| <<<Back 1 day (to 2012/03/13) | 2012/03/14 |
henrys | so if you revert marcos change everything is okay? | 00:00.48 |
| that was just checking the size of the output can't see how that would matter. | 00:01.17 |
mvrhel | I agree | 00:07.17 |
| does it fail for anyone else? | 00:07.26 |
henrys | checking the mac result | 00:08.15 |
| seems to run okay you see a crash? | 00:09.08 |
mvrhel | no the check at line 358 fails | 00:09.30 |
| which is right where marcos did all his changes | 00:09.43 |
henrys | hmmph on tiger? | 00:11.29 |
mvrhel | ok pdev->height is larger than (0xFFFFFFFF-0)/width | 00:11.36 |
| yes on tiger | 00:11.44 |
| out to tiff32nc | 00:11.51 |
| wait a minute | 00:14.49 |
henrys | so at 300 dpi width == 3300 and height would then be larger than 1301505, which isn't the height | 00:15.05 |
| sorry 2550 but same strangeness. | 00:15.50 |
mvrhel | hmm I dont see why I am going in there. I am running at 72 dpi height is 792 | 00:16.42 |
alexcher | mvrhel: the problem is reproduced on 32-bit Windows | 00:16.54 |
mvrhel | which is not larger than 4G/width | 00:16.57 |
| alexcher: ok thanks. so I am not completely crazy | 00:17.13 |
| I think there is some issue with the casting of stuff in here | 00:18.57 |
henrys | oh it's signed | 00:19.00 |
mvrhel | yes | 00:19.03 |
| ftell returns a long | 00:19.58 |
henrys | man that marcos he's really something ;-) | 00:19.59 |
| on linux and mac long is 8 bytes so it's good. | 00:20.32 |
mvrhel | alexcher: ok. so your patch is likely good | 00:22.13 |
| ok so just changing the long to unsigned long makes everything happy. henrys: is there any issues you see with that? | 00:25.26 |
henrys | you just add U suffix right? | 00:25.48 |
mvrhel | yes | 00:25.52 |
| well no | 00:26.19 |
henrys | you can give it back to marcos or just fix it. | 00:26.21 |
| it can't be a long | 00:26.46 |
mvrhel | i changed it to unsigned long | 00:27.00 |
henrys | that was the original problem it just needs to be 4 bytes unsigned. | 00:27.06 |
mvrhel | is there a def | 00:27.07 |
henrys | if you use long it won't work on mac or linux | 00:27.18 |
mvrhel | now I am confused. with long it does work on mac and linux | 00:28.20 |
| but not windows | 00:28.24 |
henrys | oh yes it will I'm sorry | 00:28.26 |
| before it was MAX_LONG - and that won't work on mac and linux because it is a much larger number. | 00:29.14 |
mvrhel | ah ok | 00:29.19 |
henrys | but the 0xFFFFFFFF-0 should work fine | 00:29.27 |
| not sure what's up with the - 0 stuff. | 00:29.42 |
mvrhel | ftell(file) was returning zero | 00:30.00 |
| it is just the amount left | 00:30.10 |
| in this case (the first time) we have the full 4Gs left | 00:30.23 |
henrys | oh I thought you were copy pasting the code - sorry. | 00:30.31 |
mvrhel | sorry | 00:30.35 |
| no | 00:30.36 |
| so I changed all the (long) 0xFFFFFFFF to (unsigned long) 0xFFFFFFFF | 00:31.13 |
| there were a few of them | 00:31.18 |
henrys | well that was an easy one - not man of those. | 00:31.26 |
| s/man/many | 00:31.33 |
mvrhel | unless we have some special type name that we use | 00:31.42 |
henrys | What we'd want is something like MAX_UINT32 which I don't think we have. | 00:32.17 |
mvrhel | oh ok | 00:32.29 |
henrys | I'd just leave it. | 00:32.30 |
mvrhel | sounds good. i need to go help with dinner. i will commit this tonight | 00:32.48 |
henrys | great | 00:33.04 |
Robin_Watts | henrys: It tested the wrong branch. | 00:42.00 |
henrys | oh I see. | 00:42.25 |
Robin_Watts | Hence all the compile fails (that branch hadn't got the updated thirdpaty) | 00:42.30 |
| now it's back to master again, it's working. | 00:42.44 |
| I know why it did it too - I did something odd with git. | 00:42.58 |
Devoid | hello | 00:48.05 |
| is ghostscript a keylogging software? | 00:50.45 |
henrys | heading out for quiz trivia night I'll see you tomorrow Robin_Watts | 00:51.03 |
Devoid | derp. | 00:51.04 |
henrys | Devoid:no | 00:51.21 |
Devoid | henrys, what is it for than? | 00:51.49 |
| i found it installed on all the netbooks at my school | 00:52.06 |
henrys | google "what is ghostscript" got to go. | 00:52.09 |
Devoid | ..why does it encrypt files? | 00:52.50 |
Robin_Watts | night henrys. | 00:53.19 |
| Devoid, I'm sorry, I think you may be confusing our Ghostscript with something else. | 00:53.58 |
| Our Ghostscript is a postscript interpreter - a piece of software that renders Postscript (and PDF) files. | 00:54.36 |
Devoid | Okay.. thank you. | 00:55.32 |
marcosw | I'm going to take casper down in about an hour. Assuming everything goes reasonably well it should be back up an hour or so later. | 05:12.40 |
mvrhel | marcosw: let me do my commit first... | 05:35.48 |
| ok. done | 05:45.46 |
| I will push the output intent stuff tomorrow | 05:46.05 |
chrisl | I don't propose to worry about it - I'm sure if you look hard enough you can find a site accusing *every* piece of software of being malicious in some way | 08:16.11 |
kens | Yeah, I was just amused | 08:16.22 |
| So, back to memory leaks :-( | 08:17.16 |
chrisl | I suppose you could, in theory, use Ghostscript to run a Postscript program which could act as a keylogger, for someone paying sufficiently little attention...... | 08:17.28 |
kens | I think it would be hard to hide :-) | 08:17.44 |
chrisl | But some people are *that* stoopid....... | 08:18.06 |
| But, you're right, back to something that will pay the bills - Freetype dropouts...... <sigh> | 08:18.52 |
| reboot..... | 08:49.34 |
ManDay | GPL Ghostscript 9.04: Error: Font Renderer Plugin ( FreeType ) return code = -1 | 08:54.17 |
| on ps2pdf | 08:54.25 |
| ideas? | 08:54.32 |
kens | Yes, report a bug. | 08:54.37 |
| And try 9.05 | 08:54.42 |
| We'll need the file to make any progress with that one if it exhibits under the current code | 08:54.58 |
ManDay | I thought that maybe it was due to some mislinkage | 08:55.48 |
kens | No, that's an error from the font renderer saying, basiclaly, 'somethign bad happened' | 08:56.09 |
ManDay | it didn't happen until today | 08:56.13 |
kens | Did you try this file before today ? | 08:56.22 |
ManDay | Yes | 08:56.26 |
kens | Are you suing shared libraries ? | 08:56.30 |
ManDay | Suing? | 08:56.39 |
kens | using | 08:56.42 |
ManDay | using | 08:56.43 |
| I guess I do | 08:57.25 |
kens | Have you recently udated FreeType ? | 08:57.39 |
ManDay | I already recompiled gs | 08:57.42 |
| kens: I don't recall | 08:57.47 |
kens | Well, its going to be hard to help if you can't tell us..... | 08:58.00 |
ManDay | What I want to know is whether there could be other components than GS that I'd have to recompile | 08:58.03 |
| Is the "Font Renderer Plugin" part of another package? | 08:58.16 |
kens | If you are using shared libraries, yes, otherwise no | 08:58.23 |
ManDay | here is my ldd: | 08:58.36 |
kens | You can use differetn font engines, teh default one is FreeType | 08:58.43 |
| ldd won't help me | 08:58.50 |
ManDay | Your paste can be seen here: http://paste.pocoo.org/show/565372/ | 08:58.55 |
kens | Won't mean anythign to me | 08:59.02 |
| You need a Linux type. | 08:59.09 |
ManDay | You mean you are not using linux? | 08:59.27 |
kens | Not exactly. | 08:59.33 |
ManDay | Not exactly using it or not exactly what you mean? | 08:59.44 |
kens | I use Linux in a VM (2 different kinds) but I compile under Windows | 08:59.47 |
ManDay | Ok | 08:59.53 |
kens | I don't speak Linux builds | 09:00.02 |
ManDay | I'll try recompile freetype then | 09:00.12 |
kens | It looks like you are using FreeType as a shared library, maybe. | 09:00.38 |
| You need a pretty recent version of FreeType | 09:00.49 |
| Or you can wait for chrisl to reboot and he can help you more with Linux builds | 09:01.14 |
ManDay | well I recompiled reetype but that didn't really help | 09:04.15 |
| chrisl ? | 09:04.37 |
chrisl | Yes | 09:04.43 |
ManDay | Does GPL Ghostscript 9.04: Error: Font Renderer Plugin ( FreeType ) return code = -1 | 09:05.09 |
| tell you anything | 09:05.12 |
| (I'm on Gentoo) | 09:05.17 |
kens | chrisl you probably want to read the log for the last 5 minutes | 09:05.22 |
ManDay | I already rebuilt gs and freetype | 09:05.23 |
| PS: ps2pdf works nonetheless | 09:06.10 |
| I guess I should have mentioned | 09:06.16 |
chrisl | As kens said, it means "something went wrong" - we don't have a lot of info as regards the internals of the font renderer | 09:06.30 |
kens | ps2pdf does not render fonts but embeds them in the output PDF file verbatim | 09:06.32 |
chrisl | ManDay: what version of Freetype are you using? | 09:06.42 |
ManDay | These are the packages that would be merged, in order: | 09:06.58 |
| Calculating dependencies ... done! | 09:07.00 |
| [ebuild R ] media-libs/freetype-2.4.9 USE="bzip2 -X -auto-hinter -bindist -debug -doc -fontforge -static-libs -utils" 0 kB | 09:07.02 |
| Total: 1 package (1 reinstall), Size of downloads: 0 kB | 09:07.04 |
chrisl | Well, that means nothing to me..... | 09:07.36 |
| ManDay: are you using the GS archive downloaded from our site, or the Gentoo package? | 09:08.16 |
ManDay | the package | 09:08.23 |
| but didn't you just say that this error is insiede of freetype? | 09:08.36 |
| (not inside of gs) | 09:08.41 |
kens | No. | 09:08.52 |
| Its a report from Ghostscritp that something in FreeType caused an error | 09:09.07 |
| THis does not mean its a FreeType 'error' | 09:09.18 |
| You say this file has worked before, so what has changed since tehn ? | 09:09.38 |
chrisl | Can you give us the file in question? | 09:09.47 |
ManDay | I can't recall, sorry. I guess I must have updated something | 09:09.51 |
| sure, sec | 09:09.58 |
| hrm, got hrm, got a pastebin for me? | 09:12.47 |
| or email? | 09:12.50 |
| omploader doesnt seem to work for me atm | 09:12.56 |
chrisl | ManDay: send it to chris.liddell (at) artifex.com | 09:13.12 |
ManDay | ah wait, i think it's working now | 09:13.29 |
| http://ompldr.org/vZDE3cg | 09:14.12 |
| There you go | 09:14.15 |
chrisl | And what command line are you using? | 09:15.00 |
ManDay | do you, in exchange, have a ps for me which should work? | 09:15.02 |
| ps2pdf file.ps | 09:15.08 |
chrisl | I thought you said ps2pdf works? | 09:15.44 |
kens | Beat me to it :-) | 09:15.51 |
ManDay | yes it works in that I get a result. But I have those errors | 09:16.19 |
kens | Then your PDF file will be missing some text | 09:16.31 |
ManDay | Maybe there are errors in the result, too, I haven't checked | 09:16.31 |
| I see | 09:16.35 |
| damn | 09:17.14 |
chrisl | Well, *our* release package runs that file without a problem. Note that 9.04 was tested and released with Freetype 2.4.3 | 09:17.18 |
ManDay | I just submitted a paper which had been converted that way | 09:17.24 |
kens | If you got those errors, then there will be missing text in the PDF | 09:17.40 |
| Best bet is to use our source code, not the package | 09:17.51 |
| Which will include a known working version of FreeType | 09:18.03 |
ManDay | All the (visible) text appears to be there | 09:18.13 |
kens | shrugs | 09:18.21 |
chrisl | Actually, there may *not* be missing text, if the error comes from FAPIrefinefont or similar, rather than FAPI_do_char | 09:18.42 |
kens | The error is telling you that it tried to create a bitmap for some of the text, and couldn't | 09:18.42 |
ManDay | A bitmap?! | 09:19.01 |
| Why would it do that? | 09:19.04 |
kens | Because it can't embed the font for some reason | 09:19.12 |
chrisl | The error can come from places other than creating the glyph bitmap | 09:19.22 |
kens | And before you ask why, I don't kow without debugging it | 09:19.23 |
ManDay | There is only one font throughout the whole document | 09:19.26 |
kens | Don't count on that | 09:19.34 |
| You may only think you are using one font. | 09:19.45 |
chrisl | ManDay: no, there's four...... | 09:19.52 |
ManDay | Oh? | 09:19.56 |
| How do you know? | 09:20.01 |
| Well ok, if italic and bold counts | 09:20.10 |
kens | Reading the PostScritp progrm | 09:20.11 |
| And yes, each face is a differnt font | 09:20.23 |
ManDay | Then we have regular, bold, italic and the font for the formulas | 09:20.29 |
| all three appear in the document, though | 09:20.36 |
kens | So 4 then | 09:20.38 |
ManDay | all four, I mean | 09:20.47 |
chrisl | So, not "one font" then..... | 09:21.07 |
kens | But it could be an individual glpyh, not the font. In which case only one glyph will be missing | 09:21.09 |
ManDay | Couldn't GS be a bit more specific with these errors?! | 09:21.13 |
| hm | 09:21.22 |
kens | specific how ? | 09:21.25 |
ManDay | I guess I'll know once the paper is reviewed | 09:21.30 |
| kens: Say, for example, which line in the ps it originates from | 09:21.41 |
kens | And as chrisl just said 'no', we can't get much back from FreeType | 09:21.42 |
| ManDay that's not possible | 09:21.50 |
| PostScript is a programming language | 09:21.59 |
ManDay | yes, so? | 09:22.05 |
kens | SO telling you that 'line 10' caused a problem wouldn't help you | 09:22.11 |
ManDay | eh, of course it would | 09:22.23 |
kens | Because that's a procedure called form line 10,000 | 09:22.24 |
ManDay | well, give a backtrace of sorts then? | 09:22.40 |
chrisl | ManDay: how much Postscript do you know? | 09:22.44 |
kens | THat's not really possible, no | 09:22.47 |
ManDay | as you said: its just a programming language | 09:22.49 |
kens | We 'could' do that, at the expense of lots of performance | 09:23.03 |
ManDay | chrisl: none, to be honest. I assumed it is just another language | 09:23.05 |
kens | Its an interpreted language | 09:23.15 |
ManDay | python is, too, so is javascript | 09:23.25 |
kens | And it has to be fast or our customers complain | 09:23.31 |
| javascrip is usually compiled | 09:23.37 |
ManDay | oh, you mean like that | 09:23.43 |
| yeah, ok, I see | 09:23.47 |
chrisl | ManDay: python is a JIT compiled language | 09:24.00 |
ManDay | yes, i got your point | 09:24.42 |
kens | Anyway, the point seems to be that there is a problem, probably using the version of FreeType you have loaded | 09:24.49 |
| Which we don't yet support. | 09:24.57 |
| THis is why we prefer that people use our sources, which include 3rd party sources that we know actually work with our code | 09:25.19 |
| I feel your best bet is to donwload and build Ghostscritp from our sources. | 09:25.47 |
| If that causes a problem then we will fix it. | 09:25.59 |
ManDay | kens: There is a good reason why things are packaged and people should not bundle their stuff. Of course everyone wants to use one specific version of a 3rd party lib, but you just shouldnt. | 09:26.20 |
| bundling stuff is not a solution | 09:26.30 |
kens | ManDay then it doesn't work | 09:26.33 |
chrisl | ManDay: then how do we QA our work? | 09:26.41 |
ManDay | You QA it with one specific version and make that version a requirement | 09:26.56 |
kens | ROFL | 09:27.04 |
ManDay | obviously, teh requirements in the ebuild are incorrect | 09:27.04 |
| otherwise it'd have not updated to an incompatible freetype | 09:27.16 |
| or slotted it, for that matter | 09:27.22 |
kens | If I may be blunt B*ll*cks | 09:27.32 |
ManDay | the rest of the world doesn't think so | 09:27.59 |
kens | If your package system is capable of dealing with that, then you should complain to your package maintainers | 09:28.00 |
| Not us. | 09:28.07 |
ManDay | kens: I will. | 09:28.08 |
| Meaning: I'll file the according bug | 09:28.15 |
| Doesn't change the whole "we bundle every 3rd party lib we want, it's more covenient to develop that way", though | 09:28.36 |
kens | Remember, we don't create he package and have no control over what is in it | 09:28.40 |
ManDay | kens: I did not mean to charge blame | 09:28.50 |
kens | ManDay : we have to test somethign | 09:28.57 |
ManDay | Huh? | 09:29.05 |
chrisl | The problem is, we are a cross platform package - so we need to ensure our users can build *easily* on all supported platforms...... | 09:29.16 |
kens | As chrisl said, we cna't test every version iof every package we use. We do publish what we use. | 09:29.24 |
| If people want to use shared libraries, that's up to them | 09:29.44 |
| But we cannot be held responsible for failures in that case. | 09:29.56 |
| Only for the versions we actually use and supply | 09:30.04 |
chrisl | Hmm, FT 2.4.9 does indeed error - I wonder why...... | 09:30.09 |
kens | And in some cases we have patched teh libraries and so they *won't* worj with the standards | 09:30.21 |
| And yes, we have supplied those patches upstream | 09:30.34 |
ManDay | No one said you should test every version of every package. You should test the current versions of each package. That's it | 09:30.42 |
chrisl | ManDay: your file should be fine - in the event of an error like that, we still fallback to using the old internal glyph renderer | 09:30.51 |
ManDay | That's how everyone does it, if I may be blunt | 09:30.53 |
kens | ManDay not reasonable, we don't have that kind of time | 09:30.54 |
ManDay | kens: Time? For what? For running ./run_test.bsh ? | 09:31.15 |
kens | ANd what does that etst ? | 09:31.26 |
| test ? | 09:31.29 |
| We have an internal test suite of 50,000 tests | 09:31.36 |
ManDay | Then the answer is 50000 tests | 09:31.49 |
kens | Which is *far* from exhaustive | 09:31.51 |
| If we were to be exhaustive that would grow by a factor of about a million I woudl guess | 09:32.11 |
ManDay | Why do you have those tests then if you don't use them? | 09:32.14 |
kens | Its not the etsts, its all the possible command line options | 09:32.29 |
| CMYK/RGB/gray/mono diffeing resolutions, the list is immense | 09:32.58 |
| We do a 'reasonable' number of configurations | 09:33.16 |
| testing the most common parameters | 09:33.29 |
| On a cluster of 12 multi-core machines that takes around 30-40 minutes to run | 09:33.57 |
ManDay | Again, what I'm saying is that you should test for ONE version only - the possibly newest one. And you make THAT a requirement. And if you can manage to test for more, older versions, you can specify them as bbeing allowed, too. There is no disadvantage to bundling things, then. | 09:33.57 |
| I do not suggest more tests, do I? | 09:34.18 |
kens | Updating to newer versions costs us time we don't have | 09:34.22 |
ManDay | Well, usually, newer versions are downwards compatible | 09:34.56 |
kens | Anyway, this is a pointless conversation, I do not agree wiht you and you are not about to persuade me otherwise | 09:35.01 |
ManDay | there isn't much updating to do | 09:35.02 |
kens | Ho ho.... | 09:35.07 |
ManDay | i said usually | 09:35.14 |
| and IF there is you are always free to just say "no" | 09:35.24 |
| and make an older version a requirement | 09:35.31 |
kens | Well we have done that IMO | 09:35.41 |
ManDay | i just don't see the point of bundling anything | 09:35.42 |
kens | Because then it actually works | 09:35.52 |
chrisl | ManDay: try building on Windows - you'll see the point, then! | 09:35.58 |
kens | Strangely, most people like it to work | 09:36.03 |
chrisl | Huh, "SFNT_Err_Unknown_File_Format"....... | 09:37.52 |
kens | That's od. | 09:38.10 |
| COuld try a bisect fot the change ? | 09:38.29 |
chrisl | I will in a bit - I may have to have some coffee first. Oh, and I have to make a phone call in a bit....... | 09:39.09 |
kens | No rush, IMO | 09:39.18 |
chrisl | Well, only if something is broken in freetype - if that's the case, I'd like to get a fix to Werner ASAP | 09:39.46 |
kens | Sure, but it won't affect any of our paying customers, the majority of free users, or anyone who uses our distribution :-) | 09:40.14 |
| But getting it fixed would be good | 09:40.34 |
| Want me to take a poke ? | 09:40.41 |
chrisl | No, it's fine, I'll keep at it - you've got enough to do! | 09:40.59 |
kens | Not really, just trying to figure out memory leaks | 09:41.12 |
chrisl | Exactly! | 09:41.21 |
kens | Yeah, but its a long slow process, and I will put it off as much as possible :-) | 09:41.44 |
| Doens't help that the PCL intepreter leaks *more* memory when rendering than pdfwrite does. | 09:42.12 |
chrisl | Hmm, well, there's nothing obvious in the environs of where the error happens, so it looks like a bisect is the only option....... | 09:45.03 |
kens | :-( | 09:45.17 |
chrisl | I'm pretty sure 2.4.6 worked - I'll try that quickly, as I happen to have it here...... | 09:46.04 |
kens | Makes sense | 09:46.11 |
chrisl | Yeh, 2.4.6 is fine..... | 09:47.55 |
ManDay | What do you need bisected? | 09:48.06 |
| freetype or gs? | 09:48.10 |
kens | Well that's a big step in the right direction | 09:48.12 |
chrisl | Freetype | 09:48.15 |
ManDay | i can to that today evening | 09:49.11 |
chrisl | ManDay: I should get to it in a little while, thanks | 09:50.35 |
ManDay | ok, great | 09:55.30 |
kens | Robin_Watts : are you around yet ? | 09:56.41 |
ManDay | FYI I'm pretty sure that 2.4.8 should have worked | 09:56.51 |
| Because that seems to be the last version I had been on, I'm downgrading to it now and let you know | 09:57.07 |
chrisl | ManDay: Yes, I'm also fairly sure 2.4.8 worked - I just remembered, it had a feature I added in that release, so I *probably* tested it. | 09:58.03 |
ManDay | chrisl: affirmed, 2.4.8 works | 09:59.39 |
chrisl | ManDay: thanks - it seems to be a problem parsing the "Private" dictionary in a Type 1 font..... | 10:00.21 |
kens | Out for a bit, back later | 10:22.48 |
Robin_Watts | kens: I'm here. | 10:23.06 |
| Damn. too slow. | 10:23.11 |
tkamppeter | chrisl, hi | 10:47.41 |
chrisl | tkamppeter: hi | 10:47.54 |
tkamppeter | chrisl, I have fixed the two bugs in cups-filters now, by inserting your code snippets. I inserted them right after %%BeginProlog. This is still before all acctive PS code but the pstops CUPS filter preserves the code insertion and even puts them into each copy if it has to send the file once per copy (printers without internal copy generation). | 10:49.22 |
| chrisl, thank you very much for your effort on fixing this! | 10:49.43 |
chrisl | tkamppeter: I was just saying to kens earlier that I was surprised how quickly we'd narrowed down the problems - and how helpful the various users had been in helping. | 10:50.32 |
| tkamppeter: I'm really glad to get them resolved! | 10:51.23 |
tkamppeter | kens, also thank you for your help on this. | 10:52.16 |
paulgardiner | Morning Tor | 11:45.30 |
tor8 | hi paul | 11:52.42 |
Robin_Watts | tor8: Did you have any clues about the metro mail ? | 12:21.41 |
tor8 | 1 - first it looks like they don't have any debugging skills at all. not a valid metro style app sounds odd, we don't touch anything win32-based in our code. and it would be rather odd if you're not allowed to touch the file system in a metro app. | 12:27.41 |
| 2 - the error message looks like it has c++ name mangling | 12:28.09 |
| perhaps they didn't do extern "C" { #include "fitz.h" } | 12:28.36 |
Robin_Watts | Radu seems competent - not sure about the others - had no experience of them. | 12:29.02 |
| I wonder whether metro apps are sandboxed in some way. | 12:29.27 |
tor8 | getting -1 on open() sounds like a bad file path | 12:29.33 |
| could be sandboxing, or disallowing POSIX api:s altogether (but that would be a bad decision on MS part) | 12:29.51 |
Robin_Watts | http://arstechnica.com/civis/viewtopic.php?f=14&t=1156133 | 12:29.55 |
tor8 | "Have universal Filesystem access" ... the scanner may be looking for file system calls there | 12:30.30 |
Robin_Watts | Yeah. | 12:30.41 |
tor8 | but they could pass in an open file descriptor with the fz_open_fd call, but I don't think windows uses file descriptors natively -- just as part of libc | 12:31.26 |
| so I doubt you can get a filedescriptor that works with read() from metro api:s | 12:31.42 |
| and probably not even a FILE* | 12:31.46 |
| a workaround to see if they can get things working is to pass in a fz_open_memory based stream. hm, we still need to add fz_open_document_with_stream don't we? | 12:33.02 |
Robin_Watts | tor8: Yeah. | 12:37.09 |
| I'll send them an email. | 12:37.18 |
| http://lunarfrog.com/blog/2011/10/03/winrt-storage-overview/ | 12:37.37 |
tor8 | huh. did marcosw change the ssh id when he upgraded casper? | 12:50.32 |
Robin_Watts | tor8: He went to a 64bit kernel I think. Putty does tell me the key has changed. | 12:51.42 |
tor8 | okay, then it's not just me | 12:52.05 |
Robin_Watts | Git 1.7.0.4, so that's a lot newer. | 12:52.05 |
| reply sent. | 12:54.14 |
| tor8: I pushed the android app last night. | 12:55.08 |
| I'm looking back through sumatraPDF.patch again. | 12:55.35 |
| and trying to close some of the new mupdf bugs. | 12:55.45 |
tor8 | hm, we should put the -t mode into sane | 12:55.58 |
| yeah, he's got a lot of xps fixes we should take on. | 12:56.12 |
| I saw his rather irate bugzilla comments :( | 12:56.24 |
Robin_Watts | I've been hiding from the font and xps ones. | 12:56.35 |
| I'm trying to avoid just blindly taking his stuff on. | 12:56.49 |
tor8 | I'll tackle the xps ones if you're feeling overwhelmed | 12:57.04 |
Robin_Watts | You have other stuff to do for the release though, right ? | 12:57.30 |
| I was planning to leave those until last. | 12:57.39 |
tor8 | I don't think we have any urgent features left, just bug fixing and polishing | 13:01.37 |
| and the android app needs to be set up on the market | 13:01.52 |
| and I have to upgrade the ios code | 13:02.00 |
Robin_Watts | Upgrading the ios code needs to happen before 1.0. | 13:04.30 |
| Getting onto the market doesn.t | 13:04.35 |
| We can just put an apk on the website initially. | 13:04.47 |
| Do you want to do the ios thing, then look at the XPS stuff? | 13:05.26 |
tor8 | yeah. I just fixed the text device things zeniko reported, and fixed a couple of memory leaks so that mudraw -t runs without leaks. | 13:08.19 |
Robin_Watts | If we have time we should do some build=memento leak tests too. | 13:09.03 |
tor8 | what do you think about his patch to change FZ_MAX_COLORS to 8 rather than 32? | 13:10.33 |
Robin_Watts | tor8: Intended to make shades take less memory ? | 13:11.01 |
tor8 | yeah. that function lookup table is rather wasteful. | 13:11.21 |
Robin_Watts | I don't know off the top of my head how that will affect things. | 13:11.36 |
tor8 | it shouldn't affect much, but we'll get errors on files that use more than 8 colors (iirc the pdf ref limit was 32 at some point) | 13:12.24 |
| fz_md5_data makes sense to take on as the public api | 13:12.38 |
Robin_Watts | I'd be reluctant to drop it if it will cause failures on valid failes. | 13:13.00 |
| Yes. fz_md5_data (fz_md5_block ?) seems reasonable. | 13:13.13 |
tor8 | considering how much he rants about his regressions, I wouldn't worry :) | 13:13.23 |
| data or memory I think, block implies other things | 13:13.55 |
Robin_Watts | I don't know that area of the code at all. Can we reasonably make it dynamic? | 13:14.37 |
tor8 | yes, it's static mainly out of laziness on my part | 13:15.00 |
| then I bumped the maxcolors to 32 when seeing the limit in the spec (and not because I ran into files that used many colors) | 13:15.26 |
Robin_Watts | Gawd, FZ_MAX_COLORS is used EVERYWHERE. | 13:15.28 |
tor8 | yup. | 13:15.46 |
Robin_Watts | So if we can make it dynamic just in the shading bit that would be nicer. | 13:16.38 |
| but... that looks a bit hairy. | 13:16.44 |
| Hi kens. You were looking for me ? | 13:29.34 |
kens | Robin_Watts : yes | 13:29.49 |
| I can't get the CTR-Alt-Q thing to work for me at all | 13:30.01 |
| I did enable the Memento_breakpoint, by changing the #ifdef | 13:30.14 |
| But it doesn't break on the memory assignment | 13:30.27 |
| Only on Memento_fin | 13:30.34 |
| Any idea what I'm doign wrong ? | 13:31.19 |
Robin_Watts | Ctrl-Alt-Q, then you enter Memento_breakAt(1234) and hit return. | 13:31.42 |
kens | Yes, did that | 13:31.49 |
Robin_Watts | And in the watchpoint window it'll say 1234 (but in hex) | 13:31.55 |
kens | Followed thee instructions from teh irc log | 13:31.57 |
| Possibly, just a moment | 13:32.03 |
| Hmm, ctr-alt-q isn't doing anythign now.... | 13:33.07 |
| oh wait | 13:33.13 |
| OK doesn't say it in hex for me, uses decimal | 13:33.59 |
| Name Memento_breakAt, Value 20982, type int | 13:34.22 |
Robin_Watts | OK. | 13:34.35 |
| So now escape, then 'play' | 13:34.41 |
| (continue, whatever...) | 13:34.49 |
kens | Ah, this time it worked | 13:34.56 |
Robin_Watts | Fab. | 13:35.00 |
kens | I wonder what I was doing wrong..... | 13:35.02 |
Robin_Watts | I sometimes manage to forget to hit return. | 13:35.12 |
kens | No I did that, I have no idea what I did | 13:35.25 |
| Anywya, its OK now, so that's great, thanks :-) | 13:35.43 |
Robin_Watts | fab. | 13:35.48 |
kens | Its even a pdfwrite allocation this teim | 13:35.52 |
| allocating a cmap, how weird, given that its PCL input... | 13:36.53 |
| Oh, its a ToUnicode CMap | 13:38.13 |
chrisl | kens: I found the freetype problem....... | 13:50.21 |
kens | Oh good! (sorry, went for a coffee) | 13:53.51 |
| What was hte change ? | 13:54.02 |
chrisl | Werner added some stream length checking in the binary decoding code used for reading Subrs and CharStrings - unfortunately, the incremental interface relies on being able to feed empty binary data into those....... | 13:55.19 |
kens | oops, oh dear.... | 13:55.32 |
| We know the incremental interface isn't widely used... | 13:55.54 |
chrisl | Yes, it's a reminder I should test every now and again with the Freetype git code...... | 13:56.21 |
kens | This is true of course, but how foten ? it gets released quite frequently | 13:56.48 |
chrisl | True, but it would be nice to catch issues like this before release. | 13:57.31 |
kens | Dream on.... :-) | 13:57.50 |
chrisl | <sigh>....... | 13:57.59 |
| It is now fixed, though. | 13:58.22 |
kens | Great, presumably you've informed Werner ? | 13:58.37 |
chrisl | I mean it's fixed in freetype's repos! | 13:58.52 |
kens | Oh! | 13:58.59 |
chrisl | By me and Werner - he tidied up my patch a bit | 13:59.14 |
kens | :-) As long as he accepted it, we're happy, right ? | 13:59.31 |
| In any event, its good t get it fixed. | 13:59.48 |
chrisl | Yep. The only thing I'd forgotten was that it was possible to build FT *without* the incremental i/f, in which case the pointer I was checking could be unintialised | 14:00.24 |
kens | Ooops...... | 14:00.37 |
chrisl | Personally, I can't see the point in keeping the configuration without the incremental interface (it really doesn't add significantly to the size!), but...... | 14:01.28 |
tor8 | Robin_Watts: now that you added FILE* to all print flags, perhaps "mudraw -o foo.txt [.xml, .html]" should invoke the text output device, as well as allowing multiple -o flags? | 14:04.04 |
| or at least enough so we can do "mudraw -5 -o out.txt" in sane | 14:04.46 |
Robin_Watts | invoking the text output device - nice. | 14:04.48 |
| multiple -o flags... maybe, yes/ | 14:05.02 |
tor8 | so we can get both text and bitmap checksum out in one pass | 14:05.11 |
DerDa_DE | Hey everyone,.. could someone just explain me why I get an error "ESP GhostScript 8.15.3: Unrecoverable error, exit code 1" when I have installed the current version of GPL GS?,... are these any plug-ins or something like this...? | 14:27.35 |
chrisl | DerDa_DE: are you running on Linux/UNix? | 14:29.08 |
DerDa_DE | yeah | 14:29.15 |
chrisl | Okay, can you do "which gs" and post what it says, please? | 14:29.31 |
