| <<<Back 1 day (to 2015/12/17) | 20151218 |
Robin_Watts | mvrhel_laptop: I shall remove the pdf_create branch from golden. | 00:32.36 |
ome | Also, as a suggestion, I think it would be better to use git subtrees instead of git modules. | 00:41.50 |
mvrhel_laptop | Robin_Watts: for the logs, thanks! | 06:50.58 |
kens | Robin_Watts : ping | 09:10.44 |
Robin_Watts | pong, brieflt :) | 09:22.13 |
kens | Umm might need some time, I can't get plink to work with Msys any more | 09:22.34 |
| If you're going out it can waiut untile later | 09:22.46 |
Robin_Watts | export GIT_SSH=plink.exe ? | 09:22.58 |
kens | Its not that | 09:23.03 |
Robin_Watts | Ok, will have to talk later then, sorry. | 09:23.33 |
kens | If I run a windows command shell and do plink casper, it works, if I do plink casper from Msys I get hte 'no authenticaton method' error | 09:23.37 |
| Later is fine, no problems | 09:24.01 |
Robin_Watts | Is it finding the same versions of plink? | 09:24.05 |
kens | good question let me check | 09:24.14 |
| Yes | 09:24.29 |
| I did run a debug session of plink and it said 'Pageant is running. Rwquesting keys." | 09:25.05 |
| THen asked for login, then barfed | 09:25.17 |
Robin_Watts | Different user names? | 09:25.27 |
kens | I used the same user name as windows command shell | 09:25.38 |
| Which worked | 09:25.45 |
| Shout at me when you get back Robin | 09:27.03 |
| Ah, some clues. When I run plink under windows and it requests pageant keys, pageant says it has 1 SSH-2 key, under Msys it doesn't report any :-( | 09:29.35 |
chrisl | Hmm, I used the ssh that comes with msysgit.... | 09:33.43 |
kens | https://opensource.org/licenses/AGPL-3.0I've always used pageant and its been fine | 09:33.56 |
| until now obviously | 09:34.11 |
chrisl | I assume you told the msysgit installer it was to use plink? | 09:34.41 |
kens | I did, and it is doing so | 09:34.55 |
| but plink form the msys bash window is not working | 09:35.10 |
| Apparntly (see above) because pageant is not reporting any keys | 09:35.21 |
| Despite that fact that it *does* when run from the windows command shell | 09:35.35 |
Robin_Watts | Are you running more than one pageant process? | 09:36.18 |
kens | No | 09:36.26 |
| You can't, it won't let you | 09:36.32 |
| Annoyingly if I start pageant from msys it never returns control to bash | 09:37.10 |
chrisl | Even if you use & ? | 09:37.35 |
kens | Dunno, how do you use & ? | 09:37.45 |
chrisl | Append it to the command line - it runs the command in the background | 09:38.14 |
kens | OK I'll try that | 09:38.20 |
| Yeah thst does return | 09:38.43 |
| and pageant is still running | 09:38.53 |
| Intriguingly, now my windows command shell has no keys...... | 09:39.17 |
| But msys does O.O | 09:39.35 |
| However, I then get 'Trying pageant key 0' | 09:40.02 |
| 'Server refused public key' | 09:40.12 |
chrisl | What happens if you specify a key explicitly running plink from msys bash? | 09:40.46 |
kens | I did that | 09:40.52 |
| Well more accurately I did with pageant | 09:41.03 |
| Can you specify a key to plink ? | 09:41.12 |
| ah -i | 09:41.22 |
chrisl | -i <key file> apparently | 09:41.24 |
| You may also need to tell it to use ssh for the connection | 09:41.44 |
kens | Hmm I may be using the wrong key | 09:42.42 |
| Nope | 09:43.10 |
| If I use the .ppk I get 'server refused opur jkey' | 09:43.31 |
| If I use the openssh key then I get 'unable ot use this key' | 09:43.53 |
chrisl | Does either work from the Windows command prompt? | 09:44.45 |
kens | Not sure, give me a moment | 09:45.05 |
| Yeah the .ppk works, though I need to give my passphrase since pageant isn't running | 09:46.23 |
kens | has basically run out of ideas | 09:47.45 |
| I'm running the same executable with the same key file. In Windows its OK in msys the server refuses our key. | 09:48.12 |
chrisl | When you ran it from msys, did you use the full path to the key file? | 09:48.24 |
kens | I used "/plink.exe -v -i keys/<key file> casper" | 09:48.57 |
| The keys being under teh curretn directory | 09:49.07 |
| I'm rtunning from the putty directory so that I cna be sure exactly where the executable is, and don't have to concern myself with spaces in the pahts | 09:49.40 |
