |
| ray_laptop : just use the file that Phil sent :-) | 16:00.25 |
ray_laptop | kens: is that the 420Mb one ? | 16:00.46 |
kens | ray_laptop : its only 11mB if you don't decompress it | 16:01.04 |
| mupdfinfo gives the MediaBox values only | 16:01.18 |
| Which is correct, because that's what's in the file | 16:01.27 |
chrisl | kens: I suspect Dan's gs isn't finding the cidfmap...... | 16:01.36 |
kens | It doesn't report the UserUnit anywhere though. | 16:01.40 |
| So it has the same problem as the released pdf_info.ps | 16:01.56 |
| chrisl I was pretty specific in the bug about how to get that to work.... | 16:02.07 |
ray_laptop | kens: I guess you need to open a bug ;-) | 16:02.14 |
kens | ray_laptop : I suppose I probably should.... | 16:02.35 |
| chrisl actually taht might be a crash (of course he hasn't go the dr watson so we can't tell). | 16:03.18 |
| Oh actually, no he has -dNOPAUSE and -dBATCH so it woudl just exit | 16:03.40 |
chrisl | kens: Dan's command line doesn't have a -I option, so unless they use a non-default build (which they might well be)....... | 16:04.02 |
kens | Yes, but there's not message about looking for the missing CIDFont either | 16:04.24 |
| I have to say that looks very like our installed version of GS though | 16:04.52 |
| Notice the directory structure. | 16:05.02 |
ray_laptop | bbiab -- coffee | 16:05.06 |
| kens: it may not be finding the cidfmap because of search path issues (-h will tell him) | 16:05.50 |
kens | chrisl the date in the image is the same as hte date of the released 9.04 | 16:06.16 |
ray_laptop | unlike fonts, I think we don't "merge" cidfmap files from all of the search paths (as we do with FontMap) so we just get the first one | 16:06.47 |
kens | I believe we only get one | 16:07.00 |
chrisl | kens: Well, reiterate the -I point, then, and just me know if you need me to test anything..... yes, we only use the first cidfmap found | 16:07.42 |
ray_laptop | kens: so if he has a path before the one where his cidfmap really lives, it will be masked | 16:07.44 |
kens | chrisl it works just fine for me, *when rendering* | 16:08.00 |
chrisl | But not with pdfwrite? | 16:08.37 |
kens | ray_laptop : I believe he is using an 'out of the box' 9.04 installation and hasn't given a '-I'. Despite the fact that I was very clear in the bug thread about the setup. | 16:08.44 |
| chrisl I am about to try pdfwrite | 16:08.53 |
| But I'm too busy keeping up in IRC | 16:09.04 |
| Yes pdfwrite throws an error (not a crash) | 16:09.55 |
| Its not hte same as what he shows in his reprot | 16:10.18 |
chrisl | What's the error? | 16:10.35 |
kens | Its form the PDF interpreter 'Error reading a content stream...' | 16:10.54 |
| sigh, can't get it to work for rendering now either... | 16:11.33 |
chrisl | Well, I can't see how pdfwrite could influence *loading* of fonts, but then..... it's fingers appear everywhere :-( | 16:11.42 |
kens | It coudl throw an error after the font is laoded if it doesn't like it. | 16:12.00 |
| Indeed it does actually *crash* if you use pdfwrite nad that setup | 16:13.28 |
| I have n idea whay | 16:13.33 |
| why* | 16:13.36 |
chrisl | 9.04 or current code? | 16:13.56 |
kens | 9.04 just switching to current code | 16:14.14 |
| current code works | 16:14.36 |
chrisl | That's interesting...... | 16:15.28 |
kens | I'm not sure I care. Its fixed in the current code. | 16:15.53 |
| They are a 'low support' customer, do I really need to bisect and find the change ? | 16:16.08 |
| ray_laptop : noice that Phil had already modified the pdf_info.ps file :-) | 16:17.03 |
| ray_laptop : I need to go, can you answer Phil's latest question please ? | 16:18.02 |
| (not fair to ask Henry to look at PostScript ;-) | 16:18.29 |
ray_laptop | kens: about the linewidth ? I just did | 16:19.16 |
| and cc'ed support | 16:19.22 |
kens | ray_laptop : linewidth ? No I meant his question about using UserUnit to calculate the boxes *in inches*. | 16:20.03 |
chrisl | kens: I haven't seen that, so it probably didn't go to support | 16:20.32 |
nem`brs | greetings. are there any plans to implement custom keybinds in mupdf? | 16:20.41 |
kens | Ohn Bum, I'll forward it to support. | 16:20.45 |
ray_laptop | kens: for cust 670, it would probably be nice to ask them if they can upgrade. (wee usually support patches whenever possible at least one release previous) | 16:20.55 |
kens | Yeah the mail only went to me. | 16:21.06 |
| ray_laptop : please feel free to do so | 16:21.24 |
| I have to go | 16:21.30 |
ray_laptop | kens: he sent the one about linewdith only to me and on the first reply I forgot to cc support | 16:21.44 |
