| <<<Back 1 day (to 2014/03/30) | 2014/03/31 |
sebras | ghostbot: time to sleep, good night! | 00:46.40 |
kens | Morning chrisl | 08:12.10 |
chrisl | morning kens | 08:12.28 |
kens | Guess what, I found a reason why we can't default CompressPages to false :( | 08:12.38 |
chrisl | Oh, why?? | 08:12.56 |
kens | Because of the way we process content streams | 08:13.20 |
| We open them with a SubFileDecode filter, and set 'endstream' as teh terminator | 08:13.41 |
| If we have any plain text in the stream which says 'endstream' then it terminates prematurely | 08:14.04 |
chrisl | Hmm, this could cause us problems elsewhere :-( | 08:14.44 |
kens | I can't think of a solution that doesn't involve some heavy rewrites. | 08:15.09 |
| My only current solution is to alter the endstream to something more unique if we have CompressPages false. | 08:15.56 |
| Unless you have a better thought ? | 08:16.19 |
chrisl | The only solutions I can think of would be to change to using the length of the stream, instead of a terminating string. Or scan the stream for instances of "endstream" and set the "Count" value appropriately | 08:16.53 |
kens | scanning the stream sounds expensive. I'm not sure we cna use the length on a subfiledecode filter, can we ? | 08:17.39 |
kens | goes off to consult the PLRM | 08:17.56 |
| Ah, I see we could set EODCount and not bother with an EODStriong | 08:18.42 |
| Well we know the count, and we know its realiable, so that probably will work, let me see if I can get it to. | 08:19.12 |
chrisl | Hmm, no, EODCount is the number of instance of EODString before we stop - that would involve scanning the stream for the string :-( | 08:20.20 |
kens | Oh, still bad then | 08:20.28 |
chrisl | I really thought subfiledecode could also take a byte length, but clearly not :-( | 08:21.04 |
kens | No, there's no provision for it, it seems | 08:21.21 |
| Actually : | 08:22.01 |
| EODString may also be of length 0, in which case the SubFileDecode filter will | 08:22.01 |
| simply pass EODCount bytes of arbitrary data | 08:22.01 |
| So there is provision | 08:22.11 |
chrisl | Oh, that's good - memory isn't failing after all..... | 08:22.39 |
kens | Yeah I just had to read on into the spec | 08:22.54 |
| OK we retrieve the length for the stream, so this should be possible | 08:23.19 |
| (Of course, it means that simple editing of the stream will go badly wrong....) | 08:23.50 |
chrisl | Well, we could take your suggestion and use something like ".ps2write_EndStream" | 08:24.47 |
kens | that's the only other thing I can think of, I'd only want to do it if CompressPages is false, but it would mean that ExecuteStream would have to know which terminating string to use. | 08:25.44 |
chrisl | Actually, I was wondering if we could use a random element if there (given you need to communicate the string to opdfread in some way anyway)? | 08:26.53 |
| s/if/in | 08:27.04 |
kens | Could do, I guess, though I'm not certain if that'sbetter :-) | 08:27.36 |
chrisl | I was thinking that it would be embarrassing if we documented this at some point, and we couldn't convert our own documentation! | 08:29.02 |
kens | Well, that's essentially the problem I ran into this time, it was an example of a PDF stream.... | 08:29.33 |
| Hmm, doing a random element would be harder, I'd need to store the random string and pass it around. | 08:31.25 |
chrisl | Well, just a thought.... | 08:34.15 |
kens | I'm going to assume this will do for now. | 08:34.43 |
Robin_Watts | Morning mace. How goes ATS? | 09:09.22 |
kens | chrisl there turn out to be several places in the pdfwrite code where we emit endstream :-( Rather than meddle with each of them I've decided to use the empty EODString and the stream length. It does, unfortunately mean that when editing problem files I/we will need to keep the stream length correct. | 09:13.25 |
chrisl | kens: Okay, well, no worse than PDF editing, then..... | 09:18.13 |
kens | Yeah, I was just worried that if I started editing multipl e places it would turn into the usual pdfwrite can of worms | 09:18.39 |
| So this will be tedious, but possible. | 09:18.59 |
Robin_Watts | Morning paul. | 09:44.19 |
paulgardiner_lap | Robin_Watts: hi | 09:47.14 |
Robin_Watts | pedro reviewed all but 2 of my changes, so they are in. | 09:47.34 |
paulgardiner_lap | I see you've replied to one of the SO support requests. I was just generating a bug for the other. | 09:48.04 |
Robin_Watts | Perfect. | 09:48.14 |
| I have a bluetooth keyboard thing should you ever feel the urge to try it with the ipad. | 09:48.44 |
paulgardiner_lap | Robin_Watts: oh good about the changes. That was my next job to check if everything had gotten reviewed | 09:48.53 |
| Robin_Watts: ah right. That would be handy | 09:49.13 |
norbertj | Robin_Watts: hello, can you setup the access for casper (regression cluster) for me, with the ssh-keyfile from peeves? Should I send the peeves ssh file to you then? | 10:06.55 |
Robin_Watts | norbertj: Hi. | 10:07.10 |
| Yes, could you mail me that file please? | 10:07.21 |
| norbertj: Step 1 is to set you up with ssh access on casper. | 10:07.56 |
| The next steps depend on whether you are using windows or unix. | 10:08.12 |
norbertj | Robin_Watts: im on windows platform. At work I still have some issues with the firewall. But at home I should be able to connect. | 10:10.16 |
Robin_Watts | norbertj: And are you at home now ? | 10:10.27 |
norbertj | Robin_Watts: (file is sent). Currently I'm at work. | 10:10.43 |
| Robin_Watts: so I can't test just now. | 10:11.05 |
Robin_Watts | OK, the normal (unix) way of working is to use rsync to transfer the files. | 10:11.18 |
| sadly, this doesn't work well on windows, so I have a way of working using the git stuff. | 10:11.36 |
| This will require a small amount of setup on your machine. | 10:11.47 |
| Hence the best way to do this is to talk you through it when you are connected from home. | 10:12.08 |
| but I will set up as much as I can now. | 10:12.17 |
| That PPK is in putty format. I need it in openssh format please. | 10:13.22 |
norbertj | Robin_Watts: It's now 12:12 (local time), so when I'm at home I'll contact you (will be around 19:00 I think). | 10:13.41 |
Robin_Watts | norbertj: Fab. | 10:13.50 |
jogux_mac | anyone know if it's deliberate that the old picsel smart-office app (the £2.99 one) is still under the picsel account (ie wasn't transferred to the artifex account?) | 10:13.53 |
Robin_Watts | jogux_mac: That's a question for miles. He may just have missed it. | 10:14.20 |
norbertj | Robin_Watts: I just checked, and I have a putty connection with peeves ;) | 10:16.44 |
Robin_Watts | excellent! | 10:17.03 |
norbertj | Filezilla also works now. | 10:17.41 |
Robin_Watts | Did you resend the key in putty format? | 10:20.13 |
norbertj | Robin_Watts: just sent the keyfiles from peeves to you | 10:20.27 |
jogux_mac | robin_watts : okay, I'll email him then | 10:20.36 |
Robin_Watts | norbertj: Just arrived, thanks. | 10:20.44 |
jogux_mac | robin_watts : what's the protocol on that - should I cc you + paul + henry ? | 10:21.24 |
Robin_Watts | jogux_mac: Yes, that would be good. | 10:22.37 |
| We also have tech at artifex.com, which goes to all the engineers, but I'm not sure if you can write to that. We should get marcosw to add you. | 10:23.12 |
| norbertj: Try ssh to casper now ? | 10:23.26 |
