| <<<Back 1 day (to 2012/06/25) | 2012/06/26 |
tor8 | Robin_Watts: const pointers to matrixes in that case! | 08:01.34 |
sebras | Robin_Watts: I'm mostly complaining because it's not consistent. | 08:25.51 |
| Robin_Watts: I can't find any text about rollover in pdfref though. it may be in the jsapi | 08:30.41 |
| Robin_Watts: yes, field/mouse enter and field/mouse exit. | 08:32.22 |
| Robin_Watts: hm... then further states are needed, because currently there is only down/up there is no enter/exit. and the x11 api would need to be taught to distinguish down/up from enter/exit and only take action for textinput widgets for the former type of events. | 08:33.44 |
kens | chrisl I just realised somethign about this stupid PostScript file I shuld have spotted before | 08:41.24 |
chrisl | kens: oh? | 08:43.14 |
kens | The parentheses have a different matrix to the otehr glyphs | 08:43.28 |
| THe parentheses are rotated, the other glyphs aren't | 08:43.45 |
chrisl | Yes, that's where I spotted the matrix disparity between rendering and pdfwrite | 08:44.27 |
kens | indeed. | 08:44.37 |
| So pdfwrite is actually emitting 2 fonts, one for each orientation | 08:44.56 |
chrisl | Well, that makes some sense | 08:45.26 |
kens | THey are using the same descendant font though | 08:45.54 |
chrisl | Hmm, that's weird, so where is the rotation coming from? | 08:46.32 |
kens | It gets baked into the text matrix | 08:46.42 |
chrisl | I don't follow - the parentheses are in a string with other glyphs..... | 08:48.01 |
kens | They are int eh PostScript, but not in the resulting PDF | 08:48.13 |
chrisl | Yes, I meant where does the rotation appear from in the PS input? | 08:48.41 |
kens | Ah, now thre's the question :-) | 08:48.51 |
chrisl | That does explain the crazy CID for them, though - they are probably the vertical "versions" of the glyphs, manipulated to work in a horizontal layout...... | 08:50.05 |
kens | Yep, I'm assuming that's what this is all about | 08:50.18 |
chrisl | The rotation must be coming from somewhere in the font, given that, in my cut down file, the three glyphs are all in one string, handled with a straight "show" | 08:52.01 |
kens | Absolutely, but I have no idea where its coming from | 08:52.17 |
| I had been assuming that the CIDFont has 2 descendants. | 08:52.32 |
chrisl | The CIDFont has *four* descendants, apparently..... | 08:55.15 |
kens | <sigh> well that's 2 more than I was exepcting... | 08:55.30 |
chrisl | Yeh, but don't forget how much stuff I cut out - we're only *using* two | 08:55.57 |
kens | OK well I guess that makes some limited sense. | 08:56.10 |
| FWIW Acrobat makes it all into one font and just meddles with the Text Matrix too. | 08:56.38 |
| It 'sort of' looks like hte parentheses are wrong because they haven't had the vertical 'flip' applied | 08:57.20 |
| Actually, that can't be true | 08:58.22 |
| Anyway, back to head-scratching. | 08:59.10 |
chrisl | Obviously, you can't keep the same font layout since PDF only allows one descendant - I'm surprised distiller munges into one font, though....... | 09:01.59 |
kens | It makes it a horizontal font too, WMode = 0 | 09:02.19 |
| The difference is pretty minor, we just end up writing one dicitonary more than Distiller, so I'm not bothered | 09:02.44 |
chrisl | I guess distiller must have a heuristic to spot things like this, and unpick it - seems like a pretty dodgy thing to do | 09:06.05 |
kens | Nothing new there then.... | 09:06.27 |
| I *think* hte actual problem is the calculation of the DW2 values, which isn't too surprising I suppose. | 09:09.28 |
chrisl | Well, I thought the problem was *all* the "PDF metrics" for the problem glyphs - *but* only the vertical ones would actually be applied when reading the PDF. | 09:11.21 |
kens | Its a vertical font (CMap has WMode = 1), so all the metrics get applied. The verticla metrics include a | 09:12.15 |
| origin shift (both x and y) and the problem seems to be that this is getting applied to the text matrix and (possibly) as the vertical metric | 09:12.51 |
| THis is why it works OK if I set the CMap WMode to 0, the shifts are already applied in the text matrix. | 09:13.32 |
chrisl | I thought we would output W and DW, too? | 09:13.48 |
kens | We do | 09:13.54 |
| But those only shift the current poitn after the glyph has been drawn | 09:14.06 |
chrisl | I would have expected those to be wrong, too, but I guess not.... | 09:14.23 |
kens | And since we emit a Tm to reposition the text after each glyph, they have no effect | 09:14.25 |
| However the vertcial metrics shiot the origin *before* drawing the glyph | 09:14.47 |
| shift* | 09:14.55 |
| Even that isn't completely true :-( | 09:15.45 |
| Setting teh WMode to 0 gets the parentheses rightm but the other text is wrong :-( | 09:16.16 |
| Anyway, its all intermingled | 09:16.21 |
chrisl | The main thing (from my point of view) is that we agree that there's more to this than CDevProc results being just ignored in certain places..... | 09:17.11 |
kens | Oh yes, that's not the problem at all. The CDevProc resutls are getting used (and should be) and the descendant font matrix is also being applied | 09:17.47 |
| The problem seems to be in the application of the actual CDevProc values. | 09:18.10 |
chrisl | As I said before, I don't think the implications of adding the CDevProc callout support in there were fully thought through :-( | 09:20.13 |
kens | I'm sure you are correct.... | 09:20.24 |
chrisl | In fact, it looks to me like it was added to support PDF input where the CDevProc applies the width metrics - in PDF, this kind of CIDFont arrangement simply can't arise..... | 09:22.08 |
kens | Hmm, just done some quick comparisons. It looks like the ontly text which is actually *correctly* placed are in fact the parentheses (comapring with Acrobat) ;-) | 09:22.31 |
| chrisl yes, you are quite correct I'm sure. | 09:22.45 |
chrisl | That's strange.... I take it our rendering is correct? | 09:23.23 |
kens | I haven't checked but I would lay odds it is | 09:23.40 |
| chrisl our rendering seems to match Acrobat as closely as can be expected | 09:25.43 |
| certainly the gross positioning is the same, even if the finer aspects differe slightly | 09:26.18 |
chrisl | Well, that's something. | 09:26.41 |
kens | Ah, if I meedle with the vertical metrics for the non-rotated text, it looks better. I think this may be the real issue here. All 0 is not good..... | 09:27.55 |
| and indeed, not correct | 09:28.02 |
| BTW I'm using your much cut-down file (3 glyphs) here, because anything larger is too hard to work with.... | 09:28.53 |
| Yep, I can get perfect rendering just by altering the W2 entries for the non-rotated text. Now to find out why those are all 0..... | 09:30.38 |
| But first, caffeine infusion... | 09:31.01 |
chrisl | I *really* hope if we can get those three glyphs right, everything else in the full file will also work...... | 09:31.03 |
kens | chrisl I'm moderately confident this is the case. | 09:31.23 |
chrisl | When I originally cut down the file, the glyphs I left were all from the same subfont, so I missed these problems :-( | 09:31.53 |
kens | given teh way /W2 entries are sued (partially reliant on the /W entry) an entry of [0 0 0] doesn't look at all correct | 09:31.54 |
| Its a pretty horrible use of the TT font.... | 09:32.29 |