| Oh, and confirm you installed GPL GS from *our* release archive..... | 14:30.19 |
DerDa_DE | Chrisl: /usr/local/bin/gs | 14:31.15 |
| Chrisl: Yes it's the "original" release | 14:31.38 |
| It's 9.05 | 14:32.14 |
chrisl | Hmm, well, the path looks right. If you give the full path to the exe, does it still report the ESP version? | 14:32.55 |
DerDa_DE | wait a second have to restart the software... | 14:35.06 |
chrisl | DerDa_DE: "restart the software"? | 14:35.35 |
DerDa_DE | Chrisl: it's a the converting Demon pdf-> jpg | 14:36.22 |
chrisl | DerDa_DE: are you using something like imagemagick which then calls gs? | 14:37.14 |
DerDa_DE | I think that was the problem strange... I never installed the esp version... | 14:37.46 |
| Chrisl: no it's a java app making e-papers... | 14:38.14 |
chrisl | It's almost certainly got the path to executable hard coded - you'll probably find the ESP executable is in /usr/bin/gs | 14:38.49 |
DerDa_DE | Chrisl: yeah you are right... just found them... okay thanks for that help... should have come over that by myself... | 14:39.54 |
chrisl | DerDa_DE: np. I these circumstances, I usually rename the original exec to something like gs-esp, and then make a link from /usr/bin/gs -> /usr/local/bin/gs | 14:40.52 |
DerDa_DE | Chrisl: the only strange think is that if I enter the commandline and use it with the path and go into commandline modus and make an --version it also says that it is gpl version... mh strange... | 14:42.07 |
chrisl | DerDa_DE: as I said, the java app probably has the path hard coded (rather than using your user search path), *or* it might be configurable..... | 14:43.21 |
DerDa_DE | Chrisl okay thanks... | 14:44.08 |
| Have a nice day... | 14:44.19 |
tor8 | Robin_Watts: we may want to add document metadata properties to fz_document (such as title, author, etc) | 14:56.41 |
| but that can wait until post 1.0 IMO | 14:56.51 |
Robin_Watts | tor8: An interface for retrieving such. And yes, post 1.0 | 14:59.50 |
tor8 | I suggest char *fz_document_property(fz_document *doc, char *s) | 15:00.36 |
| but yes, post 1.0 | 15:00.50 |
Robin_Watts | Or a more general metadata interface in the style of gxdso. | 15:04.54 |
| paulgardiner: ping? | 15:07.09 |
| SumatraPDF builds MuPDF into a windows PDF viewer, with some extensions of their own. | 15:07.31 |
| They provide a patch between our code and theirs which we are invited to take stuff back from: http://software.zeniko.ch/sumatrapdf/SumatraMuPDF.patch | 15:08.09 |
| That includes a whole bunch of code for synthesising appearance streams. | 15:08.28 |
| (About 1/2 way through, search for: "synthesize appearance streams for a few more annotations") | 15:09.03 |
| paulgardiner: http://bugs.ghostscript.com/show_bug.cgi?id=692078 may also be of interest | 15:10.32 |
paulgardiner | thanks | 15:33.57 |
Robin_Watts | tor8: Have you started to look at the xps stuff yet ? | 15:37.14 |
tor8 | Robin_Watts: in the middle of it | 15:39.06 |
Robin_Watts | OK, I just reached the xps specifics in the patch :) | 15:39.30 |
| So i'll look at the mupdf bugs. | 15:40.37 |
henrys | kens:don't know if I completely understood what you wanted leak wise yesterday. All but one of the current entries reported by memento are one time allocations which I don't consider leaks. | 15:46.24 |
kens | Hmmm. | 15:47.26 |
| I saw 2.9 Mb reproted by Memento. | 15:47.41 |
| I will have to find some way to determine which ones are not leaks so I don't chase them with pdfwrite. | 15:47.59 |
Robin_Watts | kens: If they are 1 time allocations, then presumably they are allocated at start up. | 15:48.25 |
kens | No idea, a lot of them didn't seem like that to me. | 15:48.42 |
henrys | yes and if you repeat owl owl 4 times on the command line and shouldn't change much. That's what I'm going to fix - the delta. | 15:48.42 |
Robin_Watts | So you should be safe to ignore all num < X, where X is a number to be given by henrys. | 15:48.56 |
kens | Umm, OK if that works, that's fine | 15:49.13 |
| I have to go to the opticians, back later | 15:49.24 |
henrys | unfortunately there is more than that font cache entries are not freed. the font cache simply fills up. | 15:49.54 |
Robin_Watts | tor8: RadialGradientBrush.xps gives strange results with your latest commits. | 15:50.55 |
| And RenderTransforms.xps changes in a 'not obviously better' way. | 15:53.29 |
tor8 | Robin_Watts: yeah, but RadialGradientBrush was wrong before too :) | 15:54.46 |
Robin_Watts | XPS_Examples.xps looks wrong too. | 15:54.51 |
| Before it didn't repeat. Now it has holes in. | 15:55.08 |
tor8 | the reflected(?) gradient in RenderTransforms does look odd | 15:55.24 |
Robin_Watts | ResourceDictionary.xps loses the tip of a diamond. Dunno if that's me or you though. | 15:56.21 |
henrys | Robin_Watts:is their one stash in git per branch? | 15:57.12 |
Robin_Watts | No, you can have many many stashes. | 15:57.26 |
| It's a stack. | 15:57.33 |
ray_laptop | henrys: you can see them with git stash list | 15:57.52 |
henrys | okay thanks | 15:58.04 |
Robin_Watts | wot ray_laptop said :) | 15:58.04 |
| ray_laptop: Are you coming to the next meeting in London? Karen well enough ? | 16:02.00 |
ray_laptop | chirsl: I noticed that you and ken were helping a guy (ManDay) that wanted backtrace into the PS. I wanted to mention -dESTACKPRINT, so if the error bubbles up to the interpreter you can see the tails of the PS procedures -- it's upside down compared to C stack traces 'top' is at the bottom | 16:02.57 |
| Robin_Watts: she has a visit to the surgeon in about 1/2 hour so we hope to find out how long the casts will be on and how long before she can drive | 16:04.09 |
Robin_Watts | Ah, right. | 16:04.23 |
ray_laptop | it's been 4 weeks since the surgery | 16:04.55 |
| BTW, Happy Pi Day to all (my kids get a piece of pie at lunch today) | 16:05.45 |
henrys | pi day? the tau folks aren't going to like that. | 16:08.39 |
ray_laptop | but since we have another 6 weeks until the staff meeting, I am planning on it. This is the one time when a meeting mid-week is not preferred. | 16:08.49 |
| since it means she'll be dealing with 4 school days. | 16:09.35 |
chrisl | ray_laptop: it was more a case of trying *not* to help him - but thanks, I'll have to remember that. | 16:10.40 |
henrys | kens:for the logs I don't see any leaks in owl using memento - running the job repeatedly does not increase memory use. to ppmraw I see a total of 59 blocks irrespective of how many times I repeat ow.pcl on the command line. | 16:13.08 |
ray_laptop | chrisl: yeah, I saw the logs -- that's one of the reasons I lobby for disallowing shared libs :-) | 16:15.20 |
henrys | chrisl:somehow I now have the opvp compile failure again - it seems to come and go depending on what I've installed with macports | 16:15.56 |
Robin_Watts | ray_laptop, well I hope the doctors appt goes well. | 16:15.57 |
chrisl | henrys: yeh, nothing I can do about it, sorry | 16:16.33 |
henrys | yeah I know just bitching. | 16:16.46 |
chrisl | henrys: http://www.ghostscript.com/doc/9.05/Make.htm#MacOSX | 16:17.48 |
ray_laptop | Robin_Watts: thanks. | 16:18.21 |
henrys | oh I never noticed the LDFLAGS trick. | 16:18.49 |
| thank you. | 16:19.02 |
chrisl | I only wrote it a few weeks ago..... | 16:19.05 |
henrys | hah then I missed which is just as bad. | 16:19.37 |
| missed it | 16:19.44 |
chrisl | It was add that to the docs, or throw my mac mini out the window - I spent quite a while swearing at it...... | 16:20.33 |
henrys | mvrhel:the valgrind problem alexcher reported yesterday does look like a real problem - reading past the end of a string, it isn't the typical valgrind warning we normally ignore. | 16:21.59 |
mvrhel | yes. I was going to take a look at that | 16:22.20 |
| at the time though that was not the source of my problems, which ended up being the long casting in tiff code | 16:23.50 |
henrys | right | 16:24.08 |
mvrhel | I will look at the issue that alex raised today | 16:25.29 |
henrys | chrisl:the LDFLAGS does indeed fix the opvp problem now lcms is screwy. | 16:26.19 |
chrisl | Screwy? | 16:26.41 |
henrys | well I wanted to confirm my tree is current but casper is down. This is starting out to be a bad day. | 16:27.45 |
chrisl | casper isn't down | 16:28.14 |
Robin_Watts | henrys: Oh no it's not</pantomine> | 16:28.15 |
| Pantomime, even. | 16:28.27 |
| henrys: You may need to remove the known_host key. | 16:28.46 |
henrys | yes I do need to do that - but ping doesn't work - gives a request timeout - has ping always be disabled? | 16:29.57 |
Robin_Watts | What address is it giving? | 16:30.11 |
| Yeah, I get no ping too. | 16:30.32 |
henrys | 184.73.189.105 | 16:30.34 |
Robin_Watts | same as me. | 16:30.42 |
chrisl | It may be that Ubuntu have changed to defaulting ping disabled | 16:31.08 |
henrys | that's fine I think it is commonly disabled. | 16:31.15 |
| these days. | 16:31.20 |
Robin_Watts | Would you rather have it enabled? | 16:31.22 |
ray_laptop | for msys you need to update the putty host list | 16:32.20 |
henrys | chrisl:upon link I see "undefined symbols for architecture x86_64" then a bunch of cms functions listed. | 16:32.37 |
Robin_Watts | Hmm. /proc/sys/net/ipv4/icmp_echo_ignore_all says 0, which means enabled though... | 16:32.51 |
ray_laptop | maybe amazon isn't allowing pings ;-) | 16:33.17 |
chrisl | henrys: we shouldn't be using libraries for lcms, we should be using our "local" source. | 16:33.40 |
henrys | indeed | 16:33.52 |
| just did the default build nothing fancy. | 16:34.29 |
chrisl | clean it, and try again? | 16:34.50 |
henrys | compiling now | 16:36.33 |
| yes it's okay now I thought an autogen.sh would force everything to rebuilt but I guess we have dependency issues. | 16:37.43 |
chrisl | I don't think everything is made dependent on Makefile, and they really ought to be | 16:38.29 |
mvrhel | bbiaw . have to help out at kids school for bit | 16:41.10 |
marcosw | other than the change in ssh id that tor8 mentioned any other issues with the updated casper? | 16:56.33 |
Robin_Watts | marcosw: henrys points out that we can't ping it. | 17:02.59 |
| We're not sure that we ever could, and even whether that's a bad thing :) | 17:03.23 |
marcosw | that's presumably a question of which ports are open; it's likely we never could. | 17:03.51 |
| Robin_Watts and henrys: ping away | 17:09.13 |
Robin_Watts | thanks. | 17:13.19 |
marcosw | apparently tcping is a good alternate to ping. | 17:14.47 |
MegaDS | hello | 17:30.06 |
| can i save a bookmark in mupdf? | 17:30.33 |
Robin_Watts | MegaDS: I don't believe so. | 17:32.50 |
MegaDS | really bad news:( | 17:33.44 |
Robin_Watts | tor8: Do you want the honour of closing 692912 ? | 17:34.21 |
MegaDS | there's no way to save a bookmarks? I really need to bookmark somepage in pdf and goto that page afterward:( | 17:35.02 |
Robin_Watts | MegaDS: Sorry, no. How would you want that to work? | 17:35.44 |
MegaDS | I'm working on a big project and there's alot of detail in pdf file so I think it better if i can bookmark some important page:( | 17:38.19 |
Robin_Watts | Still kinda vague... | 17:38.38 |
| What is it you'd like to be able to do? To follow a link by clicking on it, and then be able to hit a key to go back? | 17:39.36 |
| Or to be able to press a key to 'bookmark' a page, and then somehow select from a list of bookmarked pages to return to ? | 17:40.06 |
| Having the ability to go 'back' after following a link seems sensible/desirable. | 17:41.25 |
MegaDS | yes, i mean to case that i want to bookmark a page, and then select a bookmarked page to return to | 17:42.07 |
Robin_Watts | Right. That would require us to do a new window to show the list of bookmarked pages in. Can't see that happening any time soon. | 17:42.50 |
MegaDS | thank you, i will wait for your work:) | 17:44.18 |
Robin_Watts | MegaDS: The thrust of our work is on writing the core functionality of the library, and the library is clearly capable of this. | 17:45.05 |
MegaDS | I think at this time i have to save the page number to another file instread:) | 17:45.05 |
Robin_Watts | The viewer is really just a simple example of what can be done with it. | 17:45.23 |
| As such, we will accept patches that improve it, but it's not really the focus of a huge amount of our effort. | 17:45.50 |
MegaDS | oh i see | 17:47.44 |
Robin_Watts | tor8: Shouldn't copy_state be copying stroke_state too ? | 18:25.13 |
| (pdf_interpret.c) | 18:25.26 |
| (I'm looking at bug 692917) | 18:25.52 |
henrys | pantomime? some Briticism I must be missing. | 18:35.12 |
Robin_Watts | Yeah, probably. | 18:36.22 |