norbertj | works | 10:25.20 |
Robin_Watts | Fab. | 10:25.32 |
| OK, so now to try to set up the cluster. | 10:25.37 |
| What is your local username on the windows box? | 10:25.44 |
norbertj | Robin_Watts: njj | 10:26.01 |
Robin_Watts | OK. I may need to tweak stuff because of that. | 10:26.16 |
| Can you paste the results of git remote -v please? | 10:27.17 |
norbertj | Robin_Watts: alas, I think I was connecting to peeves. But casper still does not work (our firewall I think). | 10:28.34 |
Robin_Watts | Ah. | 10:29.02 |
| How do you commit to ghostscript then? | 10:29.12 |
norbertj | can I connect via peeves to casper? | 10:29.13 |
| Robin_Watts: I never committed yet. | 10:29.27 |
Robin_Watts | norbertj: Yes, you should be able to. | 10:29.29 |
| Ah, I see. | 10:29.35 |
| You would need to have your appropriate private key on peeves though. | 10:30.00 |
jogux_mac | or generate a new key on peeves | 10:32.03 |
| (and get Robin to add that to casper too :-) ) | 10:32.11 |
| that's probably easiest imho; just "ssh-keygen", press enter a few times, then cat ~/.ssh/id_rsa.pub and email the output | 10:33.13 |
Robin_Watts | yeah. | 10:34.49 |
norbertj | Robin_Watts: done it. Thanks jogux_mac for the suggestion. | 10:37.20 |
Robin_Watts | OK, you should now be able to ssh to casper from peeves? | 10:38.34 |
norbertj | Robin_Watts: yes, I'm on casper. | 10:40.42 |
Robin_Watts | great. | 10:40.48 |
| In order to use the git cluster pushing stuff, we probably need a direct connection though. | 10:41.23 |
| Or else we'd have to do hairy port forwarding. | 10:41.34 |
chrisl | Keep repo on peeves, commit to that first, and clusterpush from peeves? | 10:42.17 |
jogux_mac | if you use command line ssh on the windows side, you can do fruity (easy) stuff with .ssh/config. Not sure if that can be done with putty though. | 10:42.21 |
Robin_Watts | (on phone, sorry) | 10:42.40 |
norbertj | Robin_Watts: I do have GIT working on my PC (through http). So I just have to push my fix? | 10:43.28 |
Robin_Watts | norbertj: http will allow you to pull, but not to push. | 10:44.04 |
| So... on peeves... | 10:44.20 |
| mkdir repos. | 10:44.25 |
| cd repos | 10:44.27 |
norbertj | Robin_Watts: then git checkout ? | 10:45.28 |
Robin_Watts | norbertj: Yeah, I'm just formulating the commands. | 10:45.44 |
| We're going to have the whole bare/non-bare repo thing. | 10:46.06 |
| git clone norbert@git.ghostscript.com:/home/git/ghostPDL.git | 10:46.20 |
| That will clone you a non-bare repo. i.e. it will get you a repo with a working set. | 10:46.50 |
norbertj | Robin_Watts: /home/git/ghostPDL.git does not appear to be a git repository | 10:48.26 |
Robin_Watts | ghostpdl.git, sorry. | 10:48.57 |
norbertj | Yeah, found it too in my .git/config | 10:49.22 |
Robin_Watts | so, has that completed? | 10:55.52 |
norbertj | Robin_Watts: yes | 10:57.06 |
Robin_Watts | ok, so change into ghostpdl.git | 10:57.21 |
| And do... | 10:57.25 |
| gs/toolbin/localcluster/clusterpush.pl gs lowres | 10:57.43 |
| If we're lucky that should kick off a cluster run. | 10:59.27 |
| If not, I should hopefully be able to see what I forgot to do from the output it gives :) | 10:59.56 |
norbertj | Robin_Watts: so cd ~/repos/ghostpdl and then gs/toolbin/localcluster/clusterpush.pl gs lowres | 11:00.00 |
Robin_Watts | Yes. | 11:00.05 |
pedro_ | hi folks | 11:00.07 |
Robin_Watts | Hi pedro_ | 11:00.13 |
norbertj | Then I miss: the cluster_key | 11:00.25 |
Robin_Watts | Right, yes. | 11:00.35 |
| What's the exact error message? | 11:00.47 |
norbertj | Warning: Identity file /home/norbert/.ssh/cluster_key not accessible: No such file or directory. | 11:01.12 |
| Permission denied (publickey). | 11:01.18 |
Robin_Watts | OK. just a tick. | 11:01.21 |
| norbertj: cp ~robin/cluster_key ~/.ssh/cluster_key && chmod 600 ~/.ssh/cluster_key | 11:02.29 |
norbertj | Failed to make /home/regression/cluster/users/norbert/ghostpdl | 11:03.51 |
Robin_Watts | ok, just a mo. | 11:04.01 |
| It's copying stuff... slowly. | 11:06.50 |
| Try again now ? | 11:07.09 |
| Woo Hoo! | 11:07.47 |
norbertj | did something ;) | 11:07.50 |
Robin_Watts | Job is running. | 11:07.50 |
norbertj | With my jpeg patch. | 11:07.57 |
Robin_Watts | norbertj: I need to make you a a dashboard login... | 11:08.14 |
| norbertj: So if you visit: http://ghostscript.com/regression/ it should ask you for a username/password. | 11:10.55 |
| Your username is norbert, your password is in your email. | 11:11.04 |
norbertj | I'm in. | 11:13.14 |
Robin_Watts | OK, so the dashboard is fairly obvious. Top left is the list of machines in the cluster. | 11:13.59 |
| Under that is the job queue. | 11:14.07 |
| In the job queue: Red jobs have completed. green job is running, amber jobs are waiting to be run. | 11:14.40 |
| Whenever we do a commit to either gs or mupdf, a job is queued. These are the ones that appear in the red section of the job queue. | 11:15.18 |
| To the right of the job queue are the user links. | 11:15.25 |
| When you schedule a user job, it will appear in the job queue, while it runs, then it will disappear when it finishes. | 11:15.58 |
| You can look at the result of that run by clicking your name. | 11:16.11 |
| So, the command you used to kick off the cluster run.... | 11:16.48 |
mace | Robin_Watts: all the vm's are booted and on the virtual network | 11:17.32 |
Robin_Watts | In general, if you are in a ghostpdl directory and you do: gs/toolbin/localcluster/clusterpush.pl that will kick off a run. | 11:17.50 |
| With no options, it will run a full set of tests. | 11:18.11 |
norbertj | How long does a run take? | 11:18.35 |
Robin_Watts | You can restrict the tests it uses using: gs (just for pdf/ps), pcl (just for pcl/pxl etc), xps (just for xps) | 11:18.54 |
| In addition, you can specify lowres to force it to omit the 300 dpi tests. | 11:19.40 |
| Oh and pdfwrite for pdfwrite tests :) | 11:19.54 |
| A full test takes 20-40 minutes depending on the mood of the cluster. | 11:20.12 |
| Your test there took 11mins. | 11:20.21 |
| So if you click on "norbert" in the user links, you should get the report. | 11:20.42 |
| So your commit made changes in 530 simple rendering tests | 11:21.41 |
| Plus 135 pdfwrite tests. Plus 93 ps2write tests. | 11:22.06 |
| Were you expecting any diffs? | 11:22.25 |
norbertj | Not that many. | 11:22.39 |
Robin_Watts | norbertj: OK. So now we get to see another feature of the cluster. | 11:22.53 |
| Now run: gs/toolbin/localcluster/clusterpush.pl bmpcmp | 11:23.04 |
kens | Robin_Watts | 11:23.08 |
| I just made some commits recently,possibly norbert doesn't have those | 11:23.20 |
Robin_Watts | kens: possibly, but this is working out as a nice example :) | 11:23.37 |
kens | Yep, I noticed that, just mentioning it so you know where to look :-) | 11:24.03 |
norbertj | kens: I just cloned, But perhaps the commit from 85mins ago? | 11:24.47 |
kens | norbert, yes possibly so | 11:24.57 |
| depends what exactly your clone copied | 11:25.14 |
norbertj | bmpcmp done. | 11:25.19 |
kens | But Robin_Watts is correct, this will eb a useful exercise | 11:25.24 |
Robin_Watts | norbertj: So, you'll see that the cluster dashboard is now showing a bmpcmp job running. | 11:25.45 |
kens | Hmm, I can't get a bmpcmp result from norbert :-( | 11:25.48 |
| Oh I see, not yet complete | 11:26.01 |
norbertj | yes, the submit was done. | 11:26.10 |
kens | No problem, my misunderstanding | 11:26.28 |
Robin_Watts | The cluster is taking the jobs that rendered differently in your last user test, and rendering them to bitmaps with the current trunk version, and your version. | 11:27.01 |
| (it will only do the first 1000 files) | 11:27.23 |
| Then it compares the bitmaps, and produces a series of webpages with the diffs in. | 11:27.39 |
norbertj | Robin_Watts: so I would 'git pull --rebase' to get in sync again? | 11:30.26 |
Robin_Watts | kens: seems unlikely that it will be your commits, as your commits didn't make 530 diffs :) | 11:30.28 |
kens | Robin_Watts : true, just thought that 'some' of tehm might be mine | 11:30.46 |
Robin_Watts | norbertj: Assuming you have no uncommitted changes, yes. | 11:30.59 |
| norbertj: Did the email about the last job arrive? | 11:31.34 |
kens | So this is changes in the DCT decoder, possibly that could cause diffs | 11:33.02 |
Robin_Watts | kens: I vaguely thought the changes were pcl only. | 11:33.20 |
kens | There's some changes in sdctd.c | 11:33.33 |
| THe others are PCL only | 11:33.55 |
| None fo the files look like they would be impacted by my recent twiddling | 11:34.32 |
Robin_Watts | norbertj: So the run has now finished. Click on the 'bmpcmp' link to the right of 'norbert'. | 11:37.54 |
kens | Images gone missing it seems | 11:39.07 |
Robin_Watts | For each test (file/output device/resolution/banding) you'll get 1 or more triples of bitmaps. | 11:39.12 |
| New, Old, Diff (or Candidate, Reference, Diff) | 11:39.37 |
| As ken says, there are clearly images that aren't showing up in the rendering any more. | 11:40.07 |
norbertj | Ok, getting it. So my fix (though working for my testfile on pxl, has impact on postscript/pdf too). | 11:40.46 |
Robin_Watts | If you wave your mouse pointer in the middle image, you get a nice zoomed in magnifier thing | 11:40.58 |
| In this case, it's very easy to see the differences between the images. | 11:41.33 |
| and the green pixels in the "diff" image tells you where to look. | 11:41.51 |
norbertj | Nice bitmap compare tool. | 11:41.57 |
Robin_Watts | For some cases, you'll find that even with the diff the differences aren't easy to see, so there is 1 more trick. | 11:42.16 |
| If you hold the ctrl key down on your keyboard, and roll the mouse in/out of the left hand image, the left/middle images will swap places. | 11:42.59 |
| Thus you can do a "blink test" and it becomes MUCH easier to spot pixels turning on/off. | 11:43.26 |
| You can download individual test files that shows the problems by clicking on the link in the title of each test. | 11:46.40 |
| (either in the bmpcmp window, or within the report itself when viewed on the web). | 11:46.40 |
| And the numbers in the header/footer of the bmpcmp page take you to the appropriate page of results. | 11:46.40 |
| There are other options to bmpcmp to filter the results and do fuzzy compares etc, but we can revisit that if you ever need them :) | 11:46.41 |
norbertj | That's ok with me, I should be swamped with too much information ;) | 11:46.41 |
Robin_Watts | If you ever run tests and find that you're seeing some changes you don't understand, ask us about it. We have some files that change apparently at random. and we can tell you if this is one of those. | 11:47.15 |
kens | heads to lunch | 11:47.42 |
norbertj | I undid my changes, so if I do a clusterpush again, I should have 0 diffs. | 11:48.03 |
Robin_Watts | norbertj: Indeed. | 11:48.08 |
| When you get it set up so that you can ssh direct from your machine, let me know and I'll talk you through the git based version. | 11:48.41 |
norbertj | Ok, I'm baselining now. | 11:48.56 |
Robin_Watts | Once set up it's basically the same stuff, just a different script to kick it off. | 11:49.04 |
norbertj | Robin_Watts: will do that. For now I'm ok. | 11:49.14 |
Robin_Watts | sure. | 11:49.25 |
norbertj | but have to rehink about my fix :( | 11:49.26 |
Robin_Watts | norbertj: Ah... every time you run a test, it shows you 2 things. | 11:50.00 |
| 1) the diff from the trunk. | 11:50.06 |
| 2) the diff from your previous run. | 11:50.13 |
| So when you get your next report though you should see no diffs from the trunk, and then 530 diffs from the previous run. | 11:50.37 |
| If you are not expecting this, it can be very confusing as a quick glance makes it look like all the diffs are still there. It gets me every time. :) | 11:51.08 |
norbertj | I'll try to remember it ;) | 11:51.55 |
Robin_Watts | Jogux, pedro, paulgardiner: So... other than getting these last 2 commits in, I'm up to date with everything, I think. | 11:52.58 |
jogux_mac | cool | 11:53.05 |
Robin_Watts | I have to figure out what to look at next in SmartOffice. | 11:53.10 |
| Adding jbig2dec support to the image component is one possibility (so we'll finally be able to read baseline PDFs properly :) ) | 11:54.04 |
| or I could jump in to help with the word file appearance stuff. | 11:54.43 |
| pedro_: I don't want to duplicate effort, or get in other peoples way though. You've been looking at this. Is there a separable bit of work you can see that I could usefully do? | 11:55.40 |
| Also, both paulgardiner and I are off Thurs/Fri this week and Mon/Tuesday next week. | 11:57.17 |
pedro_ | robin> I'm still working mostly on the missing content/unsupported stuff, and haven't looked at much of the font/word-wrap/linespacing stuff in detail, which seems like it may be fairly closely related. | 12:01.11 |
Robin_Watts | closely related to the missing content stuff? | 12:01.45 |
paulgardiner | jogux_mac: Do you know how to add files onto the iOS simulator for use by an app? | 12:01.57 |
pedro_ | is also away from Friday this week until Friday next week (although planning to do a day or two while I'm away) | 12:02.15 |
jogux_mac | paulgardiner: just copy them in :-) | 12:02.22 |
| paulgardiner: as a general rule of thumb I tend to do more debug on an actual device btw | 12:03.11 |
| but let me just find the location you need to copy to | 12:03.17 |
paulgardiner | jogux_mac: I almost never use the simulator, but I have no airprint printer | 12:03.37 |
pedro_ | Robin_watts: sorry, I meant that from what I've looked at so far, a lot of the layout stuff seems attributable to font metric/linespacing differences | 12:03.38 |
Robin_Watts | pedro_: Ah, right, yes. | 12:03.49 |
jogux_mac | paulgardiner: ahhh. d'oh. | 12:04.15 |
pedro_ | much of the stuff I'm looking at just isn't supported at present :( | 12:04.17 |
Robin_Watts | I'm hoping that we should be a lot better in terms of font metrics now (so horizontal layout should be more or less OK) | 12:04.31 |
pedro_ | nods - cool | 12:04.45 |
Robin_Watts | linespacing is a different matter. Fixing the vertical spacings in the fonts does not seem to have solved the linespacing issues. | 12:05.06 |
paulgardiner | jogux_mac: just copy them in to where? I read things have changed since v 6.1 | 12:05.12 |
Robin_Watts | So, if I was to look at the linespacing stuff, that wouldn't be conflicting with you? | 12:05.39 |
pedro_ | no - that would be great. I'm mostly in the parsing/styling support in te msword DA & ooxml code plus the VML parsing (shapes and images) at the moment | 12:07.31 |
jogux_mac | paulgardiner: yeah, just checking the path. I don't seem to have a build handy :-S | 12:07.32 |