scott-san | robin are you on? | 13:02.52 |
| Robin_Watts | 13:03.08 |
kens | Hi scott-san | 13:03.12 |
scott-san | HI Ken, How goes it? | 13:03.37 |
kens | tearing hair out :-( | 13:03.47 |
| Robin_Watts : may be at lunch, cna anyone else help ? | 13:04.00 |
scott-san | Oh oh! | 13:04.00 |
| No, I need to see if he's available for a technical conference call on Thursday or Friday. | 13:04.41 |
kens | scott-san : just had an ameil to support asking for a quote, I'll forward it to you | 13:04.50 |
| Oh wow, they want HP/UX :-( | 13:05.12 |
scott-san | OK thanks. I'll send Mr. Watts an email re: the meeting. See you Ken :-) | 13:05.56 |
kens | BB | 13:06.01 |
| Hmm I see the mail is from a Mr Palatino, what really ? | 13:08.27 |
Robin_Watts | Hi scott. | 14:13.49 |
| Just replying to that email now. | 14:13.53 |
| Any time thursday or friday is fine for me. | 14:14.00 |
scott-san | Hi Robin. Okay, that's all I was checking on. Thank you. | 14:14.12 |
| Great. We'll see what Brian comes back with and put it together. Off to get a flight physical. Thanks again. | 14:14.44 |
Robin_Watts | tor8: hey. | 14:23.51 |
| just answering a customer mail for scott. When did we put DroidSansFallback into MuPDF ? | 14:24.19 |
tor8 | Robin_Watts Feb 27, 2009 | 14:24.57 |
Robin_Watts | ok, pre 0.9 then :) | 14:25.07 |
henrys | Robin_Watts:i assumed you'd want to do that meeting and volunteered you. | 14:25.38 |
Robin_Watts | henrys: Sure, no problem. | 14:25.47 |
vtorri_ | tor8: is the build system of mupdf 1.0 different from 0.9 ? | 14:42.44 |
| or relatively similar ? | 14:43.00 |
Robin_Watts | vtorri_: Not substantially. | 14:43.01 |
| The makefiles have probably been tweaked a bit, but nothing major. | 14:43.15 |
tor8 | vtorri_: the build system has not changed in any significant way the past handful of releases | 14:43.49 |
vtorri_ | good | 14:44.05 |
| maybe some files suppresion/addition ? | 14:44.13 |
Robin_Watts | no idea of the top of my head. | 14:44.26 |
vtorri_ | ok | 14:44.29 |
| i'll see | 14:44.32 |
tor8 | vtorri_: files have moved around, been added and removed. but the make targets etc are the same | 14:44.56 |
vtorri_ | i really want to use it, especially for the thread feature of 1.0 | 14:45.02 |
| we want a responsive GUI and that feature can help us a lot | 14:45.28 |
tor8 | vtorri_: the current build system dates back 2 years | 14:45.43 |
vtorri_ | from 0.8, i guess | 14:46.02 |
| it was quite different in 0.7 | 14:46.10 |
tor8 | vtorri_: you can have a responsive GUI by running mupdf in a background worker thread, like our iOS app | 14:46.35 |
| without any need for the multithreaded locks etc | 14:46.54 |
vtorri_ | anyway, iitc, there are some examples in doc/, right ? | 14:47.18 |
| iirc* | 14:47.25 |
sebras | vtorri_: yes there are, and I hope you find them useful. | 14:49.16 |
| vtorri_: if there are any problems with them let me know. | 14:49.27 |
vtorri_ | ok | 14:49.49 |
henrys | meeting day need some coffee. | 14:52.32 |
Robin_Watts | ooh, yes. | 14:53.55 |
henrys | paulgardiner:thanks for doing the item breakdown. | 15:00.16 |
paulgardiner | henrys: np. Does make much more sense that way | 15:00.49 |
henrys | about item (1) what do the other products do? Chrome, Adobe? | 15:01.49 |
paulgardiner | On the centring issue? | 15:03.01 |
henrys | oh yes... | 15:03.12 |
| I didn't know that was a legal UK spelling. | 15:04.23 |
paulgardiner | I'll take a look. I am sort of centring, but just not as accurately as it could be done. | 15:04.37 |
kens | Hmm, cluster seems to be broked | 15:04.46 |
Robin_Watts | kens: It wasn't me. I didn't do it. You can't prove anything. | 15:05.09 |
kens | It ran OK then a bmpcmp got stuck on uploading files so I ran an abort, and now its totlaly stuck | 15:05.33 |
| I will wait a while to see if it gets better | 15:05.42 |
| ah its idle again | 15:06.01 |
paulgardiner | henrys: it may not be. Spelling isn't exactly my strong point. :-) | 15:06.03 |
kens | will rerun bmpcmp | 15:06.10 |
| NB apparently henrysx6 is down | 15:06.26 |
paulgardiner | Chrome doesn't get it perfectly certral either | 15:07.08 |
henrys | well then let's not worry about it and press on. | 15:07.23 |
paulgardiner | Yeah. Best left to the polishing phase | 15:08.02 |
henrys | I'm not seeing much we need to discuss, Robin_Watts, tor8? | 15:08.25 |
Robin_Watts | Item 8, paulgardiner had something... | 15:08.43 |
henrys | looks like good progress as usual. | 15:08.44 |
tor8 | paulgardiner: I can't imagine anyone who uses comb fields caring about the spacing :) | 15:08.48 |
paulgardiner | tor8: No. Quite. :-) | 15:09.44 |
| Robin_Watts: Yes. Thanks for reminding me. | 15:09.52 |
| I'm looking at spread-sheet-like files | 15:10.07 |
| There are a load of functions not documented in the main reference. I've found possibly all the definitions in a file called Aform.js | 15:10.51 |
Robin_Watts | I think aform.js is part of Acrobat Reader (at least it was part of version 5.0) | 15:10.56 |
| Hence it's copyright Adobe. | 15:11.09 |
tor8 | depending on what app methods are called in our tests, we may want to add some way to perform the ones that make sense in our viewer(s). | 15:11.15 |
paulgardiner | That's what's worrying me. | 15:11.18 |
| The copyright that is | 15:12.09 |
henrys | are you sure it is not a liberal license, adobe does that from time to time. | 15:12.24 |
| ? | 15:12.29 |
Robin_Watts | http://sfk.gfz-potsdam.de/intern/CD/ISS%20Education%20Kit/AcrobatReader5Mac/Acrobat%20Reader%205.0/Contents/MacOS/JavaScripts/AForm.j | 15:12.32 |
| Thats the file. It looks like a expanded CD image to me. | 15:13.07 |
| I can't find the same file anywhere with a license file attached. | 15:13.21 |
paulgardiner | So far I've been including js source in MuPDF as C strings. I'd like to do that same with this (perhaps zipped), but the copyright may be a problem. | 15:13.50 |
tor8 | paulgardiner: feel free to make it a step for the build system to generate the C file from a plain text javascript file, much like we do for cmaps and fonts. | 15:14.46 |
paulgardiner | Could have mupdf read it in as an external file, but that complicates install. | 15:14.48 |
| tor8: that's a thought. | 15:15.04 |
Robin_Watts | That doesn't help the basic license problem. | 15:15.30 |
| There is, I think a javascript SDK from Adobe. | 15:15.43 |
| It's possible the same file is in that... | 15:15.55 |
henrys | hmm why wouldn't this file be in the regular acrobat release? I don't see it. | 15:18.24 |
paulgardiner | tor8: with the cmaps and fonts are we currently bothering to compress them to save on the executable's footprint? Or is that not really an issue? | 15:18.24 |
henrys | maybe compiled in now? | 15:18.38 |
Robin_Watts | http://www.scribd.com/doc/82959820/115/How-can-I-make-my-documents-accessible-A#outer_page_75 | 15:18.39 |
paulgardiner | henrys: I couldn't find it either | 15:18.40 |
Robin_Watts | The stadnard Acrobat JavaScript implementation comes with 3 Javascript files; Afrom.js which contains built-in ca-re-canned functions, Annots.js which is used by teh Annotations plug-n and ADBC.js used by the ADBC plugin. | 15:19.30 |
| These are located in the application Javascripts folder. | 15:19.38 |
paulgardiner | I have a glob.js, but that's it - plus that's empty | 15:21.29 |
henrys | I'm not finding any of those. | 15:21.55 |