chrisl | I'm wondering if the forward slash is confusing things | 09:50.34 |
kens | I cna try running form the keys directory | 09:50.51 |
| Same result. Note that we offer the public key and the server refuses it | 09:51.33 |
chrisl | And does the feedback confirm both sessions are trying the same key? | 09:52.32 |
kens | Up to the point where the server refuses the key the transcript is identical | 09:53.01 |
chrisl | What about explicitly using an ssh connection in msys? | 09:53.38 |
kens | The transcript does say "Using SSH protocol version2" | 09:54.01 |
| BTW the servere refuses the key *before* I get asked for a pass phrase for the key | 09:55.00 |
| Which suggests to me its something plink like that's the problem rather than the actual key | 09:55.30 |
| OK can't use version 1 SSH, can't use IPV6, tried also specifying -noagent -2 -4 (SSH 2 and IPV4) and it still doesn't work | 09:58.07 |
| I'm getting quite baffled by ths | 09:58.41 |
| Oh wow, I thnk I'm in | 09:59.40 |
kens2 | Network hiccup :-( | 10:03.37 |
| If I specify my username on the plink command line then the servere accepts the key offer, but plink fails my passphrase. If I don't specify user@host then it fails the offer of a key. Huh ? | 10:04.34 |
chrisl | Your local username must not match your username on casper | 10:05.15 |
kens2 | It does | 10:05.21 |
tor8 | does plink have the -vvv option to debug connections? | 10:05.42 |
kens2 | And it doesn't matter, because both ways I *specify* the user name | 10:05.42 |
| tor8 it has -v, but not -vvv | 10:05.56 |
tor8 | shame, with three v's it lists all the authentication methods it tries, IIRC | 10:06.13 |
kens2 | chrisl one way I specify the username as user@host, the other way it asks me for the user name. In neither case does it use my local user name | 10:06.42 |
| tor8 it only seems to try one | 10:06.55 |
| If I use peeves it will try passwords as well though | 10:07.05 |
| Oh that's interesting | 10:07.49 |
tor8 | kens2: casper is set up to not allow password logins | 10:08.30 |
kens2 | If I use peeves, then when it fails public key we try passwords. The password prompt is 'ke@peeves.ghostscrtip.com' when I typed (and I can see it) ken | 10:08.37 |
tor8 | cr/lf stripping gone wrong? | 10:09.16 |
kens2 | Yes, its dropping the last character | 10:09.20 |
| How odd | 10:09.24 |
| and if I say the user name is kens it accepts the offer of the key..... | 10:09.52 |
| But plink fails the passphrase | 10:10.15 |
tor8 | kens2: I set up my msysgits to not use putty, then I can just reuse my linux .ssh folder settings | 10:10.30 |
kens2 | It sort of looks like thw two ends are confused in some way | 10:10.35 |
tor8 | kens2: so I can't really help you out in any meaningful way here :/ | 10:10.54 |
kens2 | tor8 I'd then have to type my passphrase every time I open Bash | 10:10.57 |
| tor8 yeah I was hoping Robin could tell me if he has some magic in his .bashrc | 10:11.19 |
tor8 | don't use a passphrase? ;) | 10:11.34 |
kens2 | Poor security :-P | 10:11.45 |
tor8 | kens2: non-putty msysgit ships with an ssh-agent.exe | 10:13.34 |
kens2 | But I'd satill have to enter the passphrase every time I start msys git | 10:13.58 |
| Normally I enter it into pageant on startup and that's it. | 10:14.15 |
| I'm thnking I was better off with the older version of Msys, that actually seemed to work...... | 10:14.45 |
| If only I could remember what version that was :-) | 10:16.03 |
tor8 | kens2: wouldn't ssh-agent also stay on in the background? | 10:16.38 |
kens2 | When I close msys ? I don't know but I woudl assume not | 10:17.22 |
tor8 | it's probably not going to work well due to how it needs environment variables | 10:20.02 |
| to communicate the ssh-agent port | 10:20.12 |
kens2 | Yes that doesn't sound like it'll work | 10:20.43 |
| And it looks like they don't keep old binaries around for download, so I can't even go back. Damn. | 10:21.22 |
tor8 | https://github.com/git-for-windows/git/releases/tag/v2.5.3.windows.1 looks like the oldest I can find | 10:22.28 |
kens2 | I'll give that a try | 10:22.47 |
