| <<<Back 1 day (to 2014/03/17) | 2014/03/18 |
Robin_Watts | tor8: (For the logs) I could make it so that we both decompress the image and store the original compressed data for inline streams. | 00:11.58 |
| I'll check for leaks in the morning, and maybe discuss this some more with you if you're around. Thanks! | 00:12.30 |
| marcosw_: You here? | 00:30.32 |
marcosw_ | Robin_Watts: I am | 00:30.40 |
Robin_Watts | Did the NAS get backed up? | 00:30.47 |
marcosw_ | Robin_Watts: yes, it's been rsync'd to a 4 TB USB drive. | 00:32.11 |
Robin_Watts | marcosw_: Cool. And that was points? | 00:32.26 |
marcosw_ | also correct. | 00:32.36 |
| Paul actually asked me about the NAS back-up status earlier today, I guess he didn't cc you in the question (or maybe I didn't do a reply-all). | 00:33.19 |
Robin_Watts | I didn't see question or answer. | 00:33.50 |
marcosw_ | I suggested to Paul that even though the data is backed up a 4 TB external drive isn't what I would consider reliable, so I suggest making a copy on the NAS of anything that is going to be edited. | 00:34.10 |
| just checked, Paul didn't cc anyone when we emailed me. | 00:34.29 |
Robin_Watts | Wow. My old picsel home dir is on there. | 00:34.52 |
marcosw_ | Robin_Watts: hmm, there is a size mismatch between the NAS used data and the backup used data, so I wonder if some of it didn't get copied for some reason. give me a minute... | 00:35.30 |
Robin_Watts | tor8: (For the logs) Updated non leaky commit on robin/master. | 01:00.59 |
marcosw_ | Robin_Watts: I don't quite understand the picsel NAS. If I do a df it says 2.9 TB in use but if I run du on the volume it says 1.8 TB in use. The backup I made has 1.8 TB in use with df and 1.8 TB in use with du. The only thing that I can think of is the NAS is using really big block and so there is a lot of wasted space. | 01:31.19 |
gaspa | hi, is this the right place to ask things about muPDF? | 10:14.30 |
| (Just in case: i'm trying to understand how works text selection) | 10:14.34 |
Robin_Watts | gaspa: It is. | 10:15.18 |
gaspa | Robin_Watts, thanks ;) | 10:21.41 |
| so... the question is: given a string, is there a facility function to get a bounding box of the text ? | 10:22.38 |
| (given a fz_page, at least, i guess) | 10:23.12 |
Robin_Watts | gaspa: There is no potted function to do that. | 10:26.11 |
| but you can do it using the text extraction device. | 10:26.23 |
| If you run the text extraction device on a page, you'll get a structure of blocks/lines/spans/chars. | 10:26.56 |
| You can search within that for any string you want, and from that you can get the bbox. | 10:27.08 |
| We do that in various viewers (such as the android one) to highlight searched for text. | 10:27.39 |
gaspa | Robin_Watts, ok, thanks, so, more or less is what is done here: http://www.mupdf.com/docs/browse/source/fitz/stext-output.c#278 ? | 10:29.59 |
Robin_Watts | yes. stext-output turns the structures into html etc. | 10:34.08 |
lhf | I'm getting this when I do "git pull": fatal: read error: Connection reset by peer | 10:36.40 |
Robin_Watts | git remote -v ? | 10:39.13 |
lhf | origingit://git.ghostscript.com/mupdf.git (fetch) | 10:39.48 |
| origingit://git.ghostscript.com/mupdf.git (push) | 10:39.51 |
| (Hi Robin!) | 10:40.01 |
Robin_Watts | Hi. We don't use git:// | 10:40.15 |
| Try: | 10:40.17 |
| git remote rm origin | 10:40.25 |
| git remote add http://git.ghostscript.com/mupdf.git | 10:40.38 |
lhf | It was working fine till now. | 10:40.45 |
Robin_Watts | OOps. git remote add origin http://git.ghostscript.com/mupdf.git | 10:40.51 |
lhf | As per instructions in http://www.mupdf.com | 10:41.07 |
Robin_Watts | hmm. Not sure then. | 10:41.34 |
chrisl | I think we've had reliability troubles with the git protocol, it would probably be safer to use http in the future | 10:42.54 |
lhf | I've done as you've suggested and now naked "git pull" no longer works: | 10:43.44 |
Robin_Watts | git remote -v again? | 10:44.01 |
lhf | Please specify which branch you want to merge with. | 10:44.04 |
| "git pull origin master" worked. | 10:44.48 |
| git remote -v returns | 10:45.11 |
| originhttp://git.ghostscript.com/mupdf.git (fetch) | 10:45.15 |
| originhttp://git.ghostscript.com/mupdf.git (push) | 10:45.19 |
Robin_Watts | git branch -u origin/master | 10:45.40 |
| assuming you are on master now., | 10:45.52 |
lhf | I'm sorry for all these questions. I know next to nothing about git. | 10:46.30 |
Robin_Watts | np. | 10:46.36 |
lhf | BTW I can confirm that the "git clone --recursive git://git.ghostscript.com/mupdf.git" no longer works. Perhaps you could fix the instructions in the first page of http://www.mupdf.com. | 10:47.09 |
Robin_Watts | Thanks. We'll tell tor8 about that later. He maintains the web pages. | 10:47.40 |
lhf | git clone --recursive http://git.ghostscript.com/mupdf.git seems to be working. | 10:47.57 |
| or not... | 10:48.26 |
| it is. It took a long time to print "Checking connectivity... done". | 10:49.54 |
| Perhaps it's a network problem after all. | 10:50.06 |
| ah, naked git pull works fine now. | 10:51.00 |
chrisl | We have an awful lot of git-daemon processes on casper, which seems bad :-( | 10:51.23 |
lhf | so does git clone --recursive git://git.ghostscript.com/mupdf.git | 10:51.32 |
| so, it's all fine. thanks. | 10:51.51 |
| Anyway, I'd like to say that MuPDF is a fine piece of software. I've been using it to read PDF files for a project by adapting the SVG driver and it's been a pleasure. Thank you all! | 10:58.37 |
Robin_Watts | What adaptation have you done? | 10:59.31 |
lhf | I've used the SVG driver to write a driver for my own API. | 11:00.03 |
| Nothing to report except the couple of emails I've sent you. | 11:00.16 |
Robin_Watts | Aha. Morning tor8. | 11:00.21 |
tor8 | morning Robin_Watts | 11:00.27 |
Robin_Watts | I've updated that leaky commit. | 11:00.37 |
tor8 | I just tested git:// and it works fine for me | 11:00.38 |
lhf | It's back to normal, yes. | 11:01.16 |
tor8 | we've had the git: server get tied up with too many dead processes from the web server twice, but that doesn't seem to be the case now | 11:01.22 |
chrisl | I just killed all the git-daemon and git-upload-pack processes, that probably fixed it | 11:01.41 |
lhf | I was getting fatal: read error: Connection reset by peer | 11:01.42 |
tor8 | chrisl: ah, well, that's probably it then! | 11:01.52 |
lhf | Thanks for the prompt action. | 11:02.01 |
tor8 | chrisl: same thing I did last time it was stuck. | 11:02.05 |
chrisl | tor8: I know, I actually remembered :-) | 11:02.21 |
lhf | Robin_Watts: The last bug report I sent you was about the SVG driver outputting stroke-width=0, which means "don't paint", not hairline. | 11:03.22 |
chrisl | Doh, of course, why would Adobe be consistent?!?! | 11:03.59 |
Robin_Watts | right. I'm buried in other stuff at the moment, so I haven't looked at that. | 11:04.14 |
lhf | No hurry. | 11:04.22 |
| Just keep on doing what you're doing, which is fine. | 11:04.35 |
| I still have to find time to write that Lua binding to fitz. | 11:05.09 |
| Well, I'll be off now. Thanks for fixing the git repo. | 11:06.16 |
Robin_Watts | np. | 11:06.24 |
tor8 | Robin_Watts: inline images are still decompressed twice? | 11:08.24 |
Robin_Watts | Yes. | 11:11.44 |
tor8 | that's unfortunate. | 11:12.01 |
Robin_Watts | They have to be decompressed in order to know how much data they use. | 11:12.18 |
tor8 | would it be a decent solution to keep the decompressed data around for one consumption? | 11:12.21 |
Robin_Watts | possibly. I will look at that. | 11:12.34 |
tor8 | Robin_Watts: fab. I've got a simple portable dtoa function from plan9 that I intend to reuse for libjs, and I think it might help us for fitz printf as well. | 11:14.03 |
| Robin_Watts: anyway, I'm off for lunch. back in an hour. | 11:20.07 |
Robin_Watts | tor8: cool. | 11:20.48 |
henrys | tkamppeter: what's happened with printer discovery in cups? it used to work for me but seems to be broken in 13.10. I've seen other talk about it breaking as well. | 13:27.01 |
tkamppeter | henrys, which kind of printer? what exactly happens, what do "others talk about breaking it as well"? Any links? | 13:57.16 |
henrys | tkamppeter: for for find new printer - no printer found but I have a wireless epson which windows sees. | 14:02.22 |
| tkamppeter: do I need avahi now? | 14:03.02 |
tkamppeter | henrys, yes, most printers are detected by the dnssd CUPS backend and this needs avahi-daemon running. | 14:05.38 |
| henrys, only some are discovered by the snmp CUPS backend which does not ned avahi-daemon. | 14:06.20 |
henrys | tkamppeter: we were considering doing a cups port to android and iOS. Do you have any idea how difficult this would be? | 14:07.03 |
tkamppeter | henrys, for debugging you can also run avahi-discover (also needs avahi-daemon) and see what is available and compare with what the dnssd backend of CUPS finds. | 14:08.00 |
henrys | tkamppeter: why is it that snmp can't work sometimes? Is this a problem on the printer side or is the communication proprietary? | 14:10.17 |
| encrypted? | 14:10.33 |
tkamppeter | henrys, seems that more printers use Bonjour to broadcast themselves and less use SNMP. | 14:12.01 |
henrys | makes coffee before the meeting | 14:14.43 |
tkamppeter | You want to port CUPS to iOS and Android? Most things should work but there are possibly missing some infrastructure elements, like avahi, libdnssd, or a similar Bonjour infrastructure for network printer discovery, and also there is possibly no D-Bus for notification exchange (not so important). | 14:15.00 |
henrys | tkamppeter: want do some moonlighting ;-) ? | 14:20.34 |
ray_laptop | henrys: why do we want CUPS on android ? To print, all we need is bonjour or SNMP or maybe ndssd (printer discovery). Then we send data to it (raster) using IPP | 14:20.42 |
| what we *don't* get is all of the back end cups raster to printer specific drivers since those packages are linux only | 14:21.39 |
henrys | ray_laptop: I don't think you'll get the sort of printer coverage we already have in our current office app. | 14:22.14 |
| SOT | 14:22.18 |
| with your plan. | 14:22.33 |
ray_laptop | so if we drive "smart" printers (i.e., not ones with some funky proprietary raster needs) we can drive PCL, PS, PDF, ESC/P printers (the buk of them) | 14:22.53 |
| does embedded general provide some other formats than those that gs already knows how to generate ? | 14:24.12 |
henrys | ray_laptop: that discussion has to go private, maybe we should have a Skype discussion alongside the meeting. | 14:25.21 |
ray_laptop | henrys: I'd have to restart my Skype -- just let me know. I tend to shut it down because I get lots of junk messages otherwise | 14:26.19 |
Robin_Watts | Is meeting in 6 minus? Or 66 mins? Or 126 mins? | 14:27.03 |
| s/minus/mins// | 14:27.12 |
| Timezones confuse me. | 14:27.17 |
henrys | Robin_Watts: 3 minutes | 14:27.19 |
| tkamppeter: I assume there are many printers that only do snap and don't support bonjour or newer protocols right? | 14:29.47 |
| s/snap/snmp | 14:29.55 |
tkamppeter | henrys, what do you mean with moonlighting? | 14:30.01 |
ray_laptop | Robin_Watts: if you have time after the meeting, I have questions about the stuff in gsht.c (gx_ht_construct_threshold) that mvrhel said /might/ be yours | 14:30.09 |
mvrhel_laptop | good morning | 14:30.38 |
tkamppeter | henrys, ray_laptop, see also my plans for CUPS in Ubuntu Touch (Ubuntu's phone OS): https://blueprints.launchpad.net/ubuntu/+spec/client-1305-printing-stack-with-mobile-in-mind | 14:30.47 |
henrys | tkamppeter: I was thinking we would hire you to port cups ⦠this is all quite preliminary and up in the air right now. | 14:31.49 |
tkamppeter | henrys, ray_laptop, most of the important steps I have already done as you can see this Blueprint: Driver-less printing on printers with known protocols, starting daemons on-demand, sorting out what of CUPS and cups-filters is actually needed and splitting our packages appropriately, ... | 14:32.34 |
marcosw | Robin_Watts: I've created accounts for pete and josephn on points. Also given you sudo access in case you need it in the future. | 14:32.38 |
henrys | I thought I'd try another go at the meeting times ;-) | 14:32.43 |
| June 2/3 with flyout saturday before for optional russian river trip. Does that work for everyone here? | 14:33.22 |
kens | Yes,I checked, its fine for me | 14:33.31 |
Robin_Watts | Fine for me. | 14:33.55 |
ray_laptop | henrys: yes, it works for me | 14:34.01 |
henrys | please welcome joseph and peter (jogux and pedro), they will be working with us on smart office, please answer any questions they have. | 14:34.27 |
ray_laptop | June 1 is my twin's birthday, but I wouldn't have to fly out until Sat PM | 14:34.40 |
mvrhel_laptop | June 1 is sunday ray_laptop | 14:34.59 |
ray_laptop | welcome, jogux and pedro | 14:35.03 |
| mvrhel_laptop: sorry Sunday PM | 14:35.20 |
mvrhel_laptop | henrys: those dates are fine with me | 14:35.21 |
ray_laptop | but the dates are fine | 14:35.29 |
Jogux | thanks for the welcome :-) looks forward to working with all you guys | 14:35.45 |
ray_laptop | Jogux: are you in the UK along with Robin_Watts, paulgardiner, kens and chrisl ? | 14:36.39 |
henrys | for SOT stuff - we want to take out these nag screens that pop up periodically while running the smart office 2 app. | 14:36.41 |
Robin_Watts | ray_laptop: Jogu and Pete are in Scotland. Which is like a whole different country. Almost. | 14:37.29 |
kens | paulgardiner : henrys requested ok on meeting times above | 14:37.45 |
Jogux | robin> everyone knows everywhere in the UK is kinda near London ;-) | 14:38.10 |
| henrys> the "give us your email" ones? that would be awesome | 14:38.29 |
henrys | I was reading about an independent scotland vote, is that just something that comes up now and then or is this serious. | 14:38.41 |
| Jogux: right | 14:38.46 |
kens | henrys serious, but its not going to happen | 14:38.56 |
Jogux | henry> serious. We get to vote in 6 months time. | 14:38.58 |
paulgardiner | Meeting date fine for me too | 14:39.03 |
kens | I think Salmond misjudged | 14:39.09 |
Jogux | should I be paying attention to all this discussion about meeting dates? | 14:39.42 |
henrys | you're all "euros" to me ;-) | 14:39.43 |
kens | Jogux : no not really, its our next staff meeting | 14:40.01 |
Robin_Watts | It looks like it won't happen at the moment, but the yes campaign are making less asses of themselves than the no campaign at the moment, so I wouldn't like to bet either way :( | 14:40.03 |
kens | Unless Miles decides to fly you all out to Califronia | 14:40.11 |
Jogux | kens: goodgood. cos I wasn't. but have no objection in principle to being flown to California :-) | 14:40.22 |
marcosw | Jogux pedro and sebras: casper is acting strange so I'd like to reboot it. | 14:40.23 |
Jogux | marcosw: fine by me | 14:40.32 |
kens | Robin_Watts : Yes that's true, but I still think it will come down to the economy in the end | 14:40.35 |
ray_laptop | An interesting android note for finding printers on Android is http://developer.android.com/training/connect-devices-wirelessly/nsd.html#discover | 14:40.52 |
henrys | Jogux: no my goal is to convince miles that we should visit you. | 14:40.53 |
kens | Scotland is nice in the summer. | 14:41.07 |
Jogux | henrys> Scotland would be an awesome place to visit, especially if you've never been before | 14:41.16 |
kens | Of course, determining when summer will be could be a problem | 14:41.17 |
pedro | it is, but its only summer for around 3 days ;) | 14:41.23 |
Jogux | I can do that nag dialogue thing if wanted. Is there currently a plan to put a new version on the appstore(s)? | 14:42.13 |
| kens / chrisl : which part of the UK are you in? | 14:42.55 |
henrys | Jogux: yes please have at it if that is okay with paulgardiner . | 14:42.57 |
| paulgardiner might have wanted to do it. | 14:43.20 |
kens | Jogux I'm near Gatwick, but I'm actually a Scot, as is chrisl | 14:43.25 |
paulgardiner | Sounds good to me, although I seem to have missed the previous discussion on it | 14:43.26 |
Jogux | the bigger issue at least for the appstore is upgrading to the ios7 sdk & retesting. | 14:43.50 |
ray_laptop | I'm 1/4 Scot :-) | 14:43.52 |
| and 1/2 Czech | 14:44.01 |
henrys | the other sot stuff I can discuss here is we do need a printer module. Pricing for both soultions we've looked it is not doable | 14:44.16 |
tor8 | ray_laptop: and 100% american? ;) | 14:44.28 |
Jogux | ray_laptop : :-) I'm actually English. | 14:44.35 |
ray_laptop | tor8: I don't often categorize myself that way. More like a terran :-) | 14:45.01 |
henrys | chrisl: so we have a release candidate? | 14:45.19 |
kens | Its up, but there's a problem | 14:45.30 |
| We get 19 seg faults on current code | 14:45.39 |
marcosw | my wife is a 1/2 scot, clan Farquharson. | 14:45.40 |
ray_laptop | kens: did you open a bug on that ? | 14:45.45 |
henrys | and everyone will download and test? | 14:45.46 |
kens | ray_laptop : no I just mailed it to you :-) | 14:45.54 |
paulgardiner | henrys: and so we also need to update the app store versions with non-printing versions in the next couple of weeks? | 14:45.56 |
henrys | kens: I noticed that too all pcl->pdf | 14:46.00 |
| right? | 14:46.03 |
kens | Not quite | 14:46.06 |
ray_laptop | kens: hmm... I guess I better jump on that. | 14:46.08 |
kens | One is a PDF file and none are to PDF | 14:46.14 |
Robin_Watts | So, having established that we're all mongrels... | 14:46.16 |
henrys | paulgardiner: no we can leave printing as is. | 14:46.23 |
mvrhel_laptop | oh did miles get more runway? | 14:46.37 |
kens | ray_laptop : the diffs look like no problem, the seg faults are bad | 14:46.40 |
Robin_Watts | Can we be a tad careful about what we say about printing on here. Editing the logs gets dull after a while :) | 14:47.21 |
kens | My last regression test shows almost all are in pbmraw output | 14:47.27 |
| The only oine that isn't in pbmraw is in cups | 14:47.54 |
henrys | mvrhel_laptop: ? | 14:48.03 |
Robin_Watts | ray_laptop: I point the accusatory finger of blame at mvrhel_laptop :) | 14:48.13 |
| http://git.ghostscript.com/?p=ghostpdl.git;a=commitdiff;h=8024d7dbe68a1d9db1d23ca984f833cc82e055c9;hp=735a75a04f7413a80ecdf8885f3e95713c612827 | 14:48.15 |
mvrhel_laptop | what did I say? other than runway | 14:48.26 |
kens | Robin_Watts : if you mean the seg faults, I bisected them to Ray's commit | 14:48.49 |
Robin_Watts | ray said you were trying to blame me for gx_ht_construct_threshold :) | 14:48.50 |
henrys | mvrhel_laptop: was that a question about sot printing? | 14:48.51 |
sebras | marcosw: ok, do you want me to log out now? | 14:49.17 |
mvrhel_laptop | Robin_Watts: oh ray laptop had a line number where things go wrong and you were there according to blame | 14:49.24 |
marcosw | sebras: yes please | 14:49.24 |
henrys | marcosw: meeting dates okay ? | 14:49.38 |
| see above | 14:49.44 |
Robin_Watts | mvrhel_laptop: I might have tweaked the odd line, but the bulk of that code is yours AIUI. I don't claim to understand it :) | 14:49.58 |
mvrhel_laptop | marcosw: did you seem my email about the fix for the black point also | 14:50.01 |
| Robin_Watts: I actually stole this code from ray_laptop | 14:50.15 |
marcosw | marcosw: yes, assuming I can leave earlish on Tuesday (since I'm driving that shouldn't be an issue). | 14:50.22 |
| mvrhel_laptop: no. is that for the same customer that was have 16 bit color issues? | 14:50.48 |
mvrhel_laptop | ray_laptop: I can help out if you need | 14:50.49 |
| marcosw: yes | 14:50.58 |
marcosw | who | 14:50.59 |
sebras | marcosw: ok, will casper be backup soon after? | 14:51.10 |
marcosw | needs to move the mouse before typing | 14:51.11 |
henrys | marcosw: I need to coordinate the meetings now do you want to set up the web thing you suggested for next meeting. I'll go crazy doing this. | 14:51.21 |
marcosw | sebras: I can wait until after the meeting, since I believe you read irc on casper? | 14:51.36 |
Robin_Watts | http://www.doodle.com | 14:51.38 |
marcosw | henrys: it's called doodle.com | 14:51.43 |
henrys | marcosw: well chicago is fairly fixed. | 14:51.46 |
| marcosw: do we all have to sign in for that or can we just have an artifex login? | 14:52.11 |
mvrhel_laptop | marcosw: I sent you the email yesterday at 5:05. | 14:52.13 |
| with the commit for the fix | 14:52.20 |
marcosw | henrys: neither, it's one of those only those with the link can see information sort of things. | 14:52.34 |
mvrhel_laptop | and cc'd support | 14:52.39 |
henrys | marcosw: okay so I'll look at that. | 14:52.59 |
marcosw | you create a meeting date poll and send out the URL to everyone who you want to respond, the click and check out which dates work for them. | 14:53.06 |
| you can now even respond "possible but not convienent", which is useful with such a large group. | 14:53.42 |
| mvrhel_laptop: odd, I read my email this mornign already. | 14:54.09 |
| must have missed it. | 14:54.15 |
mvrhel_laptop | ok. let me resend... | 14:54.20 |
henrys | so we have a showstopper for the release? | 14:54.29 |
kens | Debatable. THe original RC didn't have Ray's fix for stripes, chrisl cherry-picked that in, but it causes segfaults. | 14:55.11 |
mvrhel_laptop | marcosw: resent. maybe you have me in your block list | 14:55.13 |
kens | We could drop it and live with the (I think long-standing) stripes, or wait until Ray fixes the seg faults | 14:55.30 |
ray_laptop | kens: the stripes have been in there for quite a while, but why they don't show up more often is a puzzle. | 14:56.04 |
henrys | ray_laptop? It's getting late in the month like to "ship this" | 14:56.06 |
kens | FWIW I think the reason ray missed the seg faults is because his commit failed to test, I suspect becaus eof the macpro problems marcosw addressed | 14:56.07 |
ray_laptop | give me a few hours to track it down. I can reproduce it on VS debug build | 14:56.42 |
Robin_Watts | It's possible the seg faults could be an easy fix. It all depends if they reproduce nicely or not. | 14:56.42 |
| bargain. | 14:56.51 |
kens | ray_laptop : you're ahead of me, I couldn't reproduce it at all | 14:56.58 |
henrys | we should finish at the hour, anything else for the meeting? | 14:57.20 |
mvrhel_laptop | can we pull in my last commit if we do this? I results in quite a few progressions at the black point | 14:57.28 |
ray_laptop | on a 32-bit build, this failed for me: debugbin/gswin32c -r300 -sDEVICE=pbmraw -o x.pbm -Z: /tests_private/comparefiles/Bug692057.pdf | 14:57.34 |
Robin_Watts | henrys: I had something, very quick. | 14:57.41 |
henrys | Robin_Watts: you guys changed time | 14:57.44 |
marcosw | mvrhel_laptop: I checked the spam folder and it wasn't there either. | 14:57.45 |
henrys | ? | 14:57.47 |
kens | AH I didn't try -Z: | 14:57.48 |
Robin_Watts | henrys: No, not til the end of the month. | 14:58.06 |
kens | henrys our clocks change at the end of the month | 14:58.06 |
henrys | Robin_Watts: go ahead | 14:58.07 |
Robin_Watts | The downscale factor stuff that various devices use. | 14:58.28 |
henrys | kens: okay Robin_Watts was confused about the time of the meeting start so... | 14:58.38 |
Robin_Watts | Normally we do -dDownScaleFactor=4 and it downscales by 4. | 14:58.42 |
kens | henrys so was I :-) | 14:58.47 |
mvrhel_laptop | its at 7:30 each morning. what is the confusion ;) | 14:59.04 |
Robin_Watts | we have special cases for 32 and 34 which are used for 3:2 and 3:4 scaling respectively. | 14:59.09 |
| but hintak (I think it was hintak) reports a case where he wanted to use downscales >= 32. | 14:59.59 |
kens | Can we go negative instead ? | 15:00.25 |
Robin_Watts | In order to accomodate that, we could push those special cases to (say) 320 and 340 or something. | 15:00.26 |
| or possibly go negative. | 15:00.34 |
ray_laptop | Robin_Watts: I think -32 and -34 was discussed for the special cases | 15:00.40 |
Robin_Watts | but that would change the api. | 15:00.40 |
| Are people happy with me changing the API ? | 15:00.51 |
marcosw | mvrhel_laptop: sorry, I have that email. I thought you said there was a second email. I sent the URL to the customer along with your comment and they said that it worked. | 15:00.58 |
kens | From my POV yes, | 15:00.59 |
ray_laptop | Robin_Watts: those few customers should be able to switch to the new values easily | 15:01.12 |
Robin_Watts | Or, I could leave stuff as 32 and 34 and allow negative values to be treated as positive values without the special cases. | 15:01.20 |
henrys | hintak has a habit of requesting features that will never be used, I guess you think this could get field use? | 15:01.33 |
kens | thinks that would be very confusing | 15:01.37 |
marcosw | Robin_Watts: that's silly. | 15:01.41 |
Robin_Watts | OK, I'll do the change. | 15:01.53 |
| Thanks. | 15:01.59 |
ray_laptop | Robin_Watts: I don't like the latter. I think making the switch (post 9.12) to -32 == 3:2 | 15:02.09 |
Robin_Watts | henrys: His use case was genuine. | 15:02.33 |
kens | I cna see it as legitimate | 15:02.42 |
Robin_Watts | He had a pcl file that only renders correctly at 600dpi. | 15:02.43 |
henrys | Robin_Watts: okay | 15:02.48 |
Robin_Watts | but it ended up way too large to actually be useful, so downscale was required. | 15:03.03 |
| So, is the meeting done? | 15:04.50 |
henrys | yes | 15:05.12 |
Robin_Watts | ray_laptop, mvrhel_laptop: The thresholding stuff then... | 15:05.21 |
| http://git.ghostscript.com/?p=ghostpdl.git;a=commitdiff;h=300c3ea8f2c82471c5e57655bd13549f1172c38a | 15:05.26 |
| That was the change that I made in here. | 15:05.34 |
| What was the problem ray was seeing? | 15:05.52 |
marcosw | can one of the git experts help me with a question? | 15:07.51 |
ray_laptop | Robin_Watts: the line that is confusing is t_level_adjust = byte2frac(t_level) * nshades / (frac_1_long + 1); (well actually the whole delta_sum as well) | 15:07.55 |
Jogux | marcosw : dunno if I qualify as an expert, but ask away :-) | 15:08.21 |
Robin_Watts | ray_laptop: delta_sum was from the original code, that you wrote :) | 15:08.38 |
sebras | marcosw: ok, bye! | 15:08.39 |
paulgardiner | henrys: I still have some confusion as to what gives with SOT printing, but maybe that needs talking about off line | 15:08.43 |
ray_laptop | Robin_Watts: the problem I get is with -dGrayValues=16 (pgmraw) the threshold t_level goes 4, 9, 7, ... NOT MONOTONIC | 15:08.56 |
Robin_Watts | I just introduced off. | 15:09.06 |
henrys | paulgardiner: it does need offline discussion, yes | 15:09.09 |
Jogux | paulgardiner: I /think/ I'm clear on that. | 15:09.09 |
marcosw | the recent 9.12rc doesn't show up for me when I do 'git tag' I see all the other ones and I've done a git fetch. | 15:09.15 |
ray_laptop | Robin_Watts: I don't think so. My original code is still in tiffsep | 15:09.21 |
Robin_Watts | ray_laptop: Well, revert 300ceea8 and see if it still does the same? | 15:09.49 |
marcosw | if I blow away the repository and run git clone the 9.12rc tag is there. | 15:09.56 |
ray_laptop | we have some spoiled soup here (too many cooks) | 15:10.07 |
Jogux | marcosw> possibly try git fetch -t - though I'm not sure that should help. | 15:10.20 |
Robin_Watts | ray_laptop: I would be entirely in favour of someone replacing this code with something that I could understand. | 15:10.44 |
marcosw | Jogux: yup, that did it. thx. | 15:10.56 |
ray_laptop | Robin_Watts: git blame has the t_level_adjust tagged as yours as well. commit 65aa942c | 15:11.07 |
henrys | so I'll keep an eye out for chrisl and tell him to read the logs about the release. | 15:11.37 |
marcosw | Jogux: it actually retrieved about half a dozen tags that were missing (e.g. ghostpdl-9.08rc1). Why is this? | 15:11.50 |
Jogux | marcosw : I think it's probably tags that are related to branches you're not tracking locally | 15:12.25 |
henrys | paulgardiner: so you will work with Jogux on the nag screen (see above)? | 15:12.34 |
Jogux | marcosw : which for these purposes probably means "branches you've checked out before" | 15:12.47 |
paulgardiner | henrys: yeah sure | 15:12.49 |
henrys | paulgardiner: I didn't know if you wanted to handle that one. | 15:12.50 |
marcosw | right, so the ghostpdl-9.12rc1 tag was in the gs912 branch. thx for clearing that up. | 15:13.08 |
ray_laptop | Robin_Watts: hmm... that was probably just whitespace :--( | 15:13.16 |
Jogux | henrys : was there an answer as to when we're targetting an appstore release? | 15:13.19 |
| marcosw : no problem | 15:13.30 |
ray_laptop | how do I ask git blame to start before the 65aa942c commit ? | 15:13.49 |
marcosw | pedro: okay with you if I reboot casper? You are still logged but it looks like you are idle. | 15:14.58 |
| chrisl: btw, when I do the release candidate testing I've been testing using the branch and not the tag (I was going to use the tag this time when I ran into the issue that Jogux just cleared up for me). Is one preferred over the other? | 15:16.15 |
pedro | marcosw> yeah, no problem | 15:16.25 |
henrys | Jogux: I think we should stick with March 31 with our modest changes. | 15:16.34 |
| paulgardiner? | 15:16.41 |
paulgardiner | henrys: yeah, I could handle that one. But I'm easy either wat | 15:17.09 |
| way even | 15:17.15 |
henrys | paulgardiner, Jogux: well draw straws | 15:17.41 |
Jogux | yeah, we'll figure it out. | 15:17.49 |
| how many app stores are we aiming to release on? ios, android, bada, symbian, amazon, windows, ...? (maybe I should just be asking Paul this) | 15:18.19 |
paulgardiner | We'll endeavour to avoid both or neither of us doing it | 15:18.19 |
henrys | paulgardiner: right xor please | 15:18.52 |
marcosw | casper is back up | 15:23.57 |
Jogux | is feeling uncomfortable working on smartoffice stuff without continuous integration in place :-S | 15:24.11 |
henrys | Jogux: I think we should be happy with iOS, android and windows. | 15:25.41 |
pedro | in the absence of ATS for the GhostDocs builds, is it worth setting up jenkins or similar just to get the bulk of the builds automated as an interim measure? | 15:26.13 |
henrys | Jogux: we are waiting for more store numbers ($) to figure out what is worth doing. | 15:26.33 |
| Jogux: I thought mace was setting all this off - how far off is that? | 15:26.54 |
| s/off/up | 15:27.05 |
Jogux | henrys : Dunno. I shall ask him :-) | 15:27.14 |
chrisl | henrys: sorry, I forgot about the time change.... | 15:28.27 |
paulgardiner | I was planning to look at getting the wiki up and running, but if anyone has previous experience... | 15:28.29 |
Jogux | paulgardiner : we don't have the picsel wiki data btw :( | 15:28.56 |
paulgardiner | Jogux: any further luck with the Windows build, btw? | 15:29.00 |
henrys | chrisl: no problem have a look at the logs but it looks like ghostbot went down a bit ago. | 15:29.03 |
Jogux | paulgardiner : pedro made some progress-ish. | 15:29.11 |
kens | logs are OK for me | 15:29.38 |
paulgardiner | Jogux: That's a pain. Strange for that to be missing | 15:29.44 |
pedro | paul> I'm working from a vs2012 build on windows7 at present (just for working through the missing content/layout issue) | 15:29.50 |
chrisl | henrys: I've been reading - from my POV nothing has changed: I was waiting to hear what ray_laptop was going to say about the seg faults, then either pull in a fix, or exclude the commit that caused them | 15:30.11 |
Jogux | paulgardiner : there's a far bit missing :-( I gave a list to Robin to pass onto Miles to pass onto admins to see if we can get any of it. | 15:30.15 |
paulgardiner | Jogux: okay great | 15:30.33 |
chrisl | marcosw: in general, I'd rather you used the RC tag, but overall, it's not a big deal | 15:31.19 |
henrys | Jogux: yes we would like to get all that stuff - I saw the email and forgot about it, sorry. | 15:31.24 |
sebras | marcosw: thanks! now I feel sufficiently restarted. :) | 15:31.45 |
Jogux | henrys : great, thanks | 15:32.07 |
marcosw | chrisl: will do going forward, now that I know about "git fetch -t" | 15:32.12 |
henrys | Jogux: so the amazon build is different than android? | 15:33.22 |
paulgardiner | Ah, so ATS vms missing too. :-( | 15:33.30 |
Jogux | henrys : very slightly - and completely different submission process (it involves a wait for review, similar to iOS) | 15:33.47 |
henrys | chrisl: from the logs I think the meeting dates and the release showstopper are for you. | 15:34.43 |
ray_laptop | chrisl: I'll let you know ASAP | 15:34.44 |
Jogux | paulgardiner : yep :-( basically nothing we hoped for was on the NAS! (apart from all the review system comments) | 15:34.59 |
ray_laptop | mvrhel_laptop: I'm afraid that the t_level_adjust issue comes from your commit: http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=5a435470d3b21c6466825bd1534a5b136a1bb7d8 | 15:35.42 |
paulgardiner | So I can read my arguments with Neil about Unidisp, but besides that not a lot | 15:35.49 |
ray_laptop | mvrhel_laptop: but I have to look into the SEGFAULT first. I'll hunt you down later | 15:36.05 |
mvrhel_laptop | ray_laptop: ok. let me review | 15:36.44 |
chrisl | henrys: so, we're currently going for June 2/3 for the staff meeting? | 15:37.27 |
henrys | chrisl: and you are off to the russian river sunday. | 15:38.17 |
Jogux | paulgardiner : sounds about right. :-) | 15:38.24 |
chrisl | henrys: yeh, that's fine with me | 15:38.42 |
Jogux | henrys : mace says he's not yet been asked to setup ats. I guess it may depend on how quickly we can get an answer from via Miles etc as to whether we can get hold of the ATS VMs. | 15:46.08 |
mace | Jogux: :) | 15:47.09 |
chrisl | Jogux: in belated answer to your question: I live in Hampshire (from Edinburgh originally......) | 15:51.33 |
Jogux | chrisl : ah right, cool. :-) | 15:52.06 |
henrys | mace:okay I'm lost, weren't you just working with marcosw on setting up SOT testing? is that not AST? | 15:53.30 |
chrisl | Argh, I think my mouse is dying.... (computer, not pet!) :-( | 15:54.47 |
ray_laptop | henrys: ATS ? | 15:59.10 |
henrys | s/AST/ATS | 16:00.39 |
pedro | ray_laptop Picsel's automated test system - build and test for all the branch/product variants | 16:00.50 |
ray_laptop | pedro: thanks | 16:01.05 |
henrys | ray_laptop: oh I thought you were correcting my transpose. | 16:02.44 |
| marcosw: weren't you and mace working on setting up SOT regression testing or was that a dream I had. | 16:03.31 |
ray_laptop | henrys: I was. But I wanted to be polite to pedro as well | 16:04.04 |
chrisl | I've always said Artifex needs more TLAs! | 16:04.07 |
Robin_Watts | henrys: SOT regression testing == ATS. | 16:04.33 |
ray_laptop | yeah. Miles bought a bunch. SOT, FBS, ATS, ... | 16:04.48 |
chrisl | Easier than coming up with our own | 16:05.04 |
Robin_Watts | Jogu (who is geographically near mace and probably is in contact with him) said earlier that "mace says he's not yet been asked to set up ATS" | 16:05.18 |
ray_laptop | unfortunately, ATS used to refer to QL Application Test Suite | 16:05.38 |
Robin_Watts | ray_laptop: We will learn to tell from context, I am sure. | 16:06.03 |
henrys | Robin_Watts: thought it was in the works ;-( | 16:06.15 |
chrisl | Well, "Application Test Suite" is a rather generic term - I bet lots of companies has things called that | 16:06.33 |
ray_laptop | robin_watts: right. Until we start testing the QL Office ATS with the ATS ;-) | 16:06.40 |
Robin_Watts | henrys: So did I. | 16:06.48 |
Jogux | henrys : it'll be magnitudes of order easier if we can get these VMs. | 16:07.00 |
mvrhel_laptop | ray_laptop: ok yes it appears that 570f1ca was mine.. let me see if I can understand what I was doing. ray_laptop: do you have the file that has the issue with this? | 16:07.49 |
henrys | mace: what is the hour estimate for setting this up? | 16:08.02 |
Robin_Watts | It may be easiest for Miles to put jogu directly into touch with the admins. | 16:08.10 |
chrisl | mvrhel_laptop: before you disappear off into another bug: your BPC commit? How sure are we about pulling it into the release (it seems to have cause a *lot* of diffs) | 16:08.48 |
Robin_Watts | They are in the same city, and if there are physical drives it'll be much easier for Jogu to pick them up for us rather than to arrange shipping etc. | 16:08.58 |
marcosw | henrys: we discussed setting up SOT regression testing but I wasn't actively working on it. I thought mace was setting up the server and I would be setting up the nodes to talk to it when it was ready. | 16:09.03 |
mvrhel_laptop | chrisl: all the diffs are at the black points and are a progression. I stepped through them with bmpcmp and also checked that the black value on the pkmraw output was correct | 16:10.58 |
mace | looks | 16:11.29 |
chrisl | mvrhel_laptop: okay, I'll pull that commit onto the release branch... thanks | 16:12.05 |
mace | i was asked for time estimates, which I said I couldn't give accurate figures for since ats is a big beast that's changed a bit since i last worked on it however many years ago :) | 16:12.11 |
mvrhel_laptop | thanks chrisl | 16:12.11 |
henrys | Robin_Watts, Jogux whatever gets it set up fastest. Tell me what you want me to do. | 16:12.19 |
mace | i didn't get a nod from miles to proceed on it | 16:12.33 |
Robin_Watts | henrys: Well, we should contact the admins quickly. | 16:12.40 |
| If they have the VMs then this is a very quick route. | 16:12.53 |
henrys | okay so mace was going to do it without the vm's, is that correct? | 16:13.19 |
Robin_Watts | If they don't (and I suspect that if they don't they will tell us quickly), we can commit to the slower route of redoing the work. | 16:13.19 |
Jogux | Robin_Watts : agreed | 16:13.32 |
mace | henrys: yeah | 16:13.48 |
Robin_Watts | mace can do it either with or without, but without is harder. | 16:13.50 |
stars | Hi all here... the question is about mudraw... How to identify bulletted lines to <ul> <li></li></ul> in the html extraction? | 16:14.01 |
Robin_Watts | And even if he does it *with* the VMs there will likely be some changes needed. | 16:14.05 |
mace | i wasn't aware of the existance of any vm's until about 15 minutes ago :) | 16:14.07 |
Robin_Watts | because our servers aren't in a single room, and we want to use git rather than cvs etc. | 16:14.27 |
| stars: MuPDF text extraction gives a set of structures based upon the data on the page. | 16:15.13 |
Jogux | without is going to be tricky as we need to setup the build clients, and we have zero instructions on how to setup the build clients (and the windows one is proving to be a PITA just now) | 16:15.16 |
Robin_Watts | blocks/lines/spans/styles etc. | 16:15.28 |
marcosw | mvrhel_laptop and chrisl: there are some recent differences that I'm still looking at. I don't know if they are caused by your black point changes, but take a look at: | 16:15.38 |
| tests_private__comparefiles__035-01.ps.pbmraw.300.0.page_1.raw | 16:15.38 |
| tests_private__comparefiles__ridt91.eps.cups.300.1.page_1.raw | 16:15.39 |
| tests_private__pdf__sumatra__656_-_crashes_while_rendering_filterless_inline_image_.pdf.pbmraw.300.1.page_1.raw | 16:15.39 |
| tests_private__pdf__sumatra__656_-_crashes_while_rendering_filterless_inline_image_.pdf.pbmraw.72.0.page_1.raw | 16:15.39 |
mace | Jogux: that assumes that you care about 90% of the old build types | 16:15.41 |
Jogux | mace : we're struggling to build the windows smartoffice :-( | 16:15.55 |
Robin_Watts | stars: Broadly you'd have to do your own analysis on the data in those structures. | 16:15.58 |
mace | Jogux: ah | 16:16.12 |
mvrhel_laptop | marcosw: ok let me look | 16:16.24 |
Jogux | mace: though pedro is getting there | 16:16.31 |
mace | vc6 build perhaps? | 16:16.39 |
Robin_Watts | stars: There is some analysis code within the html output. | 16:16.45 |
| stars: but it's not comprehensive. | 16:16.58 |
Jogux | mace : vs2005, but with directx, platform sdk, some opengl emulation layer, and a couple of other bits and pieces :) | 16:17.04 |
henrys | Jogux: all this is covered in your email and I gave the okay. What do we need from Miles? | 16:17.11 |
chrisl | marcosw: I was getting those as apparent indeterminisms when I was testing the fuzzing fix I committed earlier | 16:17.26 |
Jogux | henrys : we need Miles to ask the administrators for the missing bits | 16:17.59 |
stars | Robin_Watts, is there a strategy to detect whether the line is starting with a special character or an indent? | 16:18.09 |
marcosw | chrisl: these haven't been indeterministic before. | 16:18.34 |
Robin_Watts | stars: The analysis code looks at that, yes. | 16:18.37 |
tor8 | Robin_Watts: best way to test your printf stuff? | 16:19.05 |
Jogux | henrys : if he can copy me in on that, I'm happy to try and assist / lead this till we get to the next stage | 16:19.13 |
chrisl | marcosw: Hmm, seems odd that mvrhel_laptop's commit would cause that.... but you never know! | 16:19.22 |
Robin_Watts | tor8: I've been testing it by using it as part of the pdfclean stuff. | 16:19.30 |
henrys | Jogux: I see I read more into the last sentence of your email, i.e. you knew the admins and knew what to do. | 16:19.43 |
Robin_Watts | but it's tested as part of a normal cluster run anyway because of pdf_update_appearance. | 16:19.53 |
marcosw | I need to run to uni, I'll look at the recent regressions and enter a bug when I figure out what caused them (and if they are indeterministic). | 16:20.18 |
mvrhel_laptop | marcosw: do you have a web page where I can do these compares? I looked at my bmpcmps but dont see some of these | 16:20.20 |
Jogux | henrys : ahh. I do know the admins. but I think Miles needs to make the first contact; I'm not in a position to know what Artifex has/hasn't bought, and/or to say "hey, this random guy is an official contact for artifex" | 16:20.42 |
marcosw | mvrhel_laptop: I'll put the files on my casper and email you the URL. | 16:20.47 |
tor8 | Robin_Watts: right. I've got a rewrite of the printf stuff that I want to test. | 16:20.51 |
mvrhel_laptop | my changes would only be cmyk devices | 16:21.05 |
| e.g. pkmraw | 16:21.16 |
Robin_Watts | tor8: Well, if it's hooked in to the fz_buffer output stuff, then a normal run will test it at least partially. | 16:21.26 |
pedro | Jogux, mace: I have smrtoffice up and running in 2012 on a windows7 target (as the most obvious, supported config which avoids retrofitting all the vs2005 stuff to local build machines), but I'm doing this as a route to having some easy way to get through the bug list | 16:21.27 |
mvrhel_laptop | and pam | 16:21.29 |
pedro | if we're wanting to set up the old vs2005/xp build env then that's a different issue really | 16:21.51 |
marcosw | mvrhel_laptop: than these probably aren't yours. they are all in mono devices (which I supppose could be called CMYK devices with only the K channel :-) | 16:21.55 |
mvrhel_laptop | right. those will not be effected by my change | 16:22.11 |
Robin_Watts | pedro: speaking personally, I'd like VS2005 windows 7 :) | 16:22.15 |
| or vs2008 or vs2010 | 16:22.29 |
| I don't have vs2012 | 16:22.35 |
mvrhel_laptop | robin_watts I gave you a copy I thought | 16:22.44 |
pedro | robin> I reckon that would work fine with my changes, just that I don't have 2005 | 16:22.46 |
Robin_Watts | mvrhel_laptop: You gave me 2010 pro. | 16:22.54 |
chrisl | marcosw: maybe these are related to ray_laptop's seg fault? | 16:22.56 |
Jogux | pedro : we have 2005 :-) | 16:23.01 |
marcosw | chrisl: possibly. I'll know more later today. | 16:23.20 |
Robin_Watts | mvhrel_laptop: It's possible that paulgardiner got 2012 from you, cos he needed to do some windows 8 stuff? | 16:23.40 |
marcosw | mvrhel_laptop: I've put the files on casper, but since they aren't your issue I won't bother sending you the URL :-) | 16:23.48 |
mvrhel_laptop | ok | 16:23.53 |
marcosw | I really need to run now, will have more information later today. | 16:24.17 |
mvrhel_laptop | ok. I am going to go work on gsview for a bit today | 16:24.34 |
pedro | I think there are way too many toolchain/platform variants at present in SOT - we need 2005 for some for legacy reasons (ie nobody actually updated the builds) , and 2012 is the only option on various other builds (win8/winphone8) | 16:25.19 |
henrys | Jogux: okay message sent to miles you should be copied, he should get to it ASAP | 16:27.09 |
tor8 | Robin_Watts: hm, so there's no "mutool clean" version of the pdf-op-filter stuff yet? | 16:27.19 |
Jogux | henrys : seen email. thanks. | 16:27.21 |
Robin_Watts | tor8: It's not hooked up into that yet, no. | 16:27.38 |
tor8 | rats :) | 16:27.43 |
Robin_Watts | but one of the later commits on robin/master is close. | 16:28.00 |
tor8 | how did you debug your print_f output then? | 16:28.10 |
Robin_Watts | See 'hacky test' on robin/master | 16:28.25 |
tor8 | Robin_Watts: yes, I've cherry-picked up to "Hacky test" | 16:28.33 |
Robin_Watts | Whenever I run a page, I first clean it. | 16:28.38 |
| So if I cluster test with that in, any mistakes in the fz_printf give bitmap diffs. | 16:28.55 |
tor8 | I can hook into there and dump the streams for now, just to see if it fails completely | 16:29.09 |
henrys | tkamppeter: so you haven't considered bringing the same mobile framework (cups restructuring and additions (bonjour etc.)) to android and iOS? | 16:32.36 |
stars | Robin_Watts, Thanks! | 16:34.38 |
Robin_Watts | stars: no problem. If you make any significant headway, let us know. We're always interested in improvements. | 16:41.03 |
henrys | tkamppeter: for our project we would want level 1 and 2 only | 16:42.34 |
ray_laptop | making progress into the segfault. Now have to set a data bp to see where something is getting changed after 'reset_landscape_buffer' | 16:43.37 |
stars | Robin_Watts, Sure, Its my pleasure too doing so! | 16:45.24 |
Robin_Watts | tor8: I might have the image decode on read thing working. 246 diffs, and I was expecting some. running a bmpcmp now. | 16:49.07 |
tor8 | Robin_Watts: I'm confusing valgrind badly :( | 16:50.19 |
| it can't even give me a proper stack trace... | 16:50.30 |
Robin_Watts | that's not good.. | 16:50.34 |
tor8 | stack smash! | 16:50.48 |
tkamppeter | henrys, no, in that moment I have only thought about Ubuntu Touch, but all this would apply as well for porting CUPS to Android and iOS. I have made it even possible to use only one of Poppler and GS, but if you opt for GS, some PS printer brands with interpreter bugs will not work. | 16:57.56 |
| henrys, I thing for the porting of CUPS one would mainly need to add an interface to the system's Bonjour/DNS-SD infrastructure, to not need to introduce avahi-daemon, perhaps also add an interface to what replaces D-Bus in the mobile OSes for notification. | 16:59.42 |
| henrys, PC-hardware/Linux-specific stuff is hidden away in the backends and that backends are level-3-only anyway. | 17:00.36 |
henrys | tkamppeter: what sort of timeline do you have on the restructure described in the url you gave us. | 17:01.44 |
| ? | 17:01.45 |
mvrhel_laptop | bbiab | 17:02.15 |
tkamppeter | henrys, most is already done, so part of Ubuntu Trusty (14.04 LTS). | 17:02.35 |
henrys | tkamppeter: off topic but kens and chrisl are aware of the ps bugs right? | 17:03.58 |
tor8 | Robin_Watts: okay, found the problem. the new printf seems to work, now for some more polishing. | 17:04.00 |
Robin_Watts | But it's IEEE only ? | 17:04.11 |
henrys | tkamppeter: so you already have bonjour added to the cups-browser? | 17:04.26 |
tor8 | Robin_Watts: nope, it's independent of floating point format | 17:04.29 |
Robin_Watts | oh, cool. | 17:04.35 |
| Does it do all the %1.2g type stuff? | 17:04.51 |
ray_laptop | tkamppeter: see the link I posted above about bonjour on Android | 17:05.07 |
tkamppeter | henrys, driverless printing on IPP Everywhere (PWG Raster), PDF, PostScript, PCL is done in current upstream cups-filters, the level 1/2/3 package splitting is already done in the Debian and Ubuntu packages of current Debian unstable/Ubuntu Trusty, also on-demand starting of daemons is done in cups-filters upstream, partially in CUPS upstream and in the Ubuntu and Debian packages, for PID 1 being systemd or Upstart, | 17:05.52 |
| henrys, yes, kens and chrisl know about the PS interpreter problem. | 17:06.20 |
tor8 | Robin_Watts: it only supports %f but prints the shortest string to losslessly convert | 17:06.27 |
| though we truncate very small and very big numbers | 17:06.37 |
henrys | tkamppeter: and this is all LGPL? Apple exclusion? | 17:06.42 |
Robin_Watts | We also need... %02x | 17:06.44 |
chrisl | henrys: there may be others that we've not seen, but some of those printer bugs are ones where the user wasn't willing/able to spend the time with us to debug the problem (or, one case, spectacularly failed to follow simple instructions.....) | 17:06.46 |
tkamppeter | henrys, cups-browsed always worked with Bonjour. | 17:06.53 |
Robin_Watts | and %03o | 17:06.57 |
tor8 | Robin_Watts: zero-padding quantifier is there and works | 17:06.58 |
| but no precision width quantifier | 17:07.10 |
| I'm going to try to tweak the dtoa to work with floats directly rather than doubles | 17:07.34 |
tkamppeter | henrys, the history is as follows: CUPS 1.5.x and older used CUPS' own broadcasting and browsing, all built-in in CUPS. CUPS 1.6.x and newer switched to Bonjour broadcasting but does not do Bonjour browsing, browsing was supposed to do by GUI dialogs, and as the GUI developers did not kick in quickly and the command line is not covered I created cups-browsed which does the Bonjour browsing in a daemon running in the background for the | 17:09.58 |
| user having exactly the same experience as with CUPS 1.5.x also in 1.6.x: On a client remote CUPS queues show automatically and use the driver on the server. | 17:09.59 |
ray_laptop | chrisl: I see what I messed up with the vertical stripes "fix". Working through a fix for that (has to do with the 'done' flag) | 17:10.57 |
tkamppeter | henrys, later on, Tim Waugh from Red Hat did a contribution to cups-browsed, optional support for the old CUPS broadcasting and browsing, for backward compatibility to old CUPS on remote machines, in both directions (server/client). | 17:11.22 |
chrisl | ray_laptop: thanks, that's good to hear. | 17:12.30 |
Jogux | marcosw : Miles says we have more picsel stuff than just the NAS? | 17:12.35 |
henrys | Jogux: no marcosw just set up the NAS in miles office as far as I know. marcosw? | 17:13.55 |
tkamppeter | henrys, libcups... is all LGPL-2 with Apple exception, the non-lib part of GPL-2 with Apple exceptions. cups-filters has several different licenses because it is pulled together from several projects. Most is LGPL-2 and GPL-2 inherited from CUPS and there is also some MIT in independent parts. | 17:14.23 |
Jogux | henrys , marcosw : he seems to specifically says we have "servers" (plural) | 17:14.56 |
henrys | tkamppeter: we are happy to abide by LGPL but we can't do GPL for this project. | 17:15.19 |
Robin_Watts | henrys: Can we not? | 17:16.14 |
Jogux | GPL on the iOS appstore is... tricky. | 17:16.36 |
Robin_Watts | Ah, for ios, yes. | 17:16.46 |
tkamppeter | henrys, perhaps someone must look more closely what exactly is which license and whether it is viable to do your project. | 17:16.46 |
Robin_Watts | for android, no problem. | 17:16.50 |
henrys | Robin_Watts: we can't link it to SOT | 17:17.26 |
Robin_Watts | For android it would be no problem, AIUI. | 17:17.44 |
chrisl | henrys: we wouldn't need to link to SOT | 17:17.49 |
Robin_Watts | We produce an apk with SOT in. | 17:18.03 |
| That would be a libsot.so + some java stuff + a libcups.so | 17:18.29 |
| Anyone can then unpack the apk, replace the libcups.so and repack. | 17:18.54 |
Jogux | Robin_Watts : that's LGPL you're thinking of there? | 17:19.13 |
tkamppeter | Why CUPS on iOS? To replace their AirPrint which prints only on selcted printers directly and on all printers shared by Ubuntu servers by CUPS which prints on printers on all CUPS servers (including Mac OS X), on all IPP Everywhere printers and also on nearly all IPP PS, PDF, and PCL printers? | 17:19.19 |
Robin_Watts | They can't re-sign it as coming from us, but our obligations under the LGPL are fine. | 17:19.25 |
chrisl | Robin_Watts: didn't you and Paul work out a way to bundle separate executables in an apk? | 17:19.33 |
Robin_Watts | Jogux: Yes, that would be the LGPL case. | 17:19.33 |
| For GPL we'd need to produce 2 apps. | 17:19.49 |
chrisl | Yes, a simple "stub" exe that hands off a PDF to the CUPS machinery | 17:20.15 |
Robin_Watts | 2 apks, sorry. | 17:20.30 |
| chrisl: Yes, we can bundle an exe in with a apk and call it, but it's hacky. | 17:20.52 |
tkamppeter | henrys ^^ | 17:21.06 |
chrisl | Robin_Watts: we often tell people to use Ghostscript that way!! | 17:21.18 |
Robin_Watts | We'd have a CUPS apk and an SOT one. The CUPS one can be entirely GPL. | 17:21.23 |
| chrisl: No, I meant our way of packing it is hacky. | 17:21.33 |
chrisl | Oh, right. | 17:21.48 |
henrys | Robin_Watts: I don't think OEM's would like that packaging. | 17:22.16 |
Robin_Watts | henrys: True enough. | 17:22.33 |
chrisl | If CUPS because the de facto standard printing system on Android, though...... | 17:23.04 |
henrys | tkamppeter: we've purchased a mobile office app and need a printing module, ironic isn't it? | 17:23.14 |
| chrisl: that's what I was thinking this could be a really cool thing for us to do. Port all the cups stuff to android and iOS and see if it takes. At the same time we solve our printing problems. | 17:24.05 |
tkamppeter | henrys, nice, using what the OS provides is not viable? Not powerful enough? Range of printers to small? | 17:24.10 |
Robin_Watts | tkamppeter: At present android doesn't have a printing API in the OS. | 17:24.36 |
tkamppeter | Robin_Watts, not even 4.3 or newer? Did they not add something like that? | 17:25.09 |
henrys | tkamppeter: right legacy printers and such. | 17:25.20 |
ray_laptop | I thought I saw that it will be added, but not yet. | 17:25.46 |
Jogux | there's a new printing API thing in Android 4.4 | 17:26.16 |
henrys | yes 4.4 | 17:26.23 |
| we'll see how mature that is though and how much they want you to use google cloud printing etc. | 17:27.06 |
| will in print an add on the last page + 1? ;-) | 17:28.10 |
ray_laptop | What I see is Android 4.4 introduces native platform support for printing, along with APIs for managing printing and adding new types of printer support. | 17:28.16 |
henrys | s/in/it | 17:28.18 |
tkamppeter | henrys, I have checked on my Nexus 7 (2012) with Android 4.4 and there I can print out of Chrome. It sees the printers in the network. | 17:28.30 |
ray_laptop | but this may only be Cloud printing, not standalone printing to local printer | 17:29.03 |
| tkamppeter: can you print if you disconnect your wi-fi access point from the net (no cloud avail) ? | 17:29.54 |
henrys | tkamppeter: then maybe that route is best. | 17:30.00 |
| ugh lots of snow here. | 17:30.59 |
ray_laptop | henrys: my kids would love it | 17:31.15 |
tkamppeter | henrys, it seems rather restricted, so that CUPS even makes sense, as there are two kinds of "printer drivers", one for all HP and one for Cloud print. In stead of introducing CUPS we could perhaps make an additional "printer driver" app which is based on cups-filters and supports all IPP Everywhere, PS, PDF, and PCL printers via IPP. | 17:31.26 |
ray_laptop | (nice to visit snow, but I don't want to live there) | 17:31.32 |
| tkamppeter: that makes sense to me. | 17:32.10 |
| and for old Android devices (pre 4.4) they need to get one of the 3rd party apps | 17:32.37 |
| have to change venues. bbiab | 17:33.06 |
henrys | tkamppeter: can we do that with lgpl only? | 17:34.15 |
tkamppeter | ray_laptop, henrys, I know that the HP "printer driver" is an app (which was included from 4.4 on), the name of the app is "HP Print Service plugin". | 17:34.18 |
kens | goodnight all | 17:34.38 |
henrys | tkamppeter: I thought googles plan was to put out the framework then printer folks submit apk's that configure their devices for printing. | 17:35.08 |
tkamppeter | henrys, I think not, as as the filters originally resided in CUPS they were GPL-2. | 17:35.20 |
henrys | tkamppeter: how does apple use the filters? | 17:36.32 |
tkamppeter | henrys, we must see which filters exactly belong to level 2, but I know that one of them is pstops to support PS printers (to insert option PS code) and this still is in CUPS upstream. | 17:36.37 |
| henrys, Apple does not use cups-filters at all, they have their own proprietary CUPS filters. | 17:37.12 |
henrys | tkamppeter: okay right | 17:37.23 |
tor8 | Robin_Watts: that va_copy ifdef stuff for MSVC, any way we can get rid of it? | 17:38.06 |
henrys | of course we can do pstops with gs | 17:38.24 |
tkamppeter | henrys, pstops can pehaps be replaced by the recent advances in the ps2write device of GS, it has some facility of inserting option PS code now. | 17:38.27 |
henrys | now to enlist Robin_Watts and paulgardiner in this insanity ;-) | 17:39.52 |
Robin_Watts | tor8: Nope. | 17:43.26 |
| va_copy is actually standard C, I think. | 17:43.49 |
| It bit me horribly before. | 17:44.02 |
tor8 | Robin_Watts: anyway, 3 commits on tor/master for your perusal | 17:44.22 |
| I haven't tried compiling on win32 yet, I need to add the dtoa.c to the project file and make sure the frexp() and strtof() functions exist | 17:45.13 |
henrys | torn, Robin_Watts : this would give us a nice print module for mupdf as well. | 17:45.39 |
| s/torn/tor8 - colloquy always makes that correction | 17:45.59 |
tor8 | henrys: maybe I should change nick to torn :) | 17:46.18 |
Robin_Watts | tor8: 5 mins, and then I'll have a commit for you to look at too :) | 17:46.31 |
tor8 | henrys: this is a print module for mupdf use | 17:46.34 |
| henrys: oh, wait, nvm, I completely misunderstood | 17:46.51 |
| the dangers of concurrent irc threads | 17:47.07 |
henrys | tor8: we are talking about porting parts of cups to android and iOS. See the link above tkamppeter provided about the new "mobile" cups stuff. | 17:47.35 |
| it would be lgpl. | 17:47.47 |
tor8 | Robin_Watts: dinner. be back later. | 17:48.49 |
tkamppeter | henrys, perhasp you have a look into the COPYING file of cups-filters and we can find suitable filter paths. | 17:48.55 |
henrys | tkamppeter: I have 1.7.1 source, that okay? | 17:49.53 |
tkamppeter | henrys, pdftopdf and gstoraster are MIT. | 17:50.13 |
henrys | tkamppeter: oh the filters aren't here they've been pulled out as you said duh⦠hang on let me fetch them. | 17:51.09 |
tkamppeter | henrys, I am talking about cups-filters, current is 1.0.48, I released this version some days ago. CUPS is 1.7.1 currently, that is correct. | 17:52.18 |
henrys | okay 1.0.48.orig.tar.xz | 17:53.14 |
tkamppeter | henrys, yes. | 17:53.47 |
henrys | why is gstoraster kit? | 17:56.18 |
| s/kit/mit | 17:56.23 |
Robin_Watts | tor8: I am confused. | 17:59.21 |
| tor8: fz_ftoa can still produce numbers like 23e-10. (See fmtexp) | 17:59.47 |
| xfmtexp even. | 17:59.57 |
| Which means the fz_vsnprintf can too. | 18:00.22 |
| Also, I dislike changing what %c means in fz_vsnprintf. Can't we use for UTF8? Likewise, can we have %S for utf8 strings? | 18:01.33 |
tkamppeter | henrys, I must look? | 18:02.27 |
| s/?/./ | 18:02.35 |
Robin_Watts | The third one looks good though. | 18:02.37 |
henrys | tkamppeter: so gstoraster uses ghostscript as a gpl library but the filter itself is an MIT licensed program? | 18:05.05 |
tkamppeter | henrys, its header comments contain the MIT and I did not replace the header comments when moving gstoraster over from GS. For me it seems but I do not remember exactly that I have introduced the filter to GS and I did not write it from scratch but used some other filter, probably from the oprnprinting.jp efforts on getting filters for the PDF workflow and they made all MIT. | 18:05.21 |
| henrys, gstoraster calls GS vis command line, it does not link against libgs. | 18:06.02 |
henrys | tkamppeter: well this does give us a path to consider. | 18:06.38 |
Robin_Watts | Jogu, henrys, marcosw: Any objection to me running a recursive chmod to make everything on /mnt/picsel on points readable to everyone? | 18:07.58 |
tkamppeter | henrys, PDF, and IPP Everywhere (PWG Raster) we could support without any problems, PCl_XL is also no problem, for PCL 5c/e we would need to change the path as rastertopclx is CUPS-inherited. | 18:08.48 |
Jogux | robin_watts : fine by me. I didn't spot anything not accessible apart from the sage accounts :-) | 18:08.59 |
tkamppeter | henrys, other problem is PostScript, as pdftops is also CUPS-inherited. | 18:09.11 |
Robin_Watts | tkamppeter: We don't care about postscript. :) | 18:10.00 |
| For our purposes, we basically want raster -> printer language filters + printer discovery/communication. | 18:10.54 |
tkamppeter | Robin_Watts, you mean not supporting PostScript printers? Many legacy printers would stay unsupported then? Note that I do not mean PS input, as input I only consider PDF. | 18:11.04 |
Robin_Watts | Fair enough. | 18:11.37 |
| pdftops requires poppler, currently? | 18:11.56 |
henrys | tkamppeter: it can be configured with ghostscript no? | 18:12.02 |
tkamppeter | Robin_Watts, no, pdftops is configurable: GS, Poppler, acroread, pdftocairo, and GS/Poppler hybrid. | 18:12.40 |
henrys | licensing messes us up though we can't use ghostscript in an lgpl environment. we could shell to it. | 18:13.45 |
tkamppeter | henrys, Robin_Watts, the mobile level 2 stack is currently possible Poppler-only and also GS-only, no necessity to have both Poppler and GS on a mobile device. | 18:14.06 |
| henrys, Robin_Watts, adding MuPDF support to cups-filters I have opened as GSoC project, but no students have shown up yet, volunteers welcome. | 18:15.01 |
| henrys, AFAIK cups-filters always shells to GS, never uses libgs. | 18:16.08 |
henrys | tkamppeter: any idea how portable all the filters are? How much have you done with ubuntu mobile platforms? | 18:21.16 |
| Jogux: miles is contacting the liquidators to see if there is anything else. | 18:27.26 |
| lunch | 18:39.22 |
tkamppeter | henrys, I did not explicitly test on ARM, but all Debian and Ubuntu packages are also built on ARM by the build servers, and I also tested a full Ubuntu desktop on the Nexus 7 already and the printing stack (that time level 3) worked perfectly. | 18:42.25 |
ray_laptop | mvrhel_laptop: In the original gx_ht_thresh_planes, before my change that fixed the vertical stripes (and cause the segfault), the 'allow_reset' flag after the j < spp_out loop would always be true, so the if (!allow_reset) would not get done, so 'done' would never be set true. Or am I misreading the code ? | 18:52.53 |
mvrhel_laptop | ray_laptop: hold on need to do a context swap | 18:53.22 |
| let me jump back to gs | 18:53.34 |
ray_laptop | mvrhel_laptop: it looks like I can just get rid of the 'done' flag (which is what I am testing now) | 18:53.41 |
tor8 | Robin_Watts: no, it can't. xfmtexp is only used for the round-trip test that strtof results in the original number. | 18:54.05 |
ray_laptop | mvrhel_laptop: do you want the version of the code before I fixed the vertical stripes ? | 18:54.11 |
mvrhel_laptop | ray_laptop: what ever you want/need me to look at | 18:54.31 |
| I am just now looking at the file | 18:54.45 |
| so this will take a minute | 18:54.52 |
ray_laptop | mvrhel_laptop: the gxht_thresh.c.old is now on casper:~ray | 18:55.00 |
tor8 | Robin_Watts: do you mean %S for utf-16 strings? | 18:55.01 |
| we don't use wchar strings anywhere in mupdf (but sumatrapdf does) | 18:55.25 |
mvrhel_laptop | ray_laptop: which file do you want me to look at | 18:55.26 |
| not sure what question is | 18:55.31 |
ray_laptop | mvrhel_laptop: the old file, please. Please confirm my (static) analysis that the done flag would never get set | 18:56.02 |
mvrhel_laptop | ok hold on then | 18:56.14 |
ray_laptop | done never got set because after the j<spp_out loop, allow_reset would always be true (I don't see any way the loop would exit early -- no if (...) break;) | 18:57.30 |
| mvrhel_laptop: BTW, I am looking at the image_landscape case | 18:58.12 |
mvrhel_laptop | yes. I see where all the done values are. hold on a sec | 18:58.26 |
| it is tied to allow_reset | 18:59.17 |
| which is tied to j == spp_out - 1 | 18:59.35 |
Robin_Watts | tor8: We don't use it. yet. | 18:59.53 |
mvrhel_laptop | or rather j != spp_out - 1 | 19:00.03 |
| blah | 19:00.12 |
Robin_Watts | tor8: Let me have another look at the code then. | 19:00.47 |
| tor8: The "fix operator buffering of inline images" is ready to go now I think. | 19:01.17 |
ray_laptop | chrisl_away: my patch cured the segfaults in my gs regression test. | 19:01.19 |
tor8 | Robin_Watts: I have a slight cleanup on the way, if you can wait two minutes. | 19:01.24 |
mvrhel_laptop | ok so allow_reset means that we have done our last plane if I am reading this correctly | 19:01.24 |
Robin_Watts | tor8: sure. | 19:01.39 |
mvrhel_laptop | and if we are not in the last plane then we are setting done = true | 19:02.12 |
tor8 | Robin_Watts: do we want "0.1" or ".1"? | 19:02.26 |
mvrhel_laptop | don't like the terminology I used there | 19:02.28 |
ray_laptop | mvrhel_laptop: right, and the check (that included the done = true) was AFTER the loop | 19:02.33 |
Robin_Watts | tor8: Clipping f to a maximum of 1e10 seems wrong. | 19:02.36 |
| tor8: .1 for preference. | 19:02.43 |
mvrhel_laptop | no it is in the while loop | 19:03.02 |
| ray_laptop | 19:03.08 |
ray_laptop | in the while loop, but NOT the j loop | 19:03.23 |
| the allow_reset is set according to the j loop | 19:03.34 |
tor8 | Robin_Watts: yeah, I'm not sure what we should do though | 19:03.47 |
ray_laptop | j is the loop for each plane | 19:03.48 |
tor8 | we could print with leading/trailing zeroes in all cases | 19:03.56 |
| but then we might end up with 38+9 digits | 19:04.09 |
mvrhel_laptop | I see allow reset will always be false there | 19:04.51 |
tor8 | or maybe it's only 38 digits max, the FLT_MAX value is I believe something * 10^38 | 19:04.57 |
| and at most 9 significant digits | 19:05.06 |
mvrhel_laptop | so done will always be true their | 19:05.12 |
| there | 19:05.14 |
tor8 | Robin_Watts: what should we do for Infinity? | 19:05.24 |
ray_laptop | allow_reset will always be TRUE. The last time frhough the for loop on j, j will be == spp_out-1 | 19:05.24 |
mvrhel_laptop | ok hold on let me step myself through it one more time | 19:05.47 |
| right allow_reset will always be true | 19:06.18 |
| meaning done will never get set | 19:06.26 |
| and so is doing nothing | 19:06.36 |
Robin_Watts | tor8: What is the maximum representable value of a single precision IEEE float? | 19:06.52 |
ray_laptop | mvrhel_laptop: OK. so that is what I was thinking as well. | 19:06.53 |
mvrhel_laptop | ray_laptop: thank you for finding that (and holding my hand through it) | 19:06.54 |
| an embarrassing piece of code there | 19:07.20 |
ray_laptop | mvrhel_laptop: after you look at that, please look at the version I pushed to my repo -- it removes the setting of done and keeps the move_landscape for each plane | 19:07.28 |
mvrhel_laptop | ok let me look at that now | 19:07.40 |
tor8 | FLT_MAX=3.4028234663852886e+38 | 19:08.22 |
ray_laptop | I also discovered when stepping through that sometimes we were moving 'in place' because one image line in gave us more than 1 strip out | 19:08.36 |
| since data_length tends to be large, I added the if (postiion_curr != poistion_new) | 19:09.27 |
mvrhel_laptop | I see that | 19:09.33 |
Jogux | henrys: I've discovered more partitions on the NAS. working with Robin to investigate what's on them. | 19:09.45 |
ray_laptop | mvrhel_laptop: the segfaults were caused by exiting the whle loop before enough data was processed, and we ended up getting negative array indices. My bad | 19:11.28 |
tor8 | Robin_Watts: okay, cleaned up commit pushed | 19:12.58 |
| still need to decide what to do with huge (and tiny) numbers though | 19:13.22 |
Robin_Watts | marcosw: You awake? | 19:13.58 |
mvrhel_laptop | the landscape case was messy. ray_laptop: it looks good to me. I see the move_landscape_buffer is now in the loop on the plane for the left over stuff and it is nice to have done removed | 19:14.02 |
ray_laptop | mvrhel_laptop: OK. Thanks. I'll add the comment about the move_landscape optimization and push it. | 19:14.46 |
Robin_Watts | tor8: So can we print infinity as 3.4e+38 ? | 19:14.50 |
tor8 | Robin_Watts: hm, something is wrong with ftoa when using floats :( | 19:14.50 |
mvrhel_laptop | ray_laptop: great | 19:14.59 |
| I have to head out for bit for lunch | 19:15.05 |
Robin_Watts | tor8: IIRC C promotes everything to a double when pushing it through a var_args thing. | 19:15.22 |
| likewise chars become ints. | 19:15.28 |
tor8 | Robin_Watts: you mean "340282346638528859811704183484516925440" | 19:16.03 |
Robin_Watts | tor8: I do. | 19:16.31 |
ray_laptop | mvrhel_laptop: OK. I'll chat later about the t_level_adjust code. | 19:16.32 |
tor8 | Robin_Watts: yes, we could do that | 19:16.42 |
Robin_Watts | Certainly, I don't think we should clip at +/- 1e10 That seems overly small. | 19:17.02 |
Jogux | does a little dance. we found the picsel VM images! | 19:22.45 |
tor8 | Robin_Watts: yeah, mixing floats and doubles makes for big trouble... :( | 19:28.37 |
| "0.3" parsed as a float and back to string with full precision is 0.29999999999999999 | 19:29.39 |
| cast that to a float and back to double and you get 0.30000001192092896 | 19:29.51 |
| and then converting that back to a string won't let us end up with "0.3" anymore (since it's not losslessly converted back to 0.30000001192092896) | 19:31.07 |
| I'll have to fiddle some more with this... | 19:31.14 |
Robin_Watts | tor8: I think the "Fix operator buffering..." is good to go. | 19:47.08 |
| tor8: I am going to hook the cleaning stuff into pdfