| <<<Back 1 day (to 2012/04/08) | 2012/04/09 |
runa | does andybody know if mupdf supports synctex or will? | 02:24.20 |
jzmer | how do substitute a truetype font with a postscript/opentype one? | 10:43.10 |
| cidfnmap doesnt seem to work, and direct /foo1 /foo2 results in error | 10:43.27 |
| direct /foo1 /foo2 in Fontmap.GS | 10:43.38 |
kens | Which application ? PostScript/PDF input ? | 10:43.39 |
jzmer | a postscript generated by firefox via cairo | 10:44.02 |
kens | WHy are you trying to replace a TrueType font with a PostScrtipt font ? What do you expect to achieve ? | 10:44.08 |
jzmer | just curious, nothing more | 10:44.19 |
| as for what i expected to achieve, font substitution | 10:44.46 |
kens | Depending on the content of teh PostScript file, you may get an error if you try this, or nothign might happen at all (more liekly) | 10:44.54 |
| wPostScript will use embedded fonts in preference to system fonts | 10:45.10 |
jzmer | i dont think so. since firefox, when generating ps' doenst seem to embed the font already | 10:45.45 |
kens | Are you sure you can tell ? Its not always easy. If the font is not embedded tehn tits not a TrueType font, its a font reference, so you still can't 'replace' a TT font with a PS one. | 10:46.38 |
jzmer | unless you want me to copy paste /foo by hand in the ps file ... | 10:47.00 |
kens | I don't particularly want you to do anything. | 10:47.26 |
| But your quesation so far doesn't make sense to me | 10:47.37 |
jzmer | hmm | 10:48.19 |
kens | If you r PostScript file contains a TrueType font, then GS willl use that to render with. | 10:48.46 |
| If it doesn't then it doesn't contain a TrueType font. | 10:48.58 |
| SO either way you can't 'subsitute a TrueType font with a PostScript font' | 10:49.18 |
jzmer | i think i will check with #firefox in mozilla's irc to get an defintie answer on whether they embedded the font or not | 10:49.29 |
| thanks. | 10:49.45 |
kens | Or you could share the PostScript file and we could tell whether the font is embedded | 10:49.48 |
jzmer | i think i figured out what the issue is now | 12:20.40 |
| firefox did embed fonts in the ps file, which makes font substitution irrelevant | 12:21.08 |
| and hence, i began to try removing the embedded fonts via http://stackoverflow.com/questions/2670809/how-to-get-ghostscript-to-use-embedded-fonts-in-pdf | 12:22.03 |
| but that basically convert the file to vector graphics | 12:22.31 |
| however, per http://partners.adobe.com/public/developer/en/font/5175.CID_ATM.pdf | 12:23.08 |
kens | I can't see that code doing anythign useful. | 12:23.10 |
jzmer | it is possible to preserve the font name yet still does not embed the font | 12:23.34 |
kens | It is always possible to embed a font reference. | 12:23.45 |
| Which requests the font but does not embed it. | 12:23.56 |
jzmer | right, but then the problem is, how do you do this without hand-edit the ps file... | 12:24.05 |
kens | Fundamentally, you don't, | 12:24.19 |
| The PostScript program is created by the application/driver software. | 12:24.37 |
| Once it is created you are left with hand-editing it if you want to change it. | 12:24.54 |
| BTW that document is specific to Adobe Type Manager, I wouldn't suggest using it for anything else. | 12:25.16 |
jzmer | im just saying that pdf is an example of embed only a font reference but not the font itself | 12:25.49 |
kens | What ? | 12:26.05 |
| PDF files can contain fonts or not, just like PostScript | 12:26.23 |
jzmer | hmm, i just wanted to say how do you achieve "font reference" as in http://partners.adobe.com/public/developer/en/font/5175.CID_ATM.pdf | 12:27.12 |
kens | You could add the named fonts to the NeverEmbed list when creating the PDF file. But if the fonts are subset they will probably have 'surprising' names. | 12:27.25 |
| jzmer that document is specific to Adobe Type Manager, you cna't use it for anythign else. | 12:27.47 |
| Here is a PostScript program which does not icnlude an embedded font: | 12:27.58 |
| %! | 12:28.33 |
| 10 10 moveto | 12:28.33 |
| (Hello) show | 12:28.33 |
| showpage | 12:28.33 |
jzmer | ok , i know that you can do something like /Times 20 selectfont <blahblahblah> (XYZ) | 12:28.35 |
kens | Well there you are then, that's how you do it. | 12:28.47 |
| My example was missing a line due to IRC swallowing the font selection. Doesn't like '/' | 12:29.11 |
| jzmer, so what *is* your question ? | 12:29.44 |
jzmer | the question "was" how do you preserve font reference in a existing ps/pdf file without embedding it, and without hand-editing it, but then you answered the question. so the only chance for me to remove the embedded font is, to hand-edit it ... | 12:31.27 |
| unless i went back to #firefox in irc.mozilla.org to ask them for a way for firefox to not embed the font in the first place | 12:32.22 |
kens | Yes, once the PostScript program is created, only hand-editing, or a carefully written script will be able to do anything with it. | 12:32.27 |
| jzmer I doubt they would be interested in that, what would be the point ? | 12:32.41 |
| Systems where the font is missing would not be able to print the file properly | 12:33.01 |
chrisl | jzmer: so, why would you want to ignore the intended font, and force the use of some random other font? | 12:33.01 |
jzmer | chrisl: because STHeitiSC is truetype, and i wanna replace that with AdobeHeitiStd-Regular, which has postscript outline. however, since now you've asked, i think i can use firefox's inspect/debug to edit the css file to include AdobeHeitiStd-Regular | 12:35.43 |
kens | is unable to fathom why you would want to use a PostScript font in place of a TrueType font. | 12:36.33 |
jzmer | chrisl: and as for the font problem yesterday, it turns out to be a vendor-specific font | 12:36.37 |
| vendor-rip-specific font | 12:36.50 |
| distiller failed at that | 12:37.03 |
chrisl | jzmer: okay, so why do you want to replace a TTF with another font format - TTF outlines are supported in PS (Type 42) | 12:38.50 |
kens | Plus worries about Encodings, metrics.... | 12:39.28 |
jzmer | ... because im stupid enough to forget i can "inspect" in firefox to force another font ... | 12:40.25 |
chrisl | Doesn't Firefox use the display font when printing, then - pretty sure it does...... I'm not being stroppy, here, I'm intrigued to know, is all | 12:41.38 |
jzmer | since the targeted website has a invalid css file, and my fontconfig configuration is choppy here | 12:42.58 |
| one last question, which is irrelevant to the question above | 12:44.40 |
| what does gsx adds to gs? log says gtk+, but what does that mean? ghostscript certainly does not have a widget system | 12:45.14 |
kens | Never heard of gsx | 12:45.25 |
| Nothing to do with us | 12:45.36 |
jzmer | i think if you build ghostscript with --enable-gtk or --enable-gtk+ you will get it | 12:45.51 |
chrisl | Basically, gsx uses the display device to dump the rendered output into a gtk+ pixmap buffer - it's just an usage example, really | 12:46.07 |
kens | I'm thinking this won't work on Windows ;-) | 12:47.00 |
jzmer | hmm, sounds like what behdad's killed cairo device | 12:47.22 |
| does | 12:47.27 |
| could be useful if you wanna ditch platform-specific devices | 12:48.04 |
chrisl | The cairo device in GS? That did at least some "high level" things, but was very incomplete. | 12:48.22 |
| kens: as you can get gtk+ on Windows, the basic code could be made to work without too much hassle, but not much point...... | 12:48.57 |
| jzmer: if we were to write the x11 devices, we'd almost certainly use the display device instead of how it's currently done. | 12:50.31 |
| s/write/rewrite | 12:50.40 |
marcosw_ | morning | 13:17.57 |
kens | Morning marcosw_ | 13:18.06 |
| What is marcosw_bot ? | 13:18.16 |
marcosw_ | I was toying with the id of having the cluster send irc messages when clusterpush jobs are finished, so people wouldn't have to check their email. | 13:18.52 |
| ^id^idea | 13:19.03 |
kens | Hmm, but I'd haev to check irc instead :-) | 13:19.07 |
Robin_Watts | marcosw: Isn't that what the dashboard is for ? | 13:19.39 |
marcosw_ | I could send a sms to your phone :-) Or have a modem call your house. | 13:19.51 |
chrisl | We could have bmpcmp fax us the results....... ;-) | 13:21.01 |
marcosw_ | Robin_Watts: belts and suspenders. dashboard, email, irc, carrier pigeon, â¦. | 13:21.37 |
| Robin_Watts: btw, I was in the process of composing an email, but since I have you live I made a slight change to bmpcmp that might need a change in pngs2html (or whatever it's called). | 13:22.34 |
Robin_Watts | marcosw_: Go on... | 13:23.05 |
marcosw_ | If there are no differences found by bmpcmp the cluster now generates an empty filename.00000.meta file and uploads that to the cluster. | 13:23.24 |
Robin_Watts | oh. | 13:23.42 |
marcosw_ | previously it wouldn't upload anything and you wouldn't be able to tell if the file had even been run. | 13:24.01 |
Robin_Watts | I would prefer that we make it say "no differences found" | 13:25.12 |
| to stdout/stderr, and then the cluster could catch that. | 13:25.22 |
| meta files hold meta information 'per output image'. | 13:25.59 |
| Not per run. | 13:26.07 |
| i.e. the existence of filename.00000.meta implies that there should be filename.00000.bmp filename.00001.bmp and filename.00002.bmp too. | 13:28.02 |
| Capturing stdout/stderr to filename.out would make sense, because then we could catch other error messages that bmpcmp produces too. | 13:28.43 |
| (Such as 'refusing to diff image due to malloc failure' or 'refusing to diff images because width/height have changed') | 13:29.21 |
marcosw_ | the cluster does now catch stdout/stderr from bmpcmp and sends that email to the user. see ~regression/cluster/bmpcmpResults.txt on casper for an example. | 13:29.43 |
Robin_Watts | If you could make it capture each stdout stderr to a .out file, and upload that too, then I can include that in the pngs2html thing. | 13:30.39 |
marcosw_ | however, I feel it would be preferable to also show which files generated error or no differences on the bmpcmp generated web page. | 13:30.49 |
Robin_Watts | Right. So that's what I'm suggesting. | 13:31.42 |
| Rather than creating 1 single 'stderr/stdout' file, create one for each run. | 13:32.09 |
| and I'll update pngs2html to include it. | 13:32.30 |
marcosw_ | make sense. I'll continue to email out the list of errors but will also upload a file.out to the bmpcmp directory (an empty file indicating no error). | 13:34.45 |
Robin_Watts | cool. Let me know when that's in place and I'll tweak pngs 2html | 13:36.48 |
kens | Grrr, my code to free pattern memory works fine on Windows, seg faults on Linux on 'some' files :-( | 15:35.31 |
mvrhel | sounds like the kind of fun I had last week | 15:36.55 |
| although mine was just a rendering problem and only with the release on linux | 15:37.10 |
kens | Been there before, I'll have to use gdb :-( | 15:37.13 |
mvrhel | kens: have you used ddd at all? it is crude but useful IDE | 15:38.14 |
kens | dd is what I use in fact | 15:38.26 |
mvrhel | ok | 15:38.33 |
henrys | kens:sounds like something you should valgrind before you even get to the debugger. | 15:57.39 |
kens | seg fault should be easy to find ;-) | 15:58.00 |
Robin_Watts | run under the debugger, and then see what the backtrace is when the SEGV happens. Much easier than valgrind :) | 16:01.59 |
henrys | I was speculating an odd problem that would crash long after the damage (valgrind detectable) was done... I guess we'll see. | 16:04.41 |
Robin_Watts | henrys: Ah, right, yes, possibly. | 16:05.09 |
kens | Nah, I can guess what the problem is, I just need to find where teh memory is getting freed before I get to freeing it again | 16:05.11 |
henrys | a pcl customer found a nice leak in the pattern code. It turns out any pattern from pcl results in bumping the target device reference count | 16:06.31 |
kens | I saw that | 16:06.43 |
Robin_Watts | oops. | 16:06.53 |
| I think I've just spotted a memory overread in mupdfs image code. | 16:07.56 |
| Fixing it may be 'interesting'. | 16:08.17 |
mvrhel | whew. got the tile clip stuff working now with the new color type | 16:16.45 |
| making my way through the last 10 files that seem to show real diffs. I am hoping this fixes a few of them | 16:19.35 |
kens | OK time to go, night alll | 16:21.43 |
henrys | if the output matches and main's return code is non zero (error) does the cluster code report an issue? | 16:22.32 |
mvrhel | ok now I am down to 4 files, 3 of which appear to be a common issue with pattens and transparency | 16:30.18 |
ray_laptop | missed kens. I was going to mention that I usually resort to -ZA to see when a particular item is being freed. Sort of brute force, but that's where searching the output with an editor comes into play. | 16:31.53 |
| mvrhel: sounds good. | 16:32.02 |
mvrhel | slow but sure. | 16:32.18 |
Robin_Watts | Memento should be good for spotting double frees too - it keeps blocks around deliberately to watch for it. | 16:32.40 |
| </CompulsoryMementoPlug> | 16:32.51 |
ray_laptop | Robin_Watts: yeah, but if it's an object in a chunk, it doesn't help | 16:33.07 |
Robin_Watts | ray_laptop: Does too :) | 16:33.17 |
| Memento nobbles the chunk memory manager to only have 1 object per chunk. | 16:33.34 |
ray_laptop | Robin_Watts: is memento configured as a gs_memory_t ? | 16:33.41 |
Robin_Watts | No. | 16:33.52 |