tor8 | or https://github.com/msysgit/msysgit/releases | 10:23.20 |
kens2 | Though whether it wil install over a newer version is debatable :-) | 10:23.20 |
tor8 | before they changed names | 10:23.28 |
| I use the portablegit thing, but that's also unlikely to help you :) | 10:23.48 |
kens2 | I'll try 2.5 first | 10:23.51 |
| Nope, still can't find my pageant keys | 10:26.42 |
| Lets try a still older version | 10:26.56 |
kens2 | is pretty sure Robin said he was using 2.5.3 whcih makes me thnk its a config problem | 10:27.44 |
| Well, using 1.9.2, disabling pageant and supplying the key file works | 10:32.37 |
| But pageant doesn't :-( | 10:32.50 |
| I'm going to re-install 2.5.3 and see what Robin has for config when he comes back | 10:33.20 |
| OK if I *don't* use MinTTY but instead use the WIndows command window, then manually suplpying a key file works (as does the passphrase) but pageant doesn't. Now that could be a user name problem | 10:39.39 |
| However whoami says I'm ken | 10:40.28 |
| I have a sneaky suspicion I had something in .bashrc before which has been trashed | 10:40.53 |
chrisl | try echo $USER | 10:41.05 |
kens2 | says nothing | 10:41.17 |
| THat could be the problem | 10:41.25 |
chrisl | That is probably the problem, then | 10:41.28 |
| "export USER=ken" and try again? | 10:42.02 |
kens2 | Tried that, pageant still not working | 10:42.16 |
chrisl | I guess you'll have to swap notes with Robin - I never, ever got putty/plink/pageant to work - one of the many reasons I ended up using cygwin on Windows | 10:44.42 |
kens2 | Hmm, seems I do have a .bashrc whcih sets GIT_SSH and CLUSTER_USER | 10:44.47 |
| I've not had problems with pageant before :-( | 10:45.01 |
| However, there's clearly something badly broken in MinTTY...... | 10:45.14 |
chrisl | It worked find with the msys ssh, but that's not much help | 10:46.00 |
kens2 | I can probably make that work, but its a pain | 10:46.36 |
kens | Well I've got openssh working, mostly. I cna now run regression etst, but I have to enter my passphrase lots of times | 11:34.30 |
| ssh-agent works, but ssh-add doesn't, says if cnnot open a connetction to your authentication agent | 11:34.53 |
Robin_Watts | kens: back. | 11:40.00 |
| What is the exact plink command you are running? | 11:40.24 |
kens | 1 sec | 11:40.25 |
| plink -v casper.ghostscript.com | 11:40.56 |
| If I instead go to the plink/keys subdirectory where I keep my keys I can run plink --noagent -v -i <keyfile> casper.ghostscript.com | 11:42.12 |
| And that will work | 11:42.20 |
| Well, it will now I stopped using MinTTY | 11:42.48 |
Robin_Watts | plink -v ken@casper.ghostscript.com ? | 11:43.07 |
| Is the problem just within MinTTY ? | 11:43.36 |
kens | Using username "ken" Siconnected: No supported authentication methods available | 11:43.49 |
| No there wqere 2 problems | 11:43.56 |
| FIrstly with MinTTY it was buggering up my user name | 11:44.07 |
| Which I noticed when using peeves and it fell back to password authentication | 11:44.24 |
| If I use the other option (windows console or something) then that does not happen | 11:44.44 |
| With MinTTY I could not get plink to work under any conditions. With the soneole I cna get it to work if I specify the key file | 11:45.11 |
| Of course I have to enter the passphrase that way | 11:45.21 |
Robin_Watts | That's very odd indeed. | 11:45.42 |
kens | I know :-( | 11:45.49 |
Robin_Watts | I never instalk plink etc as part of msys. I have my own installation of them. | 11:46.07 |
kens | And if I run plink form WIndows command prompt it tells me I have a SSH-2 key in pageant, while if I run it form Msys it doesn't | 11:46.15 |
| plink isn't (or shouldn't be) installed as part of Msys | 11:46.33 |
Robin_Watts | And it's definitely the same plink binary? | 11:46.41 |
kens | My puTTY installation long predates any msys on here | 11:46.49 |
Robin_Watts | What version of plink/putty etc ? | 11:47.02 |
kens | Robin_Watts : I was running the binary directly from the directory using ./ | 11:47.06 |
| 0.60 | 11:47.09 |
Robin_Watts | You *could* try upgrading those. I believe 0.66 is out now. | 11:48.06 |
kens | Its worth a try I guess | 11:48.16 |
| I thnk I'll put a copy of my keys somewhere first though | 11:48.29 |
Robin_Watts | pageant has to be started with the keys on the command line, right? | 11:48.49 |
kens | Yes, and it is | 11:49.02 |
| Well, the one and only key I'm actually using right now | 11:49.18 |
Robin_Watts | Me, I'd just download new exes, rename the old ones to be pageant060.exe and put pageant.exe in its place. | 11:49.33 |
kens | I'm going to rename the folder | 11:49.49 |
Robin_Watts | So no uninstallation/reinstallation, nothing to lose keys. | 11:49.53 |
kens | *and* copy the keys | 11:50.02 |
| I really don't want to have to dance through getting new keys done | 11:50.18 |
| OK well new PuTTY installed | 11:53.27 |
| can still plink form windows (have aset up pagenat again) | 11:54.10 |
| and now Msys works as well..... | 11:54.53 |
kens | is baffled, but happy | 11:55.02 |
| and git works as well | 11:55.37 |
| I suppose I could try upgrading to 2.6.4 again....... | 11:55.48 |
| How hard could it be ? :-( | 11:56.02 |
Robin_Watts | kens: Fab. | 11:56.30 |
| Personally, I'm using kitty etc rather than putty most of the time these days. | 11:56.56 |
| http://www.9bis.net/kitty/ | 11:57.13 |
kens | will use anything that works | 11:57.14 |
| Well it looks like my original mistake was to allow Msys to use MinTTY, somehow that also broke Pageant, though I have no idea how | 12:00.06 |
| Since everything now appears to be working again I'm going to count my blessings and stopmeddling. Since I have 2.6.4 again hopefully I can still use notepad2 for my editor | 12:01.08 |
| Right, time for lunch | 12:15.31 |
tor8 | Robin_Watts: bah, complete_signatures() reopens the PDF output file to do read/write operations :( | 13:55.41 |
kens | I thnk it has to | 13:55.55 |
| Doesn't it compute stuff based on the contents ? | 13:56.10 |
| Or a byte range anyway | 13:56.22 |
tor8 | yeah, but I'm trying to make it work using our fz_output that can write to non-files | 13:56.41 |
| which is of course impossible when there's a post-processing pass that goes and fopens the same file in the middle of writing it | 13:57.06 |
kens | Well as long as its eekable that probably possible, if it isn't seekable, I don't really see that its possible | 13:57.14 |
tor8 | our fz_outputs are seekable, but not readable | 13:57.33 |
kens | Well that would break it too | 13:57.44 |
| It might be hard to linearize a PDF file too that way | 13:58.03 |
| I thnk you need to rewrite some data near the top of the file, based on information you don't have until the end of the file | 13:58.36 |
tor8 | of course, we could do a massive overhaul and make our fz_streams both for reading and writing but that starts to get hairy with the chained pipeline stuff for compressing then a85-encoding then encrypting, etc | 13:58.43 |
kens | That's the sort of reason that signatures and linearisation are usually done as a post-process | 13:59.25 |
Robin_Watts | tor8: fz_output can seek, but not read? | 13:59.35 |
tor8 | Robin_Watts: correct. | 13:59.43 |
| and fz_stream can seek, but not write | 13:59.49 |
Robin_Watts | We could offer a 'reopen output as stream' method. | 14:00.17 |
kens | I suppose the signature (and linearisation) code could capture what it needs during the write, tehn seek back and wreite up the result. Seems hard work though | 14:00.36 |
tor8 | ah, but I need to fix fz_output on fz_buffers to be able to seek | 14:00.48 |
Robin_Watts | kens: the linearisation stuff already works that way. | 14:00.51 |
kens | Fair enough, the GS one is totally post-process, just like Acrobat | 14:01.13 |
| Right, that should address Marcos' concerns about error messages | 14:03.38 |
| On to PDF/A-1b | 14:03.58 |
tor8 | Robin_Watts: 5 commits on tor/master. the first three should be simple enough, the latter two might need some talk. | 14:05.04 |
| I did run into a nasty hack paul did to get things working, with how the 'this' binding is set up for javascript snippets run from pdf | 14:05.53 |
| 'this' is supposed to be bound to the Doc object for the document | 14:06.49 |