| I don't think you're missing much. | 18:39.16 |
henrys | 532's mail exceeds my acceptable acronym to word ratio. | 18:46.56 |
chrisl | Even without the acronyms, it seems like borderline gibberish to me....... | 18:47.34 |
henrys | I don't think they have command of the system and that comes across in the writing. | 18:50.55 |
| chrisl:the LDFLAGS is brilliant - fixes everything on the mac server also - don't have to configure with the disable stuff anymore. | 18:52.50 |
chrisl | henrys: great - it was the only solution I could come up with. It is *theoretically* possible for the same problem to occur in "reverse" with that workaround, but I think it's pretty unlikely in the real world. | 18:53.57 |
paulgardiner | Did we say Monday for the next forms/javascipt discussion? Any chance we could make it Tuesday instead, unless that's inconvenient for someone else? | 19:05.38 |
Robin_Watts | You have an engagement on Tuesday evening :) | 19:06.36 |
| but safely later than that :) | 19:06.47 |
paulgardiner | Yes. Looking forward to that. As you say safely later. | 19:07.19 |
henrys | Tuesday is fine by me. | 19:07.26 |
paulgardiner | Ok good. If it turns out to be bad for anyone else, let me know. | 19:08.35 |
Robin_Watts | So on tuesday it'll follow the normal 5pm meeting (which might be 4pm) | 19:09.26 |
| Yeah usual meeting will be at 4pm. | 19:10.04 |
henrys | right and we try to keep the regular meeting to 1/2 hour. | 19:10.16 |
Robin_Watts | tor8: ping me when you return. | 19:14.06 |
paulgardiner | It's just that I yesterday forgot a previously plan to visit my parents. They are, as yet, unaware of the plan, so it's no dissaster if we need to stick with Monday. | 19:23.42 |
Robin_Watts | tor8: ping | 20:19.08 |
tor8 | Robin_Watts: I'm back now. | 20:31.58 |
Robin_Watts | ok. | 20:32.08 |
tor8 | Robin_Watts: we have sort of bookmarks like MegaDS was asking for. the 'm' command saves a page in a register you can then quickly go back to. | 20:32.30 |
Robin_Watts | I think there are a few mupdf bugs that may be closable now you've done the XPS stuff. | 20:32.30 |
| Oh, I stand corrected. | 20:32.44 |
tor8 | it doesn't keep them across invocations | 20:33.03 |
Robin_Watts | I was considering having a simple history stack, so when you follow a link your page is pushed onto the stack. | 20:33.21 |
| So you can hit backspace to return to the previous page. | 20:33.36 |
| but then I lost interest :) | 20:33.54 |
| I'm looking at this 'maximum number of dash entries is 32' thing. | 20:34.10 |
| I have fz_stroke_state as a ref counted dynamic object. | 20:34.32 |
| so gstate bumps the refs, grestore decrements them, the list device 'keeps' them. | 20:34.57 |
tor8 | we also have that (history stack) | 20:35.00 |
| if you follow links they get pushed and you pop back with 't' | 20:35.13 |
Robin_Watts | And we 'unshare' before every op that changes them. | 20:35.16 |
| Ah, nice. | 20:35.23 |
| where unshare will make you a new one if the old one has refs > 1. | 20:35.50 |
| It works for PDF files, but it's SEGVing for XPS files - fixing that now. | 20:36.32 |
tor8 | so '5m' will save the current page in register 5, and '5t' will go to the page remembered in register 5. m and t with no number will just posh/pop the history | 20:36.38 |
Robin_Watts | urgh. Are you a vi user? :) | 20:37.07 |
tor8 | okay (to stroke state being dynamic) | 20:37.11 |
| you can tell? ;) | 20:37.21 |
| Robin_Watts: I read your emails and back about the metro thing. | 20:41.46 |
juster1 | hi all | 20:42.34 |
Robin_Watts | tor8: Did I get anything obviously wrong ? | 20:43.08 |
tor8 | nope, but it looks like he's getting everything wrong... | 20:43.32 |
| my bet is he didn't wrap the headers with extern "C" from his c++ sources | 20:43.55 |
Robin_Watts | The functions are defined __cdecl according to one thing. | 20:44.28 |
tor8 | and the win32 kernel calls must come from libc or something else he's linked in | 20:44.36 |
| right. that suggests they're compiled as C. | 20:44.47 |
| but then his code tries to call them with C++ mangling, suggesting the headers have been compiled as C++ | 20:45.01 |
Robin_Watts | I don't read the errors that way. | 20:46.15 |
| WinRTComponent.obj : error LNK2019: unresolved external symbol "void __cdecl fz_push_try(struct fz_error_context_s *)" (?fz_push_try@@YAXPAUfz_error_context_s@@@Z) referenced in function "public: virtual int __cdecl mupdfWinRt::WinRTComponent::[mupdfWinRt::__IWinRTComponentPublicNonVirtuals]::openPdfFile(char *)" | 20:46.23 |
tor8 | oh. | 20:47.02 |
| then I have no idea. sorry. | 20:47.12 |
Robin_Watts | It's as if the name is both mangled and not mangled at the same time. It's an odd one. | 20:47.21 |
| but I do reckon he's screwing the pooch with his dll making. | 20:47.59 |
| and linking libcrt in. | 20:48.07 |
juster1 | I would like to ask a question to mupdf developers if you are here | 20:53.11 |
Robin_Watts | We are. | 20:53.19 |
juster1 | is it hard to implement color inverting for reading on dark screen.. i mean dark background with bright text? | 20:54.45 |
Robin_Watts | juster1: It's trivial. | 20:54.59 |
juster1 | that wuld be nice feature to have, i just ditched evince for mupdf and thats the only feature I miss a bit | 20:55.34 |
Robin_Watts | We export a function from the lib to invert either a whole pixmap or a rectangle, so viewers can easily do it. | 20:55.41 |
juster1 | is there any chance that it will be implemented in mupdf? | 21:03.43 |
Robin_Watts | I can't promise it. If you do it, give us a patch, and we'll consider taking it on. | 21:10.29 |
juster1 | I'm not so skillful programmer, and I'm still learning C from cs50.net.. i don't know will I be able to help, but would like to one day | 21:13.43 |
Robin_Watts | Balls. Just had to force push a fix to mupdf, sorry. Was only about 15 seconds... | 21:20.34 |
sebras | good evening. | 21:57.31 |
| Robin_Watts: now that you added the description of multi-threading to doc/overview.txt do you think it is worthwhile to add a simple POSIX-based example like my mt-examples.c to illustrate? | 22:39.24 |
| (haven't read it all yet, just wanted to know your opinion). | 22:39.41 |
Robin_Watts | sebras: Yes, I think that's worthwhile. | 23:58.27 |
| Forward 1 day (to 2012/03/15)>>> | |