Robin_Watts | I wonder if they have vanished in later Acrobats ? | 15:23.04 |
| paulgardiner: Can you name a file that uses some of the pre-canned stuff offhand ? | 15:23.30 |
paulgardiner | calc.pdf | 15:23.49 |
henrys | I was thinking they might have built them in. We can run a test to see if things like the ssn regexp is defined right? | 15:24.03 |
paulgardiner | CATX5977.pdf | 15:24.13 |
| Acrobat certainly still supports it all. Chrome too. | 15:25.23 |
Robin_Watts | Does Chromium support it ? | 15:26.33 |
| If so, then we can look in the Chromium source code. | 15:26.49 |
paulgardiner | Yes. Both files work in Chrome | 15:27.05 |
Robin_Watts | No, chromium, not chrome. | 15:27.17 |
| chromium is the proper free open source version of chrome. | 15:27.32 |
paulgardiner | Oh. I don't know | 15:27.44 |
tor8 | paulgardiner: no, we have them in ready-to-use format instead | 15:27.49 |
| the cmap data is in a fairly compact format though | 15:28.00 |
paulgardiner | tor8: Good. Means I don't need to worry about it either. | 15:28.20 |
henrys | paulgardiner:I am not following why these things can't be implemented wihout looking at the adobe file. Everything looks fairly trivial. | 15:30.09 |
Robin_Watts | henrys: The 9 lines of java code that google just got sued by Oracle over looked fairly trivial too :) | 15:30.54 |
paulgardiner | henrys: true, although the only description of them I have at the moment is the source. | 15:30.56 |
Robin_Watts | I can't find these functions in the chromium source. | 15:31.22 |
henrys | maybe I don't understand something we have a test file that say uses the ssn regexp and right now that is undefined. So to fix it you have to add a regexp for an ssn, right? | 15:33.47 |
Robin_Watts | henrys: http://commondatastorage.googleapis.com/chromium-browser-snapshots/index.html?path=Win/144169/ | 15:35.11 |
| paulgardiner, not henrys, sorry. | 15:35.24 |
henrys | I just thought we'd do some expected implementation as we trip over things that break in test files. | 15:35.43 |
ezequiel_ | Hello i'm having a problem using /D parameter in windows installer of ghostscipt | 15:36.37 |
Robin_Watts | oh. chromium doesn't have PDF viewing. | 15:36.44 |
ezequiel_ | anyone know how to use it | 15:36.49 |
Robin_Watts | henrys: Right. But how do we know what the expected implementation is? | 15:37.58 |
chrisl | ezequiel_: it's more than likely we don't implement /D (yet) on the installer | 15:39.51 |
henrys | I'm probably confused, I see a need for an SSN regexp? We know how to do that. | 15:40.22 |
Robin_Watts | Do we? (I haven't a clue what an SSN regexp is, personally) | 15:40.56 |
| If there was a functional spec we could refer to, we could do our own implementation. | 15:41.25 |
paulgardiner | Presumably these functions are described somewhere other than by the source. | 15:42.48 |
Robin_Watts | If so, then, yes,we could do our own implementations of everything. | 15:43.06 |
| There is an acrobat javascript guide thing. | 15:43.17 |
henrys | we know an ssn is xxx-xx-xxxx then a little experimenting with chrome to see about the dashes, why a spec? | 15:44.16 |
Robin_Watts | ok, so for that example you might be OK. | 15:44.30 |
| but when we hit other stuff the time to reverse engineer may not be so trivial. | 15:45.01 |
| And there is always the risk we'll get something wrong. | 15:45.09 |
paulgardiner | AFNumber_Format is the one I've been seeing a lot | 15:45.22 |
| ... although I've just found a description of that. | 15:45.41 |
| http://www.planetpdf.com/forumarchive/125041.asp | 15:46.09 |
henrys | let's just deal with them as they come up... | 15:46.14 |
| anything else for this meeting? | 15:48.07 |
Robin_Watts | No. | 15:48.35 |
paulgardiner | Not from me, although I'm still concerned about AForm.js. I'd imagined it had issues, but hadn't expected we would have no way to use it. | 15:49.20 |
Robin_Watts | Item 9 will be done as soon as marcosw figured out what I've broken in the cluster stuff. | 15:49.57 |
| I believe the technical way stuff like aform.js is supposed to be done is a clean room implementation. | 15:50.36 |
paulgardiner | We can deal with the js functions as they come up, but I fear they will mostly come up in other people's tests when we release it. | 15:50.42 |
Robin_Watts | Someone takes the job of looking at aform.js and writes a spec for the functions. | 15:51.03 |
| someone else then has to implement the functions from that spec. | 15:51.14 |
paulgardiner | What about importing it during install from a known public repository, if we can find one? | 15:52.41 |
Robin_Watts | paulgardiner: Our customers are likely to be embedded customers. | 15:52.58 |
| hence no 'install' phase. | 15:53.04 |
paulgardiner | Oh yeah. Of course | 15:53.09 |
| Lukily, people that know how bad my memory is will be able to confirm that my implementing it is more or less clean room | 15:55.03 |
sebras | Robin_Watts: fwiw: http://bfo.com/products/report/docs/api/org/faceless/pdf2/JSCoreMethods.html | 15:57.51 |
| Robin_Watts: http://www.planetpdf.com/forumarchive/125041.asp | 15:58.45 |
Robin_Watts | sebras: Yeah, paulgardiner gave that latter link above. | 15:59.14 |
sebras | Robin_Watts: right, I missed that one. | 15:59.33 |
kens | Tuesday meeting time ? | 15:59.39 |
henrys | yes let me text ray though | 15:59.52 |
Robin_Watts | Your former link is interesting, but ignoring the fact that javascript != java (so there may be subtlties) it doesn't include all the functions/regexps in AForm.js | 16:00.15 |
henrys | was mvrhel out today? | 16:01.23 |
kens | Not that I recall | 16:01.34 |
sebras | Robin_Watts: no, I know. | 16:01.43 |
kens | But I could be mistaken easily | 16:01.43 |
henrys | so 2 cluster machines have been down forever it seems, marcosw? | 16:03.37 |
| I'll look at henrysx6 after the meeting. | 16:04.48 |
marcosw_ | henrys: yes, both are in Miles' office; I'm planning on going by this week and figuring out why. I think they may be overheating and just need a good cleaning. They are some of the original cluster nodes and have been running without service for several years. | 16:05.03 |
alexcher | marcosw_: Please find the revision that introduced the bug 69315. It happened around v. 8.57. | 16:06.31 |
henrys | anybody have anything interesting this week. | 16:06.56 |
| ? | 16:06.57 |
kens | Not me. | 16:07.08 |
henrys | our bug numbers creeped up a bit. | 16:07.10 |
marcosw_ | alexcher: what was the bug number? | 16:07.16 |
alexcher | marcosw_: 693155 | 16:07.48 |
mvrhel | sorry I am late | 16:08.20 |
ray_work | Sorry I'm late -- had a meeting at the school | 16:08.44 |
henrys | mvrhel:np | 16:08.45 |
mvrhel | but not as late as ray.. | 16:08.48 |
| I just signed on too ray_work | 16:08.56 |
marcosw_ | alexcher: will do. | 16:08.56 |
mvrhel | had to run my son to lacrosse camp | 16:09.11 |
Robin_Watts | I was going to ask alexcher if he'd made any progress on 693115 | 16:09.18 |
ray_work | henrys: BTW, I did get your text. works great! | 16:09.44 |
Robin_Watts | (in a nosey, kid in the back of the car being a pain kind of way) | 16:09.49 |
henrys | ray_work:you mean sms works? | 16:10.29 |
alexcher | Robin_Watts: gs miscalculates the length of the prefix. The PS routine that does it is big and hairy. | 16:10.34 |
ray_work | henrys: yep | 16:10.37 |
henrys | I find I text more than phone call these days. | 16:10.53 |
Robin_Watts | alexcher: prefix ? | 16:11.00 |
chrisl | henrys: I'm hopeful of having the contrib license stuff sorted out later this week - one way or another...... | 16:11.26 |
ray_work | alexcher: is that comment about 693115 ? | 16:11.26 |
Robin_Watts | no, bug 693115, not bug 693155 :) | 16:11.31 |
ray_work | chrisl: thanks for doing that. | 16:11.42 |
chrisl | ray_work: all part of the fun..... | 16:12.07 |
henrys | chrisl:good I hope we get rid of some drivers. Getting rid of the open vector mess would be great, it has dependency hell but last I asked tkamppeter it was being used. | 16:12.36 |
ray_work | speaking (tongue in cheek, also) of fun -- I have _still_ having laptop problems :-( | 16:12.55 |
alexcher | Robin_Watts: nothing that is worth reporting. | 16:13.02 |
chrisl | henrys: I believe it is still used, yes - in fact, I believe there is a OpenPrinting project related to it...... | 16:13.27 |
sebras | Robin_Watts: incidentally how does foxit handle the undocumented javascript things? | 16:13.50 |
Robin_Watts | alexcher: OK. Do you have a rough feeling for how long it will take given your other task load? | 16:13.53 |
ray_work | I thought the Open Printing folks were all distracted by color management | 16:13.59 |
Robin_Watts | sebras: I'm assuming foxit behaves a lot like chrome :) | 16:14.18 |
| ray_work: Have you not replaced your laptop yet? | 16:14.34 |
chrisl | ray_work: I think this is a Google Summer of Code project proposed by the OpenPrinting people | 16:14.47 |
ray_work | Robin_Watts: I bought a new laptop. My first attempt was to take the HD out of the old one and put it in the new one and update drivers | 16:15.10 |
henrys | has everybody gone through the open customer bugs would it be useful to go through them here? | 16:15.12 |
ray_work | Robin_Watts: it seemed fine for a while, then started with the sporadic crashes. | 16:15.46 |
alexcher | Robin_Watts: I'll try to give this bug more time. | 16:16.19 |
henrys | mvrhel:you're up to 4 last bug aging report. | 16:16.24 |
Robin_Watts | ray_work: so it's either a software thing or that HD is doing something nasty like sending power spikes etc. | 16:16.50 |
ray_work | Robin_Watts: so then I tried using 'PCmover Image Assistant' to transfer stuff over to the HD from the new laptop, and many apps work, but... | 16:16.52 |
Robin_Watts | henrys: I think mvrhel has given me one of them. | 16:17.39 |
mvrhel | henrys: I started going through these yesterday. I passed one off to Robin_Watts. I am hoping to have the copy_alpha stuff done soon but I have 3 files that still segv and they are ugly | 16:17.43 |
henrys | mvrhel:sigh | 16:17.58 |
Robin_Watts | alexcher: Thanks. I think fixing that will close 2 customer bugs. | 16:18.12 |
ray_work | Robin_Watts: VS Studio, while it was transferred, reports "invalid license". I uninstalled 2008 and reinstalled it, and despite no errors during the re-install, it still reports 'invalid license' | 16:18.14 |
mvrhel | the 90,000th call into clist_copy_alpha_hl_color has a valgrind warning | 16:18.22 |
henrys | what kind of warning? | 16:18.51 |
| some are more interesting than others. | 16:19.07 |
ray_work | Robin_Watts: and it won't even let me uninstall VS 2005 :-( So now I have to "undo" the PCmover action (it has that) and try it again without letting it move VS | 16:19.11 |
mvrhel | syscall param write(buf) points to unititialised bytes(s) | 16:19.34 |
henrys | mvrhel:I see that from time to time. | 16:19.55 |
Robin_Watts | mvrhel: That need not be fatal. | 16:19.56 |
mvrhel | it occurs during the cmd_put_bits | 16:19.56 |
| oh | 16:20.01 |
ray_work | mvrhel: hmm... I wonder if it is complaining about 'pad' in a bitmap | 16:20.11 |
Robin_Watts | If you are writing some stuff to the clist, and you have padding in there, that padding can be uninitialised. | 16:20.19 |
mvrhel | ok that could be | 16:20.28 |
ray_work | Robin_Watts: right | 16:20.30 |
henrys | mvrhel:a public branch and other can give it a go? | 16:20.39 |
mvrhel | maybe | 16:20.45 |
henrys | s/other/others | 16:20.47 |
mvrhel | I cant even get it to segv for me | 16:20.52 |
| it segvs on the cluster though | 16:20.58 |
Robin_Watts | mvrhel: It may be a 64bit linux only thing. | 16:21.08 |
mvrhel | right | 16:21.13 |
kens | are you on WIndows or Linux mvrhel2 ? | 16:21.16 |
mvrhel | both | 16:21.19 |
marcosw_ | mvrhel: does it seg fault reliably or only on some nodes? | 16:21.24 |
Robin_Watts | And maybe I can spend some time and try and reproduce it for you. | 16:21.24 |
ray_work | mvrhel: have you tried a private build on peeves ? (you should have ssh access) | 16:21.25 |
henrys | oh you are still 32 bits? | 16:21.31 |
ray_work | mvrhel gets flooded with things to try | 16:21.58 |
Robin_Watts | Our windows builds tend to be 32bit ones, even on 64bit machines. | 16:22.11 |
mvrhel | the linux build is 32 bit. I do both on windows. I had not tried the 64 bit build though | 16:22.13 |
| Robin_Watts: right | 16:22.24 |
ray_work | mvrhel: the cluster build defaults to 64-bits | 16:22.31 |
Robin_Watts | mvrhel: How many colorants do we need to support? | 16:22.34 |
mvrhel | let me make a 64 bit build in windows | 16:22.36 |
| Robin_Watts: as many as the customer needs | 16:22.47 |
Robin_Watts | is 32 enough? Or 64 ? | 16:22.53 |
| or more ? | 16:22.55 |
mvrhel | Robin_Watts: for some reason Thomas wants to have 64 | 16:23.32 |
henrys | mvrhel:send your patch to tech, I doubt we'll see anything but we could get lucky and you just overlooked something. | 16:23.37 |
mvrhel | not sure what he is doing | 16:23.38 |
ray_work | Robin_Watts: the number of colorants needs to be a customer changeable option. Some customers need (and our competitors can do) 250+ | 16:23.55 |
mvrhel | right | 16:24.03 |
Robin_Watts | 64 may be doable with some minor tweaking. | 16:24.11 |
| More than that, and we hit problems. | 16:24.18 |
ray_work | Robin_Watts: The default build was 14 colorants (MAX_CLIENT_COLORS) | 16:24.33 |
Robin_Watts | pdf14 uses a bitfield for which componets are used. | 16:24.35 |
mvrhel | that is used in overprint too | 16:24.48 |
Robin_Watts | so anything > 32 goes foom. | 16:24.48 |
| And there is clist stuff in there that breaks with > 32 | 16:25.07 |
mvrhel | actually overprint uses a 64 bit word | 16:25.07 |
ray_work | Robin_Watts: is it more than 32 in the current colorspace, or 32 in the entire page ? | 16:25.26 |
kens | maximum DeviceN components in PDF is 32 | 16:25.32 |
mvrhel | the page | 16:25.35 |
alexcher | The number of components was reduced to improve the performance. | 16:25.50 |
mvrhel | kens that is in one color space though | 16:25.58 |
kens | Yes true | 16:26.05 |
ray_work | kens: yes, that was what I was asking -- each DeviceN space has a max of 32, but since you can have many different DeviceN spaces, the total in the page can be more | 16:26.30 |
kens | valid range of MaxSeparations is 0 to 250 | 16:26.43 |
mvrhel | 0? | 16:26.55 |
kens | Can't do separations :-) | 16:27.03 |
| But actually uits a typo should be 1 | 16:27.21 |
ray_work | alexcher: the performance hit was (mostly) the 'image_init_clues' that had a large structure allocated and filled for all of the MAX_CLIENT_COLORS. | 16:27.38 |
| mvrhel: didn't you get rid of the 'clues' ? | 16:27.53 |
mvrhel | it only exists in the mono image case now | 16:28.03 |
ray_work | kens: CPSI supports 252 colors | 16:28.21 |