| but in general, 'this' is bound to the global environment for non-method calls | 14:07.01 |
| so there was this hack where the Doc object was merged with the global environment >.< | 14:07.26 |
Robin_Watts | tor8: I reviewed those 4 yesterday. See the logs for my comments. | 14:08.12 |
| s/those/the first/ | 14:08.24 |
| I still prefer the idea of write_document being an fz thing rather than a pdf thing. | 14:10.18 |
tor8 | Robin_Watts: about the get_pixmap thing, I just want to be extra clear about the ref counting-ness | 14:11.14 |
Robin_Watts | {pdf,fz}_new_.... implies to me that the thing you get back has a ref count of 1, and you're free to mess with it. | 14:11.58 |
| Which is not the case for the stuff returned by fz_get_pixmap. | 14:12.24 |
tor8 | yeah, so your point about get_pixmap returning a potentially shared reference means the 'new' name for it is wrong | 14:12.24 |
Robin_Watts | Yeah. | 14:12.30 |
| We should leave it as fz_get_pixmap, and fix fz_new_pixmap_from_image to do a get, then if the ref count != 1, do a clone. | 14:12.59 |
tor8 | I don't think we need fz_new_pixmap_from_image as such | 14:13.27 |
Robin_Watts | tor8: That function exists already, I think. | 14:13.39 |
tor8 | not sure why we had that function before, nobody called it | 14:13.40 |
Robin_Watts | ah, then just lose it :) | 14:13.53 |
| Ok, the last commit looks OK too. It's not the direction I would go, but it's reasonable. | 14:16.42 |
tor8 | fz_pixmap_from_image? | 14:17.10 |
| if we ever decide to add edit capability to xps, we can put fz_save_document back | 14:17.38 |
Robin_Watts | tor8: fz_claim_pixmap_from_image ? | 14:23.26 |
| That contains the implication that we don't own it. | 14:23.49 |
tor8 | I was thinking of 'load_pixmap_from_image' but that also implies the same as 'new' | 14:24.48 |
| just fz_keep_pixmap_from_image could do as well | 14:25.19 |
| but claim does seem a bit clearer | 14:25.39 |
| or 'get' but then we still have the same ownership unclarity | 14:26.44 |
jogux | the mupdf mirror on github is missing the branches I think, eg. release_1.8 - https://github.com/ArtifexSoftware/mupdf (was it chrisl that set it up, I forget...) | 15:02.55 |
chrisl | jogux: er, yes. I wasn't aware we need branches - although, I'm not sure why it should be missing them | 15:04.15 |
Robin_Watts | jogux: We don't need no stinkin' branches. | 15:04.40 |
jogux | chrisl: yeah, I'm not sure we previously had one. We ended up using/needing one for the 1.8 iOS release. | 15:04.42 |
chrisl | jogux: for releases, we tend to use the tag | 15:04.58 |
jogux | chrisl: yeah. there was a fixup necessary for iOS though and master had already moved along | 15:05.16 |
chrisl | And the tags aren't up to date either.... I need to check this out | 15:05.35 |
jogux | chrisl: d'oh, didn't spot that | 15:05.50 |
| Robin_Watts: http://git.ghostscript.com/?p=mupdf.git;a=summary - possibly there shouldn't be a robin_master? :) | 15:06.03 |
Robin_Watts | jogux: Almost certainly not, but I'm going to claim innocence on that, cos I don't have such a branch locally. | 15:06.53 |
jogux | hehe | 15:06.59 |
| Branch_d4ab3cb0b857699c523272bb192752041c77e5e7 smells a bit too | 15:07.13 |
Robin_Watts | OK, look now? | 15:10.04 |
jogux | that looks cleaner, thanks | 15:11.07 |
| myWinRT has a smell about it too. | 15:11.16 |
Robin_Watts | I'll talk to mvrhel when he appears. | 15:13.02 |
| cos that's one of his. | 15:13.06 |
mvrhel_laptop | whats that | 15:13.19 |
| oh what would that be doing on there | 15:14.09 |
jogux | :-) | 15:14.14 |
tor8 | mvrhel_laptop: my brain is too friend to read font code today, but I will look over your code next week | 15:14.42 |
| s/friend/fried/ | 15:14.53 |
mvrhel_laptop | I guess yesterday was not the first time I did an accidental push | 15:15.01 |
| tor8: no problem | 15:15.09 |
tor8 | Robin_Watts: have you tried building the ios app recently? I worry that the 'make third' step in build_libs.sh may be trying to build glfw. | 15:20.26 |
| I don't see how, though. to build curl and glfw you need to run 'make extra' | 15:21.15 |
Robin_Watts | tor8: I don't do ios. | 15:21.23 |
jogux | fredross-perry built it only a few days ago | 15:21.33 |