kens | Phil's email is going to support now. | 16:21.48 |
| ray_laptop : same as this one then :-) | 16:21.58 |
ray_laptop | kens: I'll answer Phil's next question | 16:22.08 |
kens | Thanks, goodnight all | 16:22.16 |
ray_laptop | g'nite, kens | 16:22.33 |
| time for my coffee now... | 16:23.32 |
Robin_Watts | tor8: ping | 16:34.17 |
| oh, bad shelly. | 16:58.40 |
| I'm going to do some git surgery to resolve the merge that shelly pushed. | 16:59.20 |
chrisl | Robin_Watts: I'd offer to drive over and smack him, but he does Tai Kwando........ | 17:07.19 |
Robin_Watts | chrisl: To his credit he realised that there was a problem and mailed me. | 17:07.44 |
| I wrote a long reply and sent it about how to resolve the problem - and then realised he'd pushed already. | 17:08.06 |
| but it was an easy fix. | 17:08.19 |
chrisl | Robin_Watts: can you mail him the git incantation that makes pull default to rebase - I can never remember it...... | 17:08.37 |
Robin_Watts | I will dig it up in just a mo, yeah... | 17:08.52 |
chrisl | I have a feeling I may have promised to send it, but got distracted by something else | 17:09.33 |
Robin_Watts | The problem is, it only works for newly created branches or something. | 17:09.54 |
chrisl | Oh, that's....erm.... sub-optimal | 17:10.39 |
Robin_Watts | It's of the form "Make all new branches rebase by default: y/n" | 17:12.25 |
chrisl | Robin_Watts: git config branch.master.rebase true | 17:14.08 |
Robin_Watts | Right. That one says: make master rebase by default. | 17:14.30 |
| but there is another one that says "for all new branches make them rebase by default". | 17:14.54 |
chrisl | So you'd need to do that for all your branches, and also do "git config branch.autosetuprebase always" | 17:15.25 |
Robin_Watts | That's the puppy. | 17:15.37 |
| And for any other branches you've already created. | 17:15.46 |
chrisl | I suspect Shelly primarily uses master and very transient branches | 17:16.28 |
Robin_Watts | chrisl: So you're mailing him those? | 17:17.06 |
chrisl | Yes, I'll do that | 17:17.15 |
ray_laptop | darn, ken didn't forward the UserUnit test file. I have to go find one | 17:17.29 |
Robin_Watts | ray_laptop: It's in support, no? | 17:17.41 |
ray_laptop | Robin_Watts: no, I think Phil sent it to him directly | 17:18.03 |
Robin_Watts | no, it's in the 15:40 mail from Phil, I believe. | 17:18.37 |
| in COMBINED.zip | 17:18.47 |
ray_laptop | Robin_Watts: I found fts_01_0106.pdf and fts_06_0626.pdf (2 and 11 respectively) | 17:37.42 |
Robin_Watts | ray_laptop: Did you see alexcher's comment on the Smask bug? | 17:38.27 |
ray_laptop | Robin_Watts: yes, and I agree that using the current bbox seems like an OK method (without digging into it further). | 18:05.40 |
henrys | hmm I wonder what would happen if we changed the all the license headers in contrib/ to be consistent with gs and if they object we just tell them we can't distribute it. | 18:12.45 |
Robin_Watts | ray_laptop: Short of performing radical surgery on the pdf14 device to allow us to make it so that buffers resize upwards as required, I can't see any other solution. | 18:13.53 |
| mujstest just found another bug in the same file. | 18:18.09 |
tor8 | Robin_Watts: did you want something? | 18:32.59 |
alexcher | I'm removing 40K /tmp/gs_* files taking over 500G that were generated by gs regression from i7a and i7b. Check your nodes. | 18:33.30 |
Robin_Watts | tor8: Just to say.. were you happy for me to push my mudraw and mujstest stuff ? | 18:33.50 |
tor8 | Robin_Watts: ah. yes. | 18:34.16 |
Robin_Watts | Thanks. Will do so. | 18:34.29 |
henrys | alexcher:I had the same problem on my nodes | 18:56.54 |
d3c | Robin_Watts: found a few more bugs, it seems. http://bugs.ghostscript.com/show_bug.cgi?id=693120 and http://bugs.ghostscript.com/show_bug.cgi?id=693118 | 19:39.38 |
ray_work | henrys: alexcher: on peeves I run a cron job that does: rm -f /tmp/gs_* | 21:39.33 |
| this doesn't rm files in use -- just all of the left over | 21:39.50 |
| of course, the cron has to run as root or regression | 21:40.27 |
| oops -- make that: as root or marcos | 21:41.29 |
| (mine is root) | 21:41.35 |
| maybe this should be run at the end of the cluster script (marcos/ henrys) | 21:43.22 |
henrys | ray_work:I think we should open a bug for marcos and see what the underlying issue is. I'll do that now. | 21:54.56 |
Robin_Watts | If wanted, I can add a line to the tasks run by each cluster job to rm /tmp/gs_* | 23:10.14 |
| Forward 1 day (to 2012/06/15)>>> | |