kens | ray_work : then it doesn't conform to the spec :-) | 16:28.42 |
mvrhel | I suppose that is ps not pdf | 16:28.54 |
ray_work | mvrhel: (alexcher), right so the performance hit for a larger MAX_CLIENT_COLORS should be (famous last words) resolved | 16:29.04 |
kens | Yes, that is PostScript | 16:29.05 |
| I would be astonished if anyone used more though | 16:29.17 |
ray_work | kens: the CET tries to :-) | 16:29.37 |
mvrhel | the only case we have seen with the large number is with someone printing a swatch book | 16:29.45 |
kens | Presumably in order to provoke an error, the CET tries to do lots of illegal stuff | 16:29.58 |
ray_work | mvrhel: yes, that's the usual case for lots of colorants. And those don't use transparency | 16:30.20 |
Robin_Watts | At the moment, the code assumes that you can pack colors into a color index. It assumes a minimum of 1 bit per component, thus is limited to as many components per page as there are bits in a byte. | 16:30.29 |
| I can tweak the code to avoid that limit, but then stuff starts falling over all around it. | 16:30.51 |
alexcher | kens: I've seen a spot color sampler printed on a thermal transfer printer. | 16:30.52 |
mvrhel | we did have the one file running around here that did have a black overprint and a bunch of swatches let find that final | 16:31.02 |
ray_work | kens: the CET does 'range' checking, so depending on what the device reports for MaxSeparations, it tries a DeviceN space with those | 16:31.09 |
mvrhel | acrobat had trouble with it | 16:31.10 |
kens | merely reports what the PS specification says | 16:31.36 |
henrys | marcosw_:are we sill doing the performance monitoring? I've never seen a commit cause a problem but we have had performance regressions. | 16:31.48 |
marcosw_ | henrys: we are logging performance on a nightly basis but I'm not automatically sending out emails. I should have enabled that a while ago but need did. As soon as miles is back up I'll turn it on. | 16:33.10 |
Robin_Watts | marcosw: Did you have a chance to look at my mujstest stuff ? | 16:34.42 |
| mvrhel: I have a test .ps file that I can use to get me as many colorants as I want. | 16:35.00 |
mvrhel | Robin_Watts: I was just looking for this particular file that came from a customer | 16:35.25 |
| cant seem to find it now | 16:35.32 |
| I was curious what it had | 16:35.37 |
| Robin_Watts: so why when I select x64 in visual studio it does not build | 16:36.27 |
marcosw_ | Robin_Watts: sorry no, I was out much of yesterday and every time I had an opportunity the cluster was in use. We need a second cluster to test cluster code... | 16:36.28 |
mvrhel | is the project not set up for this? | 16:36.38 |
Robin_Watts | mvrhel: Do you have the 64bit compilers installed? | 16:36.43 |
mvrhel | good question | 16:36.55 |
| I would assume they got installed | 16:37.05 |
Robin_Watts | They don't install as standard. | 16:37.13 |
| (IIRC) | 16:37.17 |
mvrhel | great | 16:37.19 |
Robin_Watts | You need to install a platform sdk or something, maybe. | 16:37.26 |
ray_work | mvrhel: 64 bit compilers aren't part of all VS versions, and they don't get installed by default | 16:37.32 |
| mvrhel: you need the "Professional" edition to build 64-bit | 16:38.28 |
mvrhel | I have our Team edition | 16:38.48 |
chrisl | Robin_Watts, mvrhel: also I don't think the project file sets the BUILD_SYSTEM parameter for nmake - since it doesn't have any good way to know it | 16:39.08 |
ray_work | mvrhel: I don't think the 2008 Team edition has 64-bit (my installation doesn't) | 16:39.32 |
Robin_Watts | I have 64 bit install on VS2005 here, but it was an extra download job. | 16:39.52 |
ray_work | mvrhel: I can only build 64-bit with VS 2005 Professional | 16:39.58 |
Robin_Watts | You can download the windows platform SDK to get 64bit compilers added into VS2008 team edition. | 16:40.27 |
mvrhel | that is what I want to do | 16:40.36 |
| just trying to google that | 16:40.44 |
henrys | marcosw:actually for performance it seems the cluster report is pretty stable. It would be nice if you calculated the serial time and we just checked that number each commit for something weird. | 16:40.55 |
ray_work | Robin_Watts: thanks. once I get my new laptop back to working, I'll do that | 16:41.01 |
| mvrhel: (or Robin_Watts) if you know the download URL, can you paste it here ? | 16:41.36 |
Robin_Watts | I believe: http://www.microsoft.com/en-us/download/details.aspx?id=3138 | 16:42.13 |
ray_work | Robin_Watts: thanks | 16:42.21 |
mvrhel | thats where I am | 16:42.26 |
| now | 16:42.28 |
| I thought I did this before though | 16:42.35 |
chrisl | marcosw: This isn't urgent, but would it be possible to have the cluster record parallel build fails, and the error message (even if it just e-mailed them to me for a while)? | 16:42.59 |
marcosw_ | henrys: by serial time do you mean the total time to run all the jobs? | 16:43.08 |
| minus the compile times. | 16:43.23 |
ray_work | hmm... that's interesting. That download URL says "This SDK is compatible with Visual Studio® 2008, including Visual Studio Express Editions, which are available free of charge. " | 16:43.39 |
henrys | marcosw_:yes. | 16:43.58 |
Robin_Watts | ray_work: Indeed. | 16:44.18 |
henrys | I don't care about the compile times. It is the relative number we're interested in. | 16:44.41 |
Robin_Watts | It's always been completely free to build 32 and 64bit native apps on windows. Until windows 8 when they were going to nobble you to only be able to build metro apps. | 16:44.59 |
| but they backtracked on that after outcry. | 16:45.10 |
ray_work | Robin_Watts: I don't see that it mentions 64-bit anywhere | 16:45.15 |
marcosw_ | chrisl: it should be easy to do that, I think a grep statement through the node output files should be enough. I can add it to the cluster email. | 16:45.24 |
ray_work | Robin_Watts: other than the warning to upgrade to SP1 "if you are running a 64-bit version of Windows and Visual Studio 2008," | 16:45.59 |
Robin_Watts | right. but references through which I found that page suggest it includes 64bit compilers. | 16:46.20 |
chrisl | marcosw_: that would be good - I *think* the last couple of fixes should have resolved the parallel build problems, but it *very* hard to tell..... | 16:46.22 |
Robin_Watts | I'm confused. arch.h says: #define ARCH_SIZEOF_GX_COLOR_INDEX 8 | 16:46.59 |
marcosw_ | chrisl: I now, the parallel build jobs tend to be indeterministic based on what else the computer is doing at the time. | 16:47.01 |
| ^now^know | 16:47.08 |
Robin_Watts | and yet: gx_color_index_data is a ulong | 16:47.26 |
ray_work | Robin_Watts: right. color_index is 64-bit (even on 32-bit platforms) | 16:47.30 |
chrisl | mvrhel: you could just invoke the nmake build from the VS command prompt and that will show up pretty sharp whether you have 64 bit tools installed..... | 16:47.39 |
Robin_Watts | ray_work: How so? | 16:48.16 |
ray_work | mvrhel: or see if you have the 'amd64' folders | 16:48.19 |
chrisl | marcosw_: I ran 12 process parallel builds on my own i7 machine for about 1.5 hours, and never got a failure (before the fixes, the longest I had to wait was about 15 minutes), so....... | 16:49.09 |
marcosw_ | chrisl: that's encouraging | 16:49.36 |
henrys | marcosw_:I was thinking of something as trivial as just multiplying nodes by seconds - just so we have a fixed number that we get accustomed to seeing. | 16:49.42 |
ray_work | Robin_Watts: gx_color_index_data is a *gx_color_index so in a 32-bit build pointers are 32-bits | 16:49.54 |
Robin_Watts | typedef ulong gx_color_index_data; | 16:50.13 |
| I see no pointer. | 16:50.33 |
ray_work | Robin_Watts: I was looking in gxcindex.h: typedef gx_color_index_data * gx_color_index; | 16:51.12 |
Robin_Watts | ray_work: Only if TEXT_CINDEX_POINTER is defined. | 16:51.58 |
| ray_work: Only if TEST_CINDEX_POINTER is defined. | 16:52.02 |
ray_work | Robin_Watts: sorry. | 16:52.07 |
| lost in #ifdef forest | 16:52.21 |
henrys | we are way past meeting end, but I wanted to ask tor8 if he has thought more about the mupdf/pdfwrite viewer idea? | 16:53.18 |
marcosw_ | henrys: I think that would be too rough a number, adding up all the user times for the jobs should be more reliable and almost as easy. I'll report it for clusterpush jobs as well, but it won't be useful for jobs that specify lowres or a submit of languages. | 16:54.08 |
henrys | okay | 16:54.48 |
kens | How owuld MuPDF/pdfwrite work ? Is this after we've moved MuPDF to be the GS PDF interpreter ? | 16:55.07 |
| Oh, I have to go, goodnight all | 16:56.23 |
marcosw_ | brb | 16:56.25 |
henrys | tor8:is enhancing the mupdf viewer and we would bolt on pdfwrite and presumably view postscript/pcl and friends much in the style of acrobat or preview. | 16:56.25 |
kens | henrys by using pdfwtite to turn everything into a PDF as a separate prcoess ? | 16:57.02 |
henrys | yes | 16:57.12 |
kens | Slow, but no reason why it wouldn't work | 16:57.21 |
henrys | or thread. | 16:57.22 |
kens | Could use the new %d syntax to make multiple PDFs one per page for faster uinitial presentation | 16:57.41 |
henrys | good point | 16:57.55 |
kens | I really do have to run now, by all | 16:58.08 |
henrys | bye kens | 16:58.18 |
ray_work | My laptop says it "only" has 2.5 hours remaining on the "undo" of the PCmover cock up | 16:59.15 |
henrys | I've gone to mac store and looked at the retinal displays ... tempting, but I just got a new laptop. | 17:00.21 |
ray_work | then I can re-try installing VS versions, then start the 6 hour process of moving everything else again (except Adobe CS apps that don't move) | 17:00.27 |
| henrys: can you really tell the difference ? (without a magnifying glass) | 17:00.52 |
Robin_Watts | ray_work: Why not just reinstall? Why risk registry corruption by some dodgy tool? | 17:00.57 |
henrys | ray_work:yes very sharp and crisp compared with macbook pro - but you said you had a high res screen I don't know if I could tell compared with that. | 17:02.48 |
ray_work | Robin_Watts: I've got lots of apps that I'd have to hunt up the discs or re-download, _and_ find the email with the license info. That could be many times more painful | 17:02.54 |
henrys | s/macbook pro/my current macbook pro/ | 17:03.12 |
ray_work | henrys: well if marcosw gets one, we can compare to mine at the staff meeting. | 17:03.27 |
Robin_Watts | ray_work: much more painful than it would be if the "pc mover" stuff actually worked, maybe :) | 17:03.54 |
ray_work | and once I get the new laptop up and stable, I can send the old Dell out for repair (heating issue) and bring that to the staff meeting for whoever might want it | 17:04.38 |
| Robin_Watts: it warns that some apps may have license keys it can't handle, so if I don't let it try VS I _should_ be OK. Everything else worked fine. | 17:05.49 |
| Robin_Watts: but if this next try has problems, I'll resort to the reinstall the world approach (and just xfer the data over) | 17:07.13 |
Robin_Watts | ray_work: I find I just install stuff as I need it. It means I keep the amount of crap down on each new machine... | 17:08.06 |
ray_work | at least the new system hasn't crashed or hung since the PC mover (but it's only been since Sat pm) | 17:08.08 |
Robin_Watts | mvrhel: If you want me to try and reproduce the crash on 64 bit linux, please say. | 17:08.45 |
ray_work | Robin_Watts: I did leave some apps that I had installed on the old machine behind -- I was selective, just not selective enough :-/ | 17:09.00 |
mvrhel | ok. I am trying to figure out this 64 bit stuff in windoze | 17:09.05 |
Robin_Watts | There is no guarantee that it will crash under windows though. | 17:09.21 |
| long is 64bits on linux, and only 32 on windows (even in 64 bit mode) | 17:09.39 |
ray_work | mvrhel: doing the build ? should just be able to nmake WIN64=1 | 17:09.43 |
mvrhel | I want to build if from the solution | 17:10.11 |
| s/if/it/ | 17:10.16 |
Robin_Watts | The solution does nmake WIN64=1 | 17:10.27 |
ray_work | mvrhel: and you have to hope that it fails with a debug build instead of the release build the clusters run | 17:10.36 |
mvrhel | well it does not work here | 17:10.39 |
| that is what I am trying to do | 17:10.42 |
ray_work | mvrhel: what error message do you get ? | 17:11.07 |
Robin_Watts | right. there is voodoo in the makefiles to find the correct installation of x64 tools. | 17:11.15 |
mvrhel | my sdk is in program files and the visual studio is in program files x86 | 17:11.40 |
| i am wondering if that is a problem | 17:11.48 |
Robin_Watts | mvrhel: Do not try moving it by hand! | 17:12.02 |
mvrhel | ha | 17:12.05 |
| I am not going to do that | 17:12.10 |
Robin_Watts | because windows will probably zap it off into a virtual path somewhere. | 17:12.22 |
mvrhel | I am thinking that when I installed the sdk way back when I installed the 64 bit version | 17:12.32 |
ray_work | Robin_Watts: since I often do command line nmake, I usually specify MSVC_VERSION=__ | 17:12.32 |
mvrhel | but I had the 32 bit version of our team edition installed | 17:12.43 |
| this disconnect between the two may be my problem | 17:13.57 |
ray_work | I have to go now... | 17:13.59 |
alexcher | ray_work: is your platforn ppc32 or ppc64 ? | 17:15.09 |
Robin_Watts | mvrhel: I don't think that should be the issue. | 17:15.37 |
mvrhel | what is interesting is that if I try to do nmake -f psi/msvc32.mak MSVC_VERSION=9 from the command line It tries to run from "C:\Program Files\Microsoft Visual Studioa 9.0\ but it is actually in the x86 folder | 17:15.47 |
| so something is amiss | 17:15.58 |
Robin_Watts | Have you closed and reopened visual studio since you installed the x64 stuff ? | 17:16.05 |
mvrhel | Robin_Watts: I have not installed the x64 stuff as it was already installed | 17:16.20 |
| or rather reinstalled | 17:16.25 |
Robin_Watts | mvrhel: junction ? | 17:16.31 |
mvrhel | I had done this some time ago | 17:16.34 |
chrisl | mvrhel: nmake -f psi/msvc.mak MSVC_VERSION=9 WIN64=1 BUILD_SYSTEM=64 | 17:16.57 |
Robin_Watts | mvrhel: http://technet.microsoft.com/en-us/sysinternals/bb896768 Download junction from there. | 17:17.29 |
| junction gives you symbolic links on NTFS. | 17:17.43 |
mvrhel | chrisl: even the 32 bit one wont build for me from the command line. I doubt the 64 bit will.. | 17:17.50 |
Robin_Watts | So you can link C:\program files\microsoft Visual Studio 9.0\ to C:\Program Files (x86)\microsoft Visual Studio 9.0 | 17:18.29 |
| (or whatever you need to do) | 17:18.34 |
mvrhel | oh interesting | 17:18.42 |
chrisl | Which is version 9 in MS terms? | 17:18.49 |
Robin_Watts | 9 is 2008 | 17:18.56 |
mvrhel | 2008 it is our team edition | 17:18.59 |
chrisl | OKay, I've built with that recently, although I don't think it was the team edition..... | 17:19.22 |
mvrhel | let me give that a go Robin_Watts | 17:19.29 |
marcosw | I'm going to take the cluster down for a bit for mujstest testing. | 17:19.31 |
mvrhel | question for you though | 17:19.43 |
chrisl | mvrhel: are you using a "vanilla" command prompt window, or the VS specific one? | 17:20.17 |
mvrhel | VS specified one | 17:20.32 |
Robin_Watts | marcosw: Could I beg a favour of you please? | 17:21.12 |
| marcosw: I want to get a Canon 650D and lens, but it will cost almost as much in pounds as in dollars. | 17:21.36 |
| So i'm considering ordering it from the states, and getting it delivered to some kind soul for them to bring to the september meeting. | 17:22.00 |
| and you are the only one that wouldn't have to struggle to bring it hand luggage... | 17:22.36 |
marcosw | Robin_Watts: no problem. Do you need me to order it for you or will they take a UK credit card? | 17:22.40 |
chrisl | mvrhel: this works for me (for 32 bit): nmake -f psi/msvc.mak BUILD_SYSTEM=64 | 17:22.46 |
Robin_Watts | I have a US $ card which I have successfully used in the past. | 17:23.04 |
mvrhel | ok chrisl, it looks like inside C:\Program Files (x86)\Microsoft Visual Studio 9.0\VC I don't have the 64 bit stuff | 17:23.05 |
| I need to get that in place | 17:23.13 |
Robin_Watts | So I'll try to order it, and if it's a problem, I'll come back to you. | 17:23.33 |
| Thanks! | 17:23.39 |
mvrhel | so, let me install the sdk, but the 32 bit version | 17:24.02 |
marcosw | Robin_Watts: okay. you have my address? | 17:25.53 |
Robin_Watts | yeah, I've ordered to you before - thanks. | 17:26.16 |
mvrhel | ugh. this is 32 bit vs 64 bit application stuff is a mess | 17:26.26 |
| it wants to install the x64 version of the sdk | 17:26.55 |
| which it recognizes is already installed | 17:27.09 |
| but how do I get the 64 bit sdk to get my 32 bit visual studio to have the 64bit stuff | 17:27.57 |
Robin_Watts | mvrhel: It's probably a question of tweaking paths. | 17:28.21 |
| I'd try the junction trick. | 17:28.47 |
mvrhel | Robin_Watts: so for example, I dont have the AMD folder in my visual studio folders which are in x86 | 17:29.02 |
| how do those get added with this sdk? | 17:29.12 |
Robin_Watts | VS has a series of vcvars32.bat things that set the paths. | 17:29.47 |
| And when you open the VS command line window, one of those has been run. | 17:30.02 |
| All they do is add some stuff into the path. | 17:30.12 |
chrisl_away | mvrhel: IIRC those go in a different directory - the SDK installs it's own command prompt with different paths from the VS one, maybe that might work? | 17:30.15 |
Robin_Watts | Now, the makefile does some rudimentary searching for what versions of VS are installed, and chooses paths appropriately. | 17:30.43 |
| so either you need to look at the makefile and try and understand the rats nest of searching stuff in there... | 17:31.26 |
marcosw | Robin_Watts: I'm testing the mujstest code and your current clusterpush mupdf source doesn't compile. This is is the error: | 17:31.26 |
| CC build/debug/doc_interactive.o | 17:31.27 |
| fitz/doc_interactive.c:9: error: expected ')' before '*' token | 17:31.27 |
| fitz/doc_interactive.c:14: error: expected ')' before '*' token | 17:31.27 |
| fitz/doc_interactive.c:19: error: expected '=', ',', ';', 'asm' or '__attribute__' before '*' token | 17:31.27 |
| make: *** [build/debug/doc_interactive.o] Error 1 | 17:31.27 |
mvrhel | chrisl_away: for example, right now when I run the command you gave me, it tries to use C:\Program Files (x86)\Microsoft Visual Stud | 17:31.44 |
| io 9.0\VC\lib\amd64 as a lib path | 17:31.46 |
| and "C:\Program Files (x86)\Microsoft Visual Studio 9. | 17:32.28 |
| 0\VC\PlatformSDK\Lib\AMD64 | 17:32.29 |
Robin_Watts | does that path exist in C:\Program Files\ ? | 17:32.35 |
mvrhel | no | 17:33.15 |
Robin_Watts | Just looking in gs/psi/msvc.mak | 17:34.58 |
| So, it's decided that BUILD_SYSTEM=64 | 17:35.28 |
| You do have visual studio in that path, just not the amd64 stuff ? | 17:35.45 |
mvrhel | visual studio is in the x86 path | 17:36.05 |
| but no amd stuff is in there | 17:36.14 |
| which is the 64 bit compiler stuff we talked about | 17:36.24 |
| when I try to install the sdk from the website it wants to put in the x64 version | 17:36.49 |
Robin_Watts | OK... so what nmake command are you currently using? | 17:37.53 |
mvrhel | the one that chrisl gave me above | 17:38.07 |
| nmake -f psi/msvc.mak MSVC_VERSION=9 WIN64=1 BUILD_SYSTEM=64 | 17:38.20 |
Robin_Watts | I think the makefile is broken in this case. | 17:39.17 |
mvrhel | chrisl_away: said it built fine for him | 17:40.42 |
Robin_Watts | mvrhel: What path does the platform SDK live in? | 17:41.19 |
mvrhel | Robin_Watts: C:\Program Files\Microsoft SDKs\Windows\v7.1 | 17:41.48 |
Robin_Watts | try DEVSTUDIO="C:\Program Files\Microsoft SDKs\Windows\v7.1" | 17:41.55 |
| as well as all chrisl's stuff. | 17:42.08 |
| According to the notes in the makefile v6.0A is Visual Studio 2008. | 17:42.45 |
| http://www.microsoft.com/en-us/download/details.aspx?id=23719 <- That may have been a better bet to install. | 17:43.55 |
| marcosw: sorry, let me look. | 17:45.20 |
mvrhel | actually it looks like something is screwy with my install of the sdk. It did not configure itself properly it appears and install the amd stuff | 17:45.34 |
| let me resinstall | 17:45.40 |
Robin_Watts | mvrhel: Can I suggest that you uninstall that stuff, and install from the link I gave just now? | 17:46.34 |
| It looks like ray made the makefiles work with the v6.0a sdk, so you may find it easier using the same version he used. | 17:46.59 |
mvrhel | oh | 17:47.09 |
| ok | 17:47.11 |
| I cringe when I see the word vista though.... | 17:47.42 |
Robin_Watts | marcosw: The version I pushed works fine locally. | 17:47.53 |
| or at least, what I have seems to work. | 17:48.20 |
| should I repush? | 17:48.27 |
marcosw | let's try that. | 17:49.52 |
Robin_Watts | marcosw: Just trying to build on peeves. | 17:50.00 |
| Builds fine on peeves. | 17:50.08 |
| pushing now. | 17:50.19 |
mvrhel | Robin_Watts: when I do make on linux is it doing a 64 bit or 32 bit build? | 17:51.37 |
Robin_Watts | Depends on what linux box. | 17:51.57 |
| On peeves, (or any other cluster node) 64bit. | 17:52.09 |
mvrhel | I am running 64 bit Ubuntu through vmware | 17:52.14 |
Robin_Watts | Then 64bit | 17:52.21 |
| My local VMware image is 32bit, deliberately. | 17:52.49 |
mvrhel | ok. I guess I need to get the whole command line for this thing then to see if I can get it to crash | 17:53.12 |
marcosw | Robin_Watts: the code now compiles, running a cluster job now. | 17:55.32 |
Robin_Watts | marcosw: This is where casper slows to a crawl as all the cluster nodes grab a 50Meg file at once. | 17:56.24 |
| and maybe checkout 2Gig more test fils. | 17:56.39 |
mvrhel | Robin_Watts: can I send you a patch and a file to see if you get a segv? | 17:56.52 |
Robin_Watts | Sure. | 17:57.10 |
marcosw | Robin_Watts: The mujstest compile failed on macpro: | 17:59.39 |
| ld: warning: directory 'thirdparty/v8-3.9//out/x64.debug/obj.target/tools/gyp' following -L not found | 17:59.45 |
| Undefined symbols: | 17:59.45 |
| "_pdf_jsimp_new_type_cpp", referenced from: | 17:59.45 |
| _pdf_jsimp_new_type in libfitzv8.a(pdf_jsimp_cpp.o) | 17:59.45 |
| "_pdf_jsimp_toString_cpp", referenced from: | 17:59.45 |
| ... | 17:59.45 |
mvrhel | brb | 17:59.52 |
Robin_Watts | marcosw: Ass, yes, it would. | 18:00.00 |
marcosw | I'll down macpro and try again | 18:00.10 |
Robin_Watts | It needs to look in out-mac/ | 18:00.14 |
marcosw | so the mujstest code uses precompiled code? | 18:01.42 |
Robin_Watts | Yes. The v8 lib is downloaded precompiled. | 18:02.22 |
| ok, my craptastic test shows I have 64 colorants working here (CMYK + 60 others) | 18:06.19 |
mvrhel | ok and with that 64 bit windoze is building | 18:06.30 |
Robin_Watts | OK. I should add a note to the Win32/ReadMe.txt to suggest that VS2008 users use that version of the platform SDK. | 18:07.05 |
mvrhel | Robin_Watts: so I reinstalled the sdk and ran the little config tool and now visual studio runs ok | 18:07.18 |
| Robin_Watts: no I did not do the thing that you told me to do | 18:07.27 |
| I got the latest version | 18:07.32 |
| 7.1 | 18:07.37 |
Robin_Watts | oh, fair enough. | 18:07.39 |
| So, this colorant thing... | 18:07.52 |
mvrhel | I don't know what the issue was | 18:07.54 |
Robin_Watts | I have some tweaks here that make it work with 64 colorants. | 18:08.14 |
mvrhel | great | 18:08.22 |
Robin_Watts | If I try and use 65 colorants, we have problems. | 18:08.25 |
mvrhel | I say for now we limit it to 64. when someone comes forth with needing more we tackle that issue then | 18:08.54 |
Robin_Watts | because comp_bit_map_t becomes an array of things rather than a single thing. | 18:08.56 |
| OK. | 18:09.00 |
mvrhel | which may be to rehook in compressed color encoding | 18:09.06 |
Robin_Watts | compressed color won't get down to less than 1bit per component will it ? | 18:09.31 |
mvrhel | less than 1 bit? | 18:09.49 |
| oh | 18:09.56 |
| well if I have a swatch book with 2000 colors all with solid fills | 18:10.12 |
| I should be fine | 18:10.16 |
Robin_Watts | OK. | 18:10.20 |
mvrhel | right? | 18:10.27 |
| if I have only solid fills I can compress all day long | 18:10.46 |
| that is the general case, where we have had to do this before | 18:11.14 |
Robin_Watts | that seems reasonable, yes. | 18:12.10 |
| mvrhel: I've seen no patch so far... | 18:21.02 |
mvrhel | Robin_Watts: yes. hold on sec | 18:21.15 |
Robin_Watts | sorry, no hurry. | 18:21.25 |
| just didn't want to have missed it :) | 18:21.31 |
mvrhel | ok email sent | 18:23.26 |
| I am seeing what I get in windows 64 bit | 18:23.40 |
Robin_Watts | building now. | 18:30.22 |
marcosw | Robin_Watts: I fixed the code in run.pl that detects the operating system, so now macpro rsync the correct v8 file, but the link still attempts to use x64.debug. | 18:30.28 |
Robin_Watts | marcosw: Let me look at the makefile here... | 18:30.48 |
marcosw | if I down macpro the mujstest runs to completion but produces no output, is that expected? | 18:30.55 |
Robin_Watts | marcosw: Urm... no. | 18:34.58 |
| What should happen is that each job causes a mujstest-v8 job to run. | 18:35.46 |
| That job reads from a .mjs file and outputs bitmaps to the -o path given. | 18:36.07 |
| I was expecting the cluster to md5 those etc and compare them to the last one. | 18:36.29 |
| I have a new version here that should link properly on the mac. Should I clusterpush it? | 18:36.48 |
mvrhel | interesting. in windows 64, we dont segv but the psdcmyk file is not correct | 18:46.36 |
| photoshop cant open it | 18:46.47 |
| that is something I can debug | 18:47.01 |
| off to get my son | 18:47.09 |
| bbiab | 18:47.11 |
henrys | I went to fix henrysx6 and see it is working now, don't know what went wrong | 18:51.44 |
Robin_Watts | mvrhel: It SEGVs for me on peeves. | 18:52.22 |
| Trying a debug build now. | 18:52.36 |
| I get a wierd message after linking; apparently gdevtsep.c contains 2 calls to memset with constant 0 length. | 18:53.04 |
henrys | marcosw:still here? | 18:57.34 |
Robin_Watts | no segv in release mode :( | 18:57.56 |
| in debug mode I mean. | 18:58.02 |
mvrhel | Robin_Watts: that stinks. maybe this issue that I am seeing in 64 bit windows is releated | 19:16.28 |
| lunch time now | 19:16.33 |
| bbiab | 19:16.35 |
Robin_Watts | mvrhel: Let me know if you're still stuck when you quit today, and I'll look some more. | 19:16.57 |
| I'm about to give up for the night. | 19:17.02 |
| Do you have any other bugs you'd like me to look into? | 19:17.29 |
| (I'm relatively free at the moment) | 19:17.58 |
myself_ | Just trying to figure out why a certain pdf is printing wonky | 19:18.36 |
| so I'm using gswin64c to print some pdfs, most of them are fine, but any pdfs generated by QuickBooks print very oddly | 19:21.51 |
| I have tried adding papersize, paperorientation, and pdffitpage switches, doens't affect how they print | 19:22.27 |
Robin_Watts | myself_: In the absence of any better solutions being offered here, go to bugs.ghostscript.com, make a new bug, and attach an example file. | 19:22.28 |
| Unless anyone here recognises a particular issue with quickbooks (unlikely) we can't really comment unless you give us a file to look at. | 19:23.10 |
myself_ | alrighty | 19:23.13 |
| I'll post one of the .pdfs | 19:23.36 |
| Posted, my assumption is i'm doing something wrong. | 19:31.49 |
ray_work | myself_: what is the bug number ? | 19:46.13 |
myself_ | Bug no 693157 | 19:49.36 |
marcosw_ | I've reenabled the cluster. | 19:50.38 |
tkamppeter | henrys, which drivers are about to be dropped? | 19:52.02 |
Robin_Watts | marcosw_: Is there anything stopping me doing mujstest runs now ? | 19:52.25 |
Robin_Watts | is being called for food. | 19:52.44 |
mvrhel | thanks Robin_Watts | 19:57.45 |
henrys | tkamppeter:chrisl will have the complete list... but any device that does not have a proper GPL license has to go. chrisl has sent emails out and is waiting to hear back from a few. | 20:01.23 |
Ch3rryC0ke | Hey there-- I'm trying to make a win8 metro app version of mupdf, and have a quick Q-- in metro apps, you can't use the standard _wopen calls to open files | 20:43.12 |
| So I need to open the file using managed code, then pass the buffer to mupdf-- can I just create a fz_stream object and have the fz_stream->state point to the binary data of the file that is passed in from managed code | 20:44.42 |
| ? | 20:44.52 |
mvrhel | Robin_Watts: so I should be able to figure this issue out in windows. the 64 bit debug version is only creating a portion of the psdcmyk file. it doesn't segv or error out but clearly something is not right | 21:49.07 |
marcosw_ | Robin_Watts: I was out, but you can do mujstest runs if you'd like. | 21:51.19 |
ray_work | marcosw: about 'miles' and 'kilometers' you wrote: I think they may be overheating and just need a good cleaning. They are some of the original cluster nodes | 23:07.31 |
| marcosw: peeves has been around longer than those and hasn't been cleaned yet. | 23:07.54 |
| marcosw: if Miles' office is so dirty, maybe we need to send them to someone else. | 23:08.22 |
| marcosw: or sell them to someone local and have one of us buy new ones (to save shipping) | 23:09.20 |
| well, at least my new laptop is back to limping along. Next, I'll try and install stuff | 23:10.15 |
| AFTER doing a backup ! | 23:10.56 |
Robin_Watts | tor8: Makefile tweaks in a commit on my casper branch. | 23:56.16 |
| Forward 1 day (to 2012/06/27)>>> | |