Robin_Watts | but if you could check that'd be great. | 15:21.37 |
tor8 | I don't own any mac hardware | 15:22.02 |
Robin_Watts | mvrhel_laptop: Should I remove myWinRT from golden? | 15:22.19 |
jogux | hm, I just tried mupdf from golden master: | 15:25.27 |
| pdf_write_document(ctx, idoc, tmp, &opts); | 15:25.36 |
| MuDocumentController.m:112:5: error: implicit declaration of function 'pdf_write_document' is invalid in C99 [-Werror,-Wimplicit-function-declaration] | 15:25.56 |
| pdf_write_document(ctx, idoc, tmp, &opts); | 15:25.56 |
| my SOT build has finished now though, back to that. | 15:26.44 |
tor8 | jogux: no error about pdf_specifics missing a declaration too? | 15:31.26 |
| will need to add #include "mupdf/pdf.h" to that file | 15:31.43 |
jogux | tor8: that was the only error. don't see anything about pdf_specifics. | 15:38.58 |
chrisl | jogux: the github mirror should be fixed now | 15:39.23 |
jogux | chrisl: awesome, thanks | 15:39.34 |
chrisl | Just so no one gets uppity, I'm going to push a branch to mupdf golden, to make sure it gets mirrored - and then I'll delete it, and make sure it gets un-mirrored | 15:41.22 |
mvrhel_laptop | Robin_Watts: yes | 15:43.30 |
| bbiaw | 15:44.34 |
Robin_Watts | mvrhel_laptop: Buzz me when you get back, please. | 15:54.47 |
| chrisl: You about? | 16:41.57 |
chrisl | Robin_Watts: yes | 16:42.12 |
Robin_Watts | chrisl: I've just pushed a gs commit to robin/master. | 16:42.29 |
| It tweaks the memento handling in jbig2dec to alleviate bug 696183. Care to comment on it? | 16:43.00 |
| If you're happy, I'll push it, then when it regenerates the external jbig2dec repo, I'll update mupdf with the matching commit. | 16:43.52 |
chrisl | Well, okay. I'm still tempted to say let's not have memento in jbigdec at all | 16:44.06 |
| But beyond that, it looks okay | 16:45.08 |
Robin_Watts | chrisl: I think I added it for some reason. Possibly for shelly? | 16:45.29 |
| I must have had a reason to put it in. | 16:45.37 |
chrisl | Possibly | 16:45.41 |
Robin_Watts | Thanks. | 16:45.44 |
jogux | chrisl: odd... the branches are there now, but seem to point to a different place?! https://github.com/ArtifexSoftware/mupdf/commits/origin/release_1.8 vs http://git.ghostscript.com/?p=mupdf.git;a=shortlog;h=refs/heads/release_1.8 has a different commit at the top | 17:28.02 |
| unless github's web i/f is just lying to me | 17:28.42 |
chrisl | <shrug> | 17:29.06 |
jogux | hm, it's not the web i/f, I cloned and ran logg and get: | 17:30.11 |
| * 39389bb (HEAD, origin/origin/winphone, origin/origin/win_desktop, origin/origin/winRT, origin/origin/release_1.8, origin/origin/myWinRT, origin/origin/master, origin/origin/lcms2, origin/origin/fontconfig, origin/master, origin/HEAD | 17:30.28 |
| * 80949d3 Remove fz_save_document and use pdf_save_document directly instead. | 17:30.28 |
| * 02d5b22 Rename fz_halftone_pixmap to fz_new_bitmap_from_pixmap. | 17:30.30 |
| * ed1d49f Fix warning in zlib thirdparty build. | 17:30.31 |
| * 15ce216 muPDF iOS app: Support for the latest (3.4.1) Crashlytics. | 17:30.31 |
| so all the branches point at the same place as master :-( | 17:30.52 |
chrisl | jogux: check it now? | 17:36.35 |
jogux | that looks better... but, still something odd. let me just blow this away and reclone. | 17:39.18 |
| that's better. git had done something very odd (I had some origin/origin/release_1.8) but recloning has got rid of them. | 17:40.21 |
chrisl | jogux: so, I'm off work for the next couple of weeks, but if you notice anything more strange with the github mirror, ping me an e-mail, and I'll look at it in a moment of boredom | 18:07.37 |
jogux | okay, will do, ta. AFAICS it looks good now. | 18:07.50 |
| have a good break, Christmas, New Year, etc! | 18:08.03 |
chrisl | You too! | 18:09.02 |
jogux | fredross-perry: fancy reviewing http://git.ghostscript.com/?p=user/joseph/mupdf.git;a=commit;h=42fe18c3a58f75b45d055acdcdc6f36de85e8809 ? It's trivial. for both the release_1.8 branch and master. | 20:51.50 |
| fredross-perry: do you want any of these testflight releases tested btw? let me know when you do :) | 20:52.11 |
fredross-perry | thanks, Iâll let you know when to test. This commit is improtant for 1.8? | 20:52.56 |
jogux | The only other way I can get my podspec to work with 1.8 would be to do something hacky to disable the error and warning | 20:53.40 |
| (so yes, fairly important if we want the cocoapods thing to work) | 20:54.40 |
fredross-perry | jogux: ok, I understand the fz_catch(ctx); thing. But why change <MuDocRef.h> to âMuDocRef.hâ ? | 21:03.01 |
jogux | because the former is wrong - <> is only for OS headers | 21:03.18 |
fredross-perry | ok then, seems legit. | 21:03.28 |
| btw, I think {} is more readble than ; | 21:03.51 |
jogux | me too, but I suspect there are people reading that may disagree so lets keep quiet :-) | 21:04.17 |
fredross-perry | sorry, were you talking? | 21:04.28 |
| lemme know when youâve pushed | 21:05.03 |
jogux | pushed, thanks | 21:06.11 |
fredross-perry | ok | 21:13.54 |
Robin_Watts | fz_catch(ctx); is bad | 21:21.32 |
| fz_catch(ctx) { } is much nicer. | 21:21.42 |
jogux | Robin_Watts: ah, okay, good, thanks :) | 21:21.54 |
| I've pushed that fix to golden too | 21:21.58 |
Robin_Watts | mvrhel_laptop: ping | 21:22.01 |
jogux | fredross-perry: my podspec seems to almost work now - can't quite finish it until we have a release tag for 1.8 for iOS. | 21:22.31 |
fredross-perry | ok | 21:22.47 |
| I think you can go ahead and tag it. Thereâs nothing more I am planning to do in 1.8. Iâm just wrangling with Crashlytics and trying to figure out why my crashes are not getting reported. | 21:25.09 |
jogux | ah, great, so that's all private repo only stuff? | 21:28.02 |
| I'll tag it as 1.8-ios then (that seems to be the pattern we've used in the past)? | 21:29.06 |
fredross-perry | seems good then. | 21:29.28 |
| IIjust need to figure out how to update the release build project so tâs submodule for mupdf points to this updated branch. | 21:30.55 |
mvrhel_laptop | Robin_Watts: sorry I missed you | 21:37.37 |
Robin_Watts | I'm still here. | 21:37.48 |
mvrhel_laptop | oh good | 21:37.55 |
Robin_Watts | So, I've been looking into that clist bug, and I think it's to do with compositors. | 21:38.27 |
mvrhel_laptop | fun! | 21:38.33 |
Robin_Watts | Can I talk to you about it for a bit? | 21:38.37 |
mvrhel_laptop | sure | 21:38.39 |
Robin_Watts | so I've simplified the file a lot. | 21:39.11 |
| when I run through, it does an all band write for compositor. | 21:39.41 |
| and pdf_update_device_color_procs_push_c gets called. | 21:40.02 |
| That sets the clist_color_depth to 8 (instead of 0x20) | 21:40.12 |
mvrhel_laptop_ | oops I dropped for a sec | 21:40.19 |
Robin_Watts | I'll start again... | 21:40.31 |
| so I've simplified the file a lot. | 21:40.33 |
| when I run through, it does an all band write for compositor. | 21:40.39 |
| and pdf_update_device_color_procs_push_c gets called. | 21:40.42 |
| That sets the clist_color_depth to 8 (instead of 0x20) | 21:40.45 |
mvrhel_laptop_ | ok | 21:40.54 |
Robin_Watts | Then we do a copy_alpha, which does a set_color[1]. | 21:41.05 |
mvrhel_laptop_ | is it a soft mask? | 21:41.11 |
jogux | fredross-perry: okay, 1.8-ios tag made and pushed to golden. | 21:41.11 |
Robin_Watts | It is. | 21:41.15 |
| The set_color[1] encodes the color as a single byte because of the depth. | 21:41.48 |
| When we come to read it, we read the compositor, but nothing changes the color depth. | 21:42.00 |
| consequently when we read the set_color[1], it tries to read 4 bytes. | 21:42.15 |
| and hence we get out of sync and crash. | 21:42.22 |
| I couldn't see where in the compositor reading we were *supposed* to be changing the color depth. | 21:43.06 |
mvrhel_laptop | let me look | 21:43.21 |
| Robin_Watts: sorry someone came to the door | 21:59.29 |
Robin_Watts | np. | 21:59.36 |
mvrhel_laptop | so I would expect to see PDF14_BEGIN_TRANS_MASK | 22:00.29 |
| coming from the clist | 22:00.32 |
| i.e. in c_pdf14trans_read | 22:01.33 |
Robin_Watts | Sorry, just having to rebuild, cos I was in the middle of looking at another bug. | 22:02.48 |
mvrhel_laptop | which should end up making the color settings gray | 22:02.55 |
Robin_Watts | OK, I see that being read for band 0. | 22:03.17 |
mvrhel_laptop | but not the band where the set_color is happening? | 22:03.49 |
| I know ray recently rejiggered some stuff with respect to bands and softmaks | 22:04.25 |
| allbands vs not allbands | 22:04.34 |
| I think | 22:04.48 |
Robin_Watts | ok, band 12... | 22:04.58 |
| I get a PDF14_PUSH_DEVICE | 22:05.15 |
mvrhel_laptop | one problem is that the soft mask extent clipping area is much larger than any mask image in certain cases due to the backdrop default | 22:05.44 |
Robin_Watts | and a PDF14_SET_BLEND_PARAMS | 22:05.49 |
| and thats all. | 22:06.32 |
| Neither of them touch the color depth. | 22:06.40 |
mvrhel_laptop | where is the mask? | 22:07.22 |
| those are the only compositor actions you get? | 22:07.43 |
Robin_Watts | Those are the last 2 I get before the crash, yes. | 22:07.56 |
mvrhel_laptop | do you ever get a softmask push from the reader? | 22:08.21 |
Robin_Watts | Where would I check for that? | 22:08.38 |
| PDF14_BEGIN_TRANS_MASK ? | 22:08.55 |
mvrhel_laptop | PDF14_BEGIN_TRANS_MASK line 6009 in gdevp14.c | 22:09.11 |
| this is the switch statement for compositor actions coming from the clist | 22:09.38 |
Robin_Watts | ok, band 11 gets one of them. | 22:09.45 |
| band 11 gets lots of compositor actions. | 22:09.57 |
mvrhel_laptop | what band is the issue in? | 22:10.26 |
| band 12? | 22:10.28 |
Robin_Watts | yes. | 22:10.55 |
| but at no point, even in band 11, does the color depth get set. | 22:11.08 |
mvrhel_laptop | if a soft mask push occurs the color space should change | 22:11.30 |
Robin_Watts | ooh. even in band 11, that routine is going through the 'should not occur' branch. | 22:12.41 |
mvrhel_laptop | ah! | 22:12.48 |
| where is that | 22:12.52 |
Robin_Watts | params.pdf14_op = 6. | 22:12.54 |
| c_pdf14trans_read | 22:13.01 |
mvrhel_laptop | oh so the compositor operation is bogus | 22:14.15 |
| that is an odd one. I have not seen that before | 22:15.01 |
| that would certainly cause things to go badly | 22:15.28 |
| wait PDF14_END_TRANS_MASK is 6 | 22:17.24 |
| I thought you were saying that we were hitting the default case | 22:17.45 |
Robin_Watts | I must have misread the debugger, sorry. | 22:18.35 |
| So ignore the last 10 mins. | 22:18.40 |
| All the way through setting the smask compositor and unsetting it, the color depth never changes. | 22:19.47 |
| (I'm pretty sure it never gets written to, cos I had a breakpoint on it earlier) | 22:20.05 |
mvrhel_laptop | Robin_Watts: ok. if you want I can look this over this weekend | 22:24.05 |
Robin_Watts | mvrhel_laptop: I think I've hit the point where it's going to be hard for me to make progress. | 22:24.27 |
mvrhel_laptop | do you have any suggestions about breakpoints etc | 22:24.48 |
Robin_Watts | So, thanks. | 22:24.50 |
| Only what I put in bug 696364 | 22:25.12 |
| and I'm not sure that's hugely helpful. | 22:26.17 |
mvrhel_laptop | Robin_Watts: very much so | 22:26.37 |
| Thanks! | 22:26.39 |
| this looks like a real ugly one since it is mixing AA in here | 22:27.10 |
jogux | fredross-perry: Damn. First podspec didn't fix everything necessary. I ended up with a MuPDF.a and a mupdf.a, which really upsets things working with case insensitive file systems.... | 23:52.19 |
| fredross-perry: long story short, I don't see a way of getting myself out of the whole I'm in without retagging exactly the same commit as 1.8.1-ios. | 23:52.45 |
fredross-perry | call it 1.8.0.2. Then we can do .3, .4, etc, but itâs still 1.8.0. | 23:53.25 |
jogux | nah, can't :( cocoapods insist strictly on semantic versioning | 23:55.29 |
| ( http://semver.org ) | 23:55.39 |
| they also /insist/ that the tag name contains the version number, and don't let you update a podspec once it's published. (so I should've been more careful, but it seems they've made non-backwards compatible changes since the last time I looked at this which meant my pod that would've worked fine before didn't) | 23:56.57 |
| tbh I also really don't understand how the mupdf.a vs MuPdf.a thing worked last time either, that must've been pure luck | 23:57.51 |
| Forward 1 day (to 2015/12/19)>>> | |