ray_laptop | Robin_Watts: does it also do that for the GC memory ? | 16:34.07 |
Robin_Watts | Pass. | 16:34.28 |
| The chunk memory manager has a line in it that says "if this object is >= x bytes, then put it in a chunk of it's own" | 16:34.51 |
| Memento sets 'x' to be 1. | 16:34.59 |
mvrhel | aha! found the issue. when the target is pdf14cmykspot, the pattern tile is not getting set as a planar type | 16:37.08 |
Robin_Watts | henrys: I can't answer your cluster question - you need marcosw_. My pidgin perl lets me down I'm afraid. | 16:37.29 |
ray_laptop | Robin_Watts: nm. I see in gsalloc.c and #ifdef MEMENTO iimem->large_size = 1 | 16:37.53 |
| Robin_Watts: I just hadn't looked. | 16:38.08 |
Robin_Watts | ray_laptop: That's the puppy. | 16:38.16 |
henrys | Robin_Watts:np I thought you might happen to know. | 16:38.16 |
mvrhel | brb | 16:38.35 |
Robin_Watts | henrys: I suspect the cluster looks at return codes BEFORE comparing md5 sums. | 16:38.49 |
ray_laptop | mvrhel: isn't there a special op call to see if a device is planar ? | 16:39.03 |
marcosw_ | henrys: are you asking if the cluster detects the return code from main? | 16:39.13 |
henrys | marcosw_:more or less. | 16:39.57 |
Robin_Watts | marcosw_: henrys asked "if the output matches and main's return code is non zero (error) does the cluster code report an issue?" | 16:40.14 |
marcosw_ | I actually don't remember, let me check. | 16:40.17 |
henrys | when it's detected is more important - I expect the output to match but get an error code. | 16:40.36 |
| for the test I imagine doing, can I do that? | 16:41.20 |
marcosw_ | the answer is no, if the md5sums match and no error message is printed out the error code is ignored. I presume you'd like that behavior to change? | 16:42.43 |
henrys | yes for my experiment, but it seems reasonable to always look at the error code, don't you think? | 16:44.36 |
ray_laptop | marcosw_: BTW, I just happened to scroll to the end of the 'aging' report and see a whole mess of "bugs that have been resolved but not closed" -- seems like this was to be a reminder to you to communicate back to the customer, then "CLOSE" it, but with such a long list, it probably isn't helping you. | 16:45.06 |
marcosw_ | ray_laptop: true, I don't really use that list. I keep telling myself I should either use the information or stop emailing it out, preferably the former :-) | 16:46.00 |
ray_laptop | mvrhel: avadhut sure is persistent :-( | 16:46.21 |
mvrhel | ray_laptop: yes there is. the pdf14 device was not planar in the sense that it was not using the planar device buffers but its own planar buffers | 16:46.22 |
| hehe. yes | 16:46.27 |
| I am going to try to get to that this week | 16:46.49 |
Robin_Watts | henrys: Seems reasonable to me. A test file that always returns an empty bitmap and an error would never fail otherwise. | 16:46.51 |
mvrhel | I *really* want to get this tiffsep thing wrapped up | 16:47.09 |
| if I step away from it it is too hard to get started back into again | 16:47.20 |
marcosw_ | henrys: I agree we should not be ignoring the error code. I'll make that change but I have to attend a presentation that starts in 15 minutes, so it will have to wait for a couple of hours. | 16:47.23 |
henrys | marcosw_:okay | 16:47.42 |
ray_laptop | mvrhel: I agree that wrapping up the planar spot color stuff is higher priority (although avadhut would probably disagree) | 16:48.00 |
mvrhel | shaving .02 secs of of his time vs. me getting something useful done | 16:48.46 |
marcosw_ | henrys: I believe the cluster now detects non-zero exit codes, independently of md5sum matches or error messages (or at least I think it's supposed to, I can't test this since I can't easily force a non-zero return value without also causing an error). | 17:08.11 |
mvrhel | oh my | 17:08.31 |
| need to add copy_planes into the pdf14 device now... | 17:08.40 |
| which will actually speed things up when drawing patterns into this device | 17:09.05 |
| argh. things are a bit messed up though as the number of planes in the tile is 3 and we are going into a pdf14gray device. | 17:11.05 |
| oh actually in this case the pdf14gray device has changed itself to rgb due to a group push so all is well | 17:20.42 |
| of course two of my final issues appear to be cairo files with a huge patterns | 18:32.48 |
| this thing is insane | 18:33.12 |
| hmm somehow I have broken the pdf14 device big time. off for lunch for a bit | 18:58.36 |
| bbiaw | 21:00.36 |
| Forward 1 day (to 2012/04/10)>>> | |