Robin_Watts | pedro_: Fab. | 12:07.45 |
jogux_mac | paulgardiner: something like ~/Library/Application Support/iPhone Simulator/7.1-64/Applications/<guid-directory-that-has-so.app-in-it>/Documents | 12:10.15 |
| actually the easiest way to get that path; start the app in simulator then ps -axf | grep -i smartoffice or whatever | 12:10.44 |
paulgardiner_lap | jogux_mac: magic thanks | 12:10.45 |
jogux_mac | or just cheat and use the getting started or template docs that are bundled with the app :-) | 12:11.38 |
jogux_mac | is very tempted to suggest we should add much of the customer config information into git (ie. so the build commands in the release spreadsheet become more like scripts/tgvbuild.py --release=ios-appstore) | 12:13.20 |
Robin_Watts | ohgodyes. | 12:14.03 |
jogux_mac | paulgardiner: I can probably try and test airprint | 12:14.22 |
| I suspect to make tracking changes to a customer's config easier we should have a .txt per release type with one line per build option or something | 12:14.51 |
paulgardiner_lap | jogux_mac: great. I'll invite you to out testflight thingy in a minute | 12:15.42 |
jogux_mac | paulgardiner_lap: http://www.netputing.com/handyprint/ may be handy. | 12:16.08 |
| paulgardiner_lap : oh, we already use testflight? cool :-) | 12:16.22 |
Robin_Watts | jogux_mac: Or add comments to the tgvbuild.py script where we handle the --release thing? Although git history will probably help us there. | 12:16.33 |
jogux_mac | I think, given it's current size, ideas that add the least possible to tgvbuild.py should be encouraged :-) | 12:17.28 |
| actually I wonder if it should be a simple outer wrapper, as not all releases are tgvbuild based. (some releases seem to be done randomly with sdkbuild.py at least) | 12:18.56 |
paulgardiner_lap | He he. I was just going to say outer wrapper | 12:20.04 |
emarsden | Hi, the pdfwrite output mechanism seems to break PDF file attachment annotations and document-level file attachments. Is this a known problem? | 12:26.49 |
Robin_Watts | So, can anyone suggest a suitable build line for the windows version of smart office? To save me a few iterations of guesswork? | 12:27.10 |
kens | pdfwrite does not support certain kinds of annotations | 12:27.11 |
| see gdevpdfm.c for supported annotations | 12:27.29 |
pedro_ | tgvbuild.py -vs2010 --dispman-ue2fileviewer app/simple | 12:27.47 |
jogux_mac | would throw a -debug into pedro's suggestion. | 12:28.02 |
pedro_ | well, yeah, that's implied ;) | 12:28.14 |
jogux_mac | :-) | 12:28.32 |
Robin_Watts | will play with -vs2005 and -vs2008 switches too. | 12:29.03 |
pedro_ | nods | 12:29.14 |
jogux_mac | -vs2005 may be the point you discover you're missing the directx sdk or winhttp.h, itrying to nstall could be the final straw that pushes you over the edge :-) | 12:29.56 |
Robin_Watts | jogux_mac: Indeed. | 12:30.16 |
pedro_ | quick question: for bugs in ghostdocs where we appear to be adhering to the ooxml specs (and match OpenOffice rendering) are we still aiming to match MSWord? | 12:32.20 |
Robin_Watts | pedro_: Leave a bugzilla bug open, but lower the priority. | 12:33.10 |
| Our initial priority is to match cases where LibreOffice/OpenOffice/MSWord etc all agree. | 12:33.27 |
pedro_ | ok, ta. | 12:33.33 |
emarsden | kens: OK, thanks | 12:39.37 |
Robin_Watts | TTOTD: Before going hunting through the build system looking for where a define is being enabled, check to see if you are setting it on the command line. | 12:47.13 |
jogux_mac | :-) | 12:47.38 |
Robin_Watts | pedro: Should I expect to have to do -windows7, or is that detected automatically? | 13:04.58 |
pedro_ | believes that's a maual switch | 13:05.30 |
Robin_Watts | Ass, so I need to restart my build :( | 13:06.21 |
Robin_Watts | lunches | 13:06.33 |
jogux_mac | possibly daft question, what does -windows7 do? | 13:06.53 |
pedro_ | robin> you'll still get a binary which runs on win7 | 13:06.53 |
| at present, doesn't package anything (just sets up a bunc of build settigns to sensible values (as per windows8) but doesn't currently package anything | 13:08.20 |
pedro_ | doesn't think Robin_Watts needs to use -windows7 | 13:09.34 |
| (basically because we have the dispman-ue2fileviewer dev build working without that, and aren't planning to do a win7-based installer at present | 13:12.00 |
Robin_Watts | Missing winhttp.h. | 13:46.06 |
| Jogu: Was that one of your predictions? | 13:46.14 |
| yeah. Let me move to 2008. | 13:46.34 |
jogux_mac | it was | 13:48.06 |
Robin_Watts | actually, sod it. I'll use 2010. | 13:48.58 |
jogux_mac | I'm tempted to suggest we nominally standardize on a vs version everyone has access to (which would almost certainly be one of the free express editions) | 13:49.17 |
| 2013 is a requirement for winrt stuff I think too. | 13:49.39 |
Robin_Watts | jogux_mac: Yeah, it would make sense. | 13:49.39 |
| 2013 requires windows 8 to run on. | 13:49.52 |
| AIUI. | 13:49.54 |
pedro_ | jogux> the 'express' editions will probably involve more path hacking vs. the 'full' versions of the same vintage | 13:50.30 |
jogux_mac | robin_watts : afaik 2013 runs on win7.... however you can only submit windows store apps from 8.1. | 13:52.11 |
| pedro : seems likely, albeit I would wager that the -vs2012 flag is already setup for the free version... | 13:53.21 |
| I'm fairly certain I tried -vs2012 and it worked with the free one, anyway | 13:54.48 |
safa_ | congrats on mupdf, a great tool | 13:58.06 |
kens | thanks safa_ on behalf of the actual developers | 13:59.08 |
Robin_Watts | safa_: Thanks. Do you have a question? I have to pop out for 15 mins in a mo. | 13:59.41 |
safa_ | running on old debian laptop, everything works perfectly | 14:00.16 |
| only question, after looking through source, there's no way opening a pdf could reveal ip address? | 14:00.33 |
| i don't see any socket calls, etc | 14:00.37 |
Robin_Watts | nope. no networking at all. | 14:00.50 |
safa_ | sweet | 14:01.07 |
| you guys rock | 14:01.10 |
| if only other pdf readers cared as much about security :) | 14:03.50 |
| their users' security, that is | 14:04.00 |
Robin_Watts | pedro_: Doing a build under vs2010, but without -windows7, it died with: EXEC : error : Failed to execute command python D:\cvs\artifex\sot.git\epage/platform/tgv/apps/simple/postlink.py | 14:20.52 |
| Should I expect that? | 14:20.56 |
pedro_ | that's an interesting postlink.py location | 14:21.56 |
Robin_Watts | It is? It exists. | 14:22.28 |
jogux_mac | is there not another more detailed error hidden somewhere? | 14:22.48 |
pedro_ | nods - just an odd one to fail on | 14:23.04 |
Robin_Watts | http://pastebin.com/xpRXDPqQ | 14:23.51 |
pedro_ | thinks I'd have had app/simple rather than appsimple | 14:24.37 |
| not sure whbut probably unrelated to the error | 14:25.01 |
| : rephrases | 14:25.10 |
Robin_Watts | retries with app/simple | 14:25.37 |
| oh, that changed the path, so it's rebuilding everything. Joy. | 14:26.12 |
pedro_ | :( | 14:26.24 |
jogux_mac | robin_watts : I wouldn't necessarily rule out their been an error earlier in that log; the output streams sometimes seem rather "interleaved" | 14:26.41 |
| I can't otherwise understand why your system would now be refusing to start a python script it's succeeded in running before | 14:27.06 |
Robin_Watts | previously, I've seen such errors and they meant "we ran that script, but didn't show you any output from it, and it returned a failing error code" | 14:29.57 |
| ooh, yes, earlier errors. | 14:31.12 |
| 1>d:\cvs\artifex\sot.git\epage\platform\tgv\apps\windows-alien-common\tcp\alien-tcp-types.h(32): fatal error C1083: Cannot open include file: 'openssl/ssl.h': No such file or directory [D:\cvs\artifex\sot.git\epage\platform\tgv\apps\windows-alien-vs2010\win32ui.vcxproj] | 14:31.51 |
| 1> alien-tcp-api.c | 14:31.53 |
| where should it be finding openssl ? | 14:32.23 |
pedro_ | mm, the project files are hardcoded to assume you have OpenSSL in C:\OpenSSL | 14:32.38 |
Robin_Watts | Ah. | 14:32.49 |
| jogux_mac: Do you have VS2010? Pedro obviously does, I do, and I think Paul does. | 14:33.05 |
pedro_ | thinks that *should* have been the same with vs2005 though... | 14:33.13 |
Robin_Watts | pedro_: I've been doing -nonet builds until now. | 14:33.38 |
pedro_ | ah | 14:33.50 |
Robin_Watts | and avoiding tgv. | 14:33.51 |
paulgardiner | I have the Express version | 14:34.05 |
pedro_ | envies your approach | 14:34.12 |
Robin_Watts | Working with test-shell is simpler/faster. | 14:34.15 |
jogux_mac | robin_watts: I don't think I have it installed, but I appear to have a vs2010 pro key from somewhere. | 14:34.20 |
| I'm tempted to believe we should change openssl to be installed in the default location. | 14:34.48 |
Robin_Watts | Well, VS2005/2008 basically use the same project file format. | 14:34.55 |
jogux_mac | robin_watts : it's the /32 bit/ openssl you want btw. | 14:35.17 |
paulgardiner | New MuPDF Ad Hoc release is a 13MB ipa file, previous (Jan) was 8MB. Not sure what has caused the difference. Can't be mujs, surely? | 14:35.22 |
| tor8: ^ | 14:35.35 |
jogux_mac | paulgardiner> have you changed xcode version since? | 14:36.10 |
Robin_Watts | jogux_mac: I bet there is an archive on points somewhere. | 14:36.15 |
jogux_mac | robin_watts : yeah, I'd hope I can find the iso somewhere | 14:36.40 |
Robin_Watts | jogux_mac: I meant for openssl :) | 14:36.52 |
tor8 | paulgardiner: no, can't be mujs by any reasonable explanation | 14:37.07 |
jogux_mac | robin_watts : oh, right. well I'd just use the latest. | 14:37.17 |
paulgardiner | jogux_mac: possibly. | 14:37.21 |
jogux_mac | since all the earlier versions have security holes | 14:37.31 |
pedro_ | the latest works fine | 14:38.10 |
jogux_mac | paulgardiner : if you've moved xcode 4 -> 5, it may be an armv7/armv7s binary now rather than a armv6/armv7. though not sure that explains it either really. | 14:38.13 |
pedro_ | http://slproweb.com/download/Win32OpenSSL-1_0_1f.exe | 14:38.14 |
paulgardiner | But I thought I had to swap to XCode 5 to target my iPad so I would have been forced to do so in Jan | 14:38.35 |
jogux_mac | yeah, just install to c:\openssl (as opposed to c:\openssl-32bit or whatever it defaults to) | 14:38.38 |
paulgardiner | XCode has probably updated in that time | 14:39.03 |
Robin_Watts | jogux_mac: And that path is hardcoded in a .vcproj? or a .bld ? | 14:39.21 |
jogux_mac | probably 4 billion vcprojs | 14:39.39 |
pedro_ | the windows project files - incudes and libs | 14:39.58 |
Robin_Watts | Urgh. Would be nicer as a sysvar, but I bet that's not possible. | 14:39.59 |
jogux_mac | probably some .bld or similar too, but they're less relevant. | 14:40.02 |
Robin_Watts | We should use PENSSL_DIR% or something, and then set that smartly in the .py code. | 14:40.29 |
| bah. stupid irc. | 14:40.41 |
jogux_mac | wonders if openssl installer sets up any env variables. | 14:40.47 |
pedro_ | reckons it could probably be added to the relevant toolchain files | 14:40.58 |
Robin_Watts | cos C is my SSD and I'd rather not waste space on that. | 14:41.32 |
jogux_mac | paulgardiner : might be worth unzipping both and diffing. | 14:41.38 |
paulgardiner_lap | jogux_mac: yeah good thought | 14:42.18 |
pedro_ | jogu: OPENSSL_CONF could be parsed to get the base (points at OpenSSL-Win32\bin\openssl.cfg) | 14:42.27 |
jogux_mac | pedro : sounds like a plan. | 14:42.42 |
Robin_Watts | pedro, jogu: I'll look into that then? | 14:42.59 |
jogux_mac | has no objection. | 14:43.10 |
pedro_ | robin> cool | 14:43.12 |
Robin_Watts | I'll invent an OPENSSL_DIR env var, and set that (if it's not already set) by deriving it from the OPENSSL_CONF one. | 14:43.42 |
jogux_mac | cool. | 14:43.53 |
| for bonus marks, bail out saying "install openssl from htto://blalbhah/" if that var is unset :-) | 14:44.11 |
Robin_Watts | something is not right with my PC. Going to reboot. | 14:56.38 |
paulgardiner_mac | Jogux: re mupdf, the actual binaries are wildly different in size 16MB vs 26MB. I guess I cannot easily analyse any further | 15:04.00 |
jogux_mac | ah right. there's probably more you can do, but it may not be worth the time :-) | 15:04.38 |
paulgardiner_mac | I guess anything silly I might have done would show up in git history. | 15:07.11 |
Robin_Watts | jogux_mac: no openssl env vars here :( | 15:07.24 |
pedro_ | Robin_Watts: that's the same installer I used and it created them for me :-S | 15:08.27 |
jogux_mac | checks my win8.1 install; that definitely has openssl_conf set. | 15:08.44 |
Robin_Watts | pedro_: Did you change any options on the install other than install dir ? | 15:09.52 |
jogux_mac | definitely didn't. | 15:10.03 |
pedro_ | Robin_Watts: no | 15:12.33 |
Robin_Watts | You let it copy DLLs to the windows dir? Me too. | 15:12.39 |
pedro_ | nods | 15:12.48 |
Robin_Watts | Just re ran the installer. Still no vars. | 15:13.24 |
| I have a space in my path. Maybe that's it? | 15:13.34 |
pedro_ | re-installs | 15:14.05 |
Robin_Watts | no, even with no space, no sysvar. | 15:14.30 |
pedro__ | boggles - I've just removed the env var and the OpenSSL dir, reinstalled and now its not creating the env var when I reinstall | 15:21.10 |
pedro | wonders if it only creates that var the first time you run the DLLs... | 15:23.48 |
| nope | 15:29.11 |
Robin_Watts | don't worry, I'll cope. | 15:29.11 |
pedro_ | Robin_Watts: just rebooted, et voila! | 15:30.38 |
mvrhel_laptop | kens: are you there? | 15:30.46 |
Robin_Watts | so it only shows up on a reboot? | 15:30.49 |
kens | mvrhel_laptop : yes, is it quick ? | 15:30.55 |
pedro_ | apparently so | 15:30.56 |
mvrhel_laptop | kens: bug 695118 | 15:31.16 |
kens | yes ? | 15:31.26 |
mvrhel_laptop | kens never mind I see you comment | 15:32.42 |
| s/you/your | 15:32.46 |
kens | :-D | 15:32.49 |
mvrhel_laptop | typing slow today | 15:32.50 |
| I will see if I can figure this out before the next staff meeting | 15:33.22 |
| it really should not be too hard | 15:33.27 |
kens | ok thx michael | 15:33.33 |
mvrhel_laptop | have a good night kens | 15:33.41 |
kens | not gone yet ;-) | 15:33.49 |
darkdrgn2k | Hi All | 15:37.30 |
| im using the GhostScript Win32 API and it works great... except when it reandomly doesnt yielding -100 durring gsapi_new_instance | 15:38.06 |
| its running in an IIS process and works fine if i switch pools so makes me thing it may be a memory leak of some sort? | 15:38.25 |
chrisl | darkdrgn2k: I assume you are also shutting down Ghostscript properly? | 15:41.10 |
darkdrgn2k | yep looks liek it | 15:41.53 |
| gchandleArgs.Free() api_exit(intGSInstanceHandle) gsapi_delete_instance(intGSInstanceHandle) | 15:42.03 |
| found that my If (intReturn < 0) Then didnt gsapi_delete_instance(intGSInstanceHandle) but i dont know if that could be an issue | 15:42.23 |
| and if i swich the IIS pool everythign starts working again | 15:42.45 |
chrisl | darkdrgn2k: I don't think that shuts down ghostscript, just deletes the dll instance...... | 15:45.08 |
| darkdrgn2k: I think before that, you need a call to gsapi_exit() | 15:46.38 |
darkdrgn2k | here is the whole function | 15:46.55 |
| http://pastebin.com/k1wM5Xur\ | 15:46.55 |
| http://pastebin.com/k1wM5Xur | 15:46.57 |
chrisl | As best I can tell, it looks okay, so you may have found a problem. | 15:48.40 |
darkdrgn2k | i hope so | 15:49.32 |
| memory ussage on server is growing after i swiched the pool | 15:50.26 |
| but that could just be normal oprations.. | 15:50.34 |
chrisl | You're best bet would be to produce a minimal C program to reproduce your issue, and find a single job that, run repeatedly, shows memory increasing | 15:50.45 |
| The report a bug | 15:50.52 |
| s/The/Then | 15:50.58 |
darkdrgn2k | yeh... well its not my coding thought tirhgt | 15:51.06 |
| it looks right | 15:51.08 |
| what would cause -100? | 15:51.16 |
chrisl | -100 is gs_error_Fatal - we couldn't even initialise enough to give a meaningful error | 15:52.18 |
darkdrgn2k | so it could be memory or somethign else rigt | 15:52.36 |
chrisl | Yes, but it's going to be very hard to tell without debugging it | 15:52.59 |
| darkdrgn2k: actually, what version of gs are you using? | 15:56.43 |
darkdrgn2k | just upgraded to 9.02 i think | 15:57.04 |
kens | O.O | 15:57.09 |
darkdrgn2k | sorry | 15:57.16 |
| gs910w64 | 15:57.16 |
| i upgraded from 906 | 15:57.35 |
chrisl | We just released 9.14.... but I'm not sure that'll help | 15:57.47 |
kens | what's the device ? | 15:57.59 |
darkdrgn2k | windows 2008R2 server | 15:58.45 |
| just using GS to thumbnail pdfs.. | 15:58.57 |
kens | what GS device ? | 15:59.05 |
darkdrgn2k | ? sorry i dont understand the question | 15:59.30 |
chrisl | What kind of output are you creating? | 15:59.47 |
darkdrgn2k | oh | 15:59.54 |
| jpg | 15:59.54 |
kens | at least its not pdfwrite :-) | 16:00.13 |
darkdrgn2k | LoL no PDF->JPG :) | 16:00.28 |
| astrArgs(9) = "-sOutputFile=c:\temp\" & tmp.ToString & "-%03d.jpg" | 16:00.29 |
| and only 1 page | 16:00.48 |
chrisl | So, you should have something like -sDEVICE=jpeg in your args | 16:00.50 |
darkdrgn2k | "-dNOPAUSE" "-dBATCH" "-dSAFER" "-dFirstPage=1" "-dLastPage=1" "-dNOPAUSE" "-sDEVICE=jpeg" "-r72" | 16:01.32 |
chrisl | Seems fair enough | 16:02.23 |
darkdrgn2k | gs doensst do office docs right? | 16:02.56 |
chrisl | No, it only does Postscript and PDF | 16:03.14 |
darkdrgn2k | meh good enough for me | 16:03.37 |
chrisl | Office formats have problems as general purpose "printing" formats | 16:04.08 |
darkdrgn2k | yeh and being propriatery its a pain | 16:04.35 |
chrisl | The recent XML based ones are "open" | 16:04.54 |
| darkdrgn2k: how easily could you temporarily handle either file types other than jpeg, or even no image files at all? | 16:06.10 |
darkdrgn2k | what do you mean? | 16:06.53 |
| the generation of imnages is not a critical component of the page in question. | 16:07.10 |
chrisl | Well, could you accept PNG, for exmaple? | 16:07.49 |
| Or better yet, ppm? | 16:08.08 |
kens | OK I'm off, if henrys pops up could someone direct him to bug #695123 and its two companions ans ask him for an opinion regarding rasterops please / | 16:12.36 |
| goodnight all | 16:12.48 |
chrisl | goodnight kens | 16:12.55 |
Robin_Watts | wasn't henrys away today? I could be misremembering. | 16:13.20 |
chrisl | I don't see a mail from him about it, but I could have deleted it..... | 16:14.28 |
darkdrgn2k | chrisl: jpgs are smaller that the only reason i was using them but pngs are fine too | 16:20.06 |
Robin_Watts | jpgs are a very poor choice for rendering pdf and ps to. | 16:21.31 |
chrisl | darkdrgn2k: all I wanted was for you try something other than the jpeg device to see if it was the device causing the problem, or if it device independent | 16:22.02 |
Robin_Watts | pdf and ps generally contain text. Text has lots of high frequency edges, and so will look awful with jpeg compression; lots of ringing. | 16:22.07 |
| -sDEVIVE=png16m will give nicer results. | 16:22.27 |
darkdrgn2k | ok ill try it out | 16:22.33 |
henrys_ | norbertj: you can just remove the OCE #define and commit if you want now that you have an account. We do like changes to have a review by any staff member before committing | 16:28.00 |
| norbertj: I reviewed that one. | 16:28.14 |
Robin_Watts | norbertj: You found the diffs then? | 16:28.53 |
| extra parens in the if (A || (B && C) ) ? | 16:29.26 |
ray_laptop | Robin_Watts: we generally don't like sending SSh private key files via email (unless they are encrypted, and the password is transmitted verbally) It's too easy for emails to go awry | 16:40.48 |
Robin_Watts | ray_laptop: Fair enough. | 16:41.35 |
ray_laptop | Robin_Watts: but if it was just his public keys for "authorized_keys", then that's fine | 16:41.50 |
Robin_Watts | It was. | 16:42.01 |
ray_laptop | I couldn't tell from the exchange for sure. Thanks | 16:42.16 |
henrys_ | oh I didn't see kens question because he was missing the _ in my irc name. | 16:42.48 |
ray_laptop | henrys: I just trigger on 'ray' so all of my various nicks are picked up | 16:43.37 |
| which will work until we get someone hanging out here that confuses that | 16:44.10 |
henrys | ray_laptop: oh I didn't know I could do that. | 16:44.12 |
ray_laptop | henrys: maybe not -- it's a client feature. Chatzilla allows for it | 16:44.35 |
chrisl | I think most gui clients let you specify highlight key words, or similar | 16:46.36 |
ray_laptop | oops. have to run something home. bbiab | 16:48.11 |
mvrhel_laptop | ray_laptop: you still there? | 16:48.28 |
| the email from customer 885 seems crazy. | 16:50.29 |
Robin_Watts | pedro: Did you footle with "browse-tags" ? | 16:53.54 |
pedro | Robin_Watts: not that I recall - let me check | 16:55.52 |
Robin_Watts | pedro: No, you didn't. sorry. | 16:57.01 |
pedro | np - I did tweak some of the toolchain files so wasn't 100% sure ;) | 16:59.19 |
norbertj | henrys: you mean the plexity-patch I assume | 17:03.29 |
henrys | norbertj: right | 17:04.21 |
jogux_mac | pedro : robin_watts : paulgardiner : latest commit in my repo contains a very simple go at earlier discussion about keeping release build commands in git; comments welcome. | 17:08.16 |
Robin_Watts | pedro: So, I've built (without -windows7) and it says it's completed... but where is my binary? | 17:08.21 |
jogux_mac | robin_watts : scripts\showbuild.py ? | 17:08.33 |
Robin_Watts | jogux_mac: Doesn't work trivially. | 17:08.48 |
jogux_mac | how does it fail? | 17:09.04 |
pedro | platform/tgv/apps/mindows-vs201-alien iirc | 17:09.11 |
| or windows | 17:09.17 |
Robin_Watts | D:\cvs\artifex\sot.git\epage\platform\tgv\apps\windows-alien-vs2010\Debug | 17:09.45 |
| jogux_mac: It wasn't liking the fact that I don't have GEN_ROOT set. | 17:10.00 |
pedro | tries to type 'windows-alien-vs2010 (or 2005 or 2012, depending on toolchain) | 17:10.15 |
Robin_Watts | but if i run it under solution_build.py it worked. | 17:10.16 |
jogux_mac | robin_watts : wtf. | 17:10.29 |
Robin_Watts | solution_build.py sets GEN_ROOT for me :) | 17:10.43 |
jogux_mac | oh. right. yes. | 17:11.10 |
Robin_Watts | Still building into the tree. <gerph>sigh</sign> | 17:11.26 |
pedro | Robin_Watts: mm, always has iirc | 17:11.43 |
Robin_Watts | pedro: Always has, yes. But the longevity of the shiteness does not excuse the shiteness itself. | 17:12.07 |
jogux_mac | robin_watts : we tried not building in the tree at some point, but the only sane solution resulted in two copies of all the files, and people always editted the wrong ones :-) | 17:12.08 |
pedro | apart from host tools and a few other resourcey bits and pre-processed files | 17:12.26 |
jogux_mac | although for the iphone build someone seems to have gone out of their way to build stuffi n the tree :-( | 17:12.49 |
Robin_Watts | I ran the binary... and it asserted. | 17:15.45 |
Robin_Watts | sobs. | 17:15.49 |
jogux_mac | aw | 17:16.02 |
Robin_Watts | Is this the point where I need a -windows7 flag? | 17:16.34 |
pedro | is on windows7 and doesn't need it | 17:16.58 |
| what was the assert? | 17:17.03 |
Robin_Watts | assert(NULL == model) | 17:17.14 |
| from within UE2FvModel_Busy_disable | 17:17.21 |
jogux_mac | try dragging it from debug up a directory? | 17:17.31 |
pedro | do you have the patch for the temp dir path fix? | 17:17.37 |
jogux_mac | I suspect it's failing to find the mcf files. | 17:17.39 |
| actually, no, not sure I suspect that. | 17:18.22 |
jogux_mac | reckons pedro was closer. | 17:18.30 |
| can you run it in the debugger and catch the console output? | 17:18.39 |
Robin_Watts | pedro: the what for the what? | 17:18.42 |
jogux_mac | robin_watts : pedro's patch that you fixed the url_destroy position | 17:18.55 |
pedro | Robin_Watts: that's the one | 17:19.05 |
Robin_Watts | jogux_mac: I am running it in the debugger, and it looks to me like it's failing to find the mcf files. | 17:19.13 |
jogux_mac | oh, I was right | 17:19.19 |
Robin_Watts | Yes, I have all pedros fixes (which are on golden now I believe). | 17:19.30 |
jogux_mac | try movning it up a level in the file system :-) | 17:19.31 |
| the older builds don't generate the binary in the 'Debug' subdirectory, for whatever reason. so I suspect that of being an issue. | 17:20.08 |
pedro | dispman-ue2fileviewer builds the mcfs into an 'mcf' directory in the same location as the binary, so should work in or outside the debugger | 17:20.16 |
| robin> which toolchain are you using? | 17:20.42 |
Robin_Watts | There is no "busy-background.png" | 17:20.45 |
| vs2010 | 17:20.47 |
| The paths look plausible. | 17:20.57 |
| 0.463|[09]:Th 0x05071BD4:( dispManAppThread):Ignored error 'D:\cvs\artifex\sot.git\epage\foundation\platform\tgv\libraries\io\fs\file-veneer.c[1310]:Open of /d:/cvs/artifex/sot.git/epage/D:/cvs/artifex/sot.git/epage/platform/tgv/apps/windows-alien-vs2010/DEBUG/mcfs/carbon-blue-480.mcf/busy-background.png failed (:)' [30F], at D:\cvs\artifex\sot.git\epage\libraries\image\create-url.c:323 | 17:21.18 |
| oh, so scratch what I said about the busy-background.png | 17:21.48 |
pedro | genrootness | 17:22.46 |
| ? | 17:23.01 |
Robin_Watts | 0.458|[00]:Th 0x05071BD4:( dispManAppThread):Ignored error 'D:\cvs\artifex\sot.git\epage\foundation\platform\tgv\libraries\io\fs\file-veneer.c[1310]:Open of /d:/cvs/artifex/sot.git/epage/epage_dir/Resource/image/placeholder.png failed (:)' [30F], at D:\cvs\artifex\sot.git\epage\libraries\image\decoder.c:1200 | 17:23.07 |
| I think EPAGE_DIRECTORY is not being picked up. | 17:23.16 |
jogux_mac | the first one sounds fatal; the latter one less so. | 17:23.39 |
| actually, no, I think that busy-background.png is non-fatal too | 17:24.06 |
Robin_Watts | let me pastebin the whole lot. | 17:24.18 |
pedro | the first one appears to have your EPAGE_DIRECTORY twice in the path | 17:24.29 |
jogux_mac | true, true. | 17:24.46 |
Robin_Watts | http://pastebin.com/nAX1Du1H | 17:24.52 |
| Aha. | 17:24.54 |
jogux_mac | robin may well be the first person to try doing this from anything other than c: | 17:24.59 |
Robin_Watts | The problem is the font init, I think. | 17:25.08 |
jogux_mac | yes | 17:25.37 |
Robin_Watts | and that's an EPAGE_DIRECTORY problem. | 17:25.47 |
jogux_mac | yeah. but it should 'just work'. I don't have to manually set EPAGE_DIRECTORY. | 17:26.30 |
| (I'm not sure /why/ it just works) | 17:26.40 |
pedro | Open of /d:/cvs/artifex/sot.git/epage/D:/cvs/artifex/sot.git/epage/platform/tgv/apps/windows-alien-vs2010/DEBUG/mcfs/carbon-blue-480.mcf/busy-background.png failed | 17:27.06 |
jogux_mac | oh. hm. did you set EPAGE_DIRECTORY? | 17:27.15 |
| the fonts should be in romfiles. | 17:27.22 |
Robin_Watts | EPAGE_DIRECTORY=d:\cvs\artifex\sot.git\epage | 17:27.31 |
jogux_mac | don't do that :-) | 17:27.35 |
Robin_Watts | Ah. Let me unset it. | 17:27.38 |
jogux_mac | we should, at some point, fix the app to not assert in an utterly random place if startup fails :( | 17:28.21 |
Robin_Watts | I suspect I have it set by default in the environment. | 17:28.54 |
| Bingo! | 17:29.48 |
jogux_mac | :) | 17:30.00 |
Robin_Watts | Thanks guys. | 17:30.06 |
| I will now prepare some reviews with what I've learned. | 17:30.16 |
| Was anyone waiting on reviews from me ? | 17:30.28 |
jogux_mac | robin_watts : I'm not sure it's ready for review, but comments on my release script welcome (last commit in my repo) | 17:30.54 |
Robin_Watts | yes, of course, goldfish memory strikes again. | 17:31.14 |
jogux_mac | or less politely put, it's ready to go but I'm not saying so and instead betting on you lot deciding there are 4 billion 'must have' features :-) | 17:31.38 |
pedro | does seem worthwhile for us to do some sanitisation pre-build, post-build and runtime - we seem to have a bunch of errors reported (and lost in the scrollback). Wasn't so easy when we had huge numbers of configurations, but should be feasible now with a few SOT/lib builds | 17:32.36 |
Robin_Watts | Jogux, pedro: Minor thing, but can we try to keep log file messages to a 72 char width please? | 17:42.29 |
| otherwise gitk zaps them off the right hand side of the window. | 17:42.44 |
| And not just gitk, git log etc too. | 17:43.11 |
| Jogux: Can we have comments in the text files? | 17:44.31 |
| Like "# Very important to use -define=WE_LOVE_STEVE with ios" etc. | 17:45.02 |
| oh, it seems we can. excellent. | 17:45.18 |
| Are we sure that calling subprocess is safe on windows? If it was safe, then why the hell are we using os.spawnve everywhere? | 17:45.54 |
| Jogux: I'm not the best person to review the ios stuff, but it all looks sane to me. | 17:47.36 |
Jogux | robin_watts : I'm presuming subprocess postdates most of the epage build system. or at least, am hoping that. | 18:10.33 |
Robin_Watts | ok. | 18:11.10 |
| We can try it and see. If it causes problems we can cut and paste the doCommand stuff back. | 18:11.32 |
Jogux | nods. | 18:11.41 |
Robin_Watts | If it doesn't cause problems we can probably simplify the rest to use subprocess :) | 18:11.50 |
Jogux | robin_watts : actually maybe I should allow comments at the end of lines too. I only allow purely comment lines just now. | 18:12.12 |
Robin_Watts | Jogux: ah, possibly. | 18:12.33 |
Jogux | I'm pondering taking any extra params to release.py and shoving them onto the build line too. | 18:13.05 |
| and maybe hardcoding it to look it epage/release-configs, and not require the .txt. | 18:13.25 |
| so it'd be release.py ios-appstore -debug to build the appstore release, but with debug enabled. or something. | 18:13.46 |
Robin_Watts | Yeah. | 18:13.56 |
| Only that might encourage people to do release builds without updating the scripts. | 18:14.09 |
Jogux | yeah, sort of. | 18:14.23 |
Robin_Watts | but I do like the idea of -debug and -fortify being available. | 18:14.25 |
Jogux | need to do something about --setvers too, as we need to pass that through to releases. | 18:14.36 |
| (filled out with issue number etcetc) | 18:14.51 |
| I'm assuming we would still put the build command into the release spreadsheet, it would just be that the commands are a hell of a lot shorter. | 18:15.25 |
Robin_Watts | Jogux: 3 more commits on robin/master | 18:26.16 |
| pedro: OK. I've got my commits on robin/master now (for the openssl and updated solution etc). Now I can finally start looking at the line spacing stuff :) | 18:57.17 |
| pedro: In your travels through the docs have you found a good example .docx for poor line spacing? | 19:00.05 |
| bug 694998 talks about a .doc file, and we've since decided that .docx is higher priority than .doc | 19:00.32 |
norbertj | Robin_Watts: I'm looking through the cmpbmp output for the jpeg-patch, pxl jobs. It seems that the images are more elongated (i.e. not missing scans). Can we compare those bitmaps with reference output? | 19:00.46 |
Robin_Watts | norbertj: What do you mean by reference output? | 19:01.28 |
norbertj | Robin_Watts: like paperoutput from e.g. QualityLogic or a reference printer | 19:01.55 |
Robin_Watts | norbertj: No. The cluster doesn't have such reference output available. | 19:02.24 |
| It purely works spotting changes between versions of our own code (i.e. it's regression testing, not correctness testing) | 19:02.53 |
norbertj | Robin_Watts: Ok, I was just looking for a way that proves that the output now generated is correct. E.g. by printing on a HPLJ4600 and verify against ghost-output to see if it looks the same size. Some of the differences are quite clear. | 19:05.44 |
Robin_Watts | 24 and 26 would seem to be odd. | 19:05.52 |
| 24 and 28 probably easier to see. | 19:06.10 |
| 30 and 33 too. | 19:06.24 |
| 36 and 38 | 19:06.37 |
| but I could believe the earlier ones were improvements. | 19:07.16 |
pedro | Robin_Watts: cool - I'll take a nosey at the commits. I've found most of the .doc files reproduce the same layout issues when you save them as .docx | 19:07.18 |
Robin_Watts | pedro: D'Oh. That didn't occur to me :) | 19:07.42 |
pedro | as long as they reproduce the same fonts/styles they should be ok | 19:08.20 |
Robin_Watts | pedro: I am tempted to start with some very simple 1 font, 1 font size tests. | 19:08.49 |
pedro | nods | 19:08.56 |
| I'm doing the same with most of the other MSWord bugs; creating degenerate cases. Its quite painful to debug otherwise | 19:09.28 |
| Robin_Watts: the other thing I've noticed in passing is that there are various properties which *may* be set such as kerning levels in the document styling xml which presumably gets read by Word to modify the font rendering; iirc we only use the built-in kerning tables when rendering fonts don't we? | 19:12.54 |
| things like condensed/expanded etc and font synthesis in general will probably cause us a bit of grief | 19:13.35 |
Robin_Watts | pedro: AFAIK, we just use the defaults. | 19:14.19 |
| Yeah, if we 'only' go wrong with stuff that does silly things with condensed etc, I would count that as progress :) | 19:14.52 |
pedro | heh | 19:15.04 |
Robin_Watts | Oh great. I just constructed a file with libreoffice. Just some lorem ipsum text. | 19:18.04 |
| I save it, load it back in, and open office renders it differently. | 19:18.15 |
Jogux | :-D | 19:18.36 |
Robin_Watts | Let me install 4.2.2 and retry. | 19:18.57 |
pedro | its good that they standardised ooxml | 19:24.19 |
Robin_Watts | And 4.2.2 loads it in a completely different way. | 19:27.00 |
fdncred | Greetings fellow mupdfers/gsers. I need some help if possible. I have a pdf that mupdf.exe will not view. It keeps throwing an exception in colorspace.c on this line: assert(srcs->to_rgb && dsts->from_rgb); | 19:37.17 |
| the dsts->from_rgb is null | 19:37.35 |
| I can email the PDF if that would be helpful, if anyone has time to take a look. | 19:38.28 |
Robin_Watts | fdncred: Are you using the latest version from git? | 19:44.00 |
| If so, open a bug and attach it there. | 19:44.09 |
fdncred | Robin_Watts: I downloaded it on 2/25 - so it's fairly recent | 19:48.31 |
| Robin_Watts: downloaded from git that is | 19:48.53 |
| i just noticed this error message in the debug console "jpx file and dict colorspaces do not match" | 19:51.20 |
Jogux | robin_watts : your commits nominally look okay to me. | 20:06.24 |
pedro | Robin_Watts: yup, all looks good to me too | 20:37.00 |
Jogux | hi marcosw_ - I think Robin was suggesting that me/pedro should be added to tech if that's possible/sensible? | 21:45.12 |
| marcosw_ : also, we could probably do with getting at least one copy of testdocs (~576G) onto the beefy machine in your garage; I guess that means at least a 1T disc, and if we're doing one it's probably sense to upgrade 2 or 3 or nodes which would also let us spread out ATS a bit if that becomes necessary. | 21:47.48 |
marcosw_ | Jogux: will do both (adding you and pedro to tech and upgrading the storage of a couple of the machines in my garage). | 21:48.58 |
Jogux | marcosw_ : brilliant, many thanks :) | 21:49.12 |
henrys | marcosw: not sure if you saw but I uploaded a file for your plotter | 21:50.18 |
marcosw_ | henrys: I won't be home until tonight but will plot it then. | 22:05.08 |
henrys | Robin_Watts: is there a script to add files to the vcproj files or have you been doing it manually? | 22:06.56 |
| or mvrhel_laptop ^^^ I guess you'd know about that too. | 22:08.40 |
marcosw_ | Robin_Watts: can you email me the preferred email addresses for Jogux and for pedro to use for the tech alias? | 22:10.01 |
Robin_Watts | henrys: Which vcproj files? | 22:45.54 |
| The gs ones I add files manually. | 22:46.07 |
| Likewise mupdf (though it harder to forget there, cos that's a real solution rather than a wrapper around a makefile) | 22:46.27 |
| I have a script to do a recursive pass over a directory tree and spit out XML that I used to create something I could hand hack into the original Smart Office solution, but additions/maintenance will be done manually from now on. | 22:47.30 |
henrys | Robin_Watts: yes gs, I thought there was some way you could get nmake to output dependencies and we could use that. No big deal. | 22:47.47 |
Robin_Watts | henrys: There might be, but that wouldn't gather the files into functional groups. | 22:48.19 |
| (which mvrhel can take credit for) | 22:48.25 |
henrys | Robin_Watts: ah right ⦠I just noticed because I was testing gdevxps and noticed it wasn't in the solution explorer. | 22:49.04 |
| in the ide | 22:49.09 |
Robin_Watts | henrys: Ah, if it's a new file that you added, then I probably missed it. | 22:49.27 |
henrys | Robin_Watts: Does it produce an error if the user takes devices out of the makefiles without changing the vcproj files? | 22:50.47 |
Robin_Watts | henrys: Not while building. | 22:51.08 |
| Possibly while searching, but even then I expect VS should cope. | 22:51.23 |
| Forward 1 day (to 2014/04/01)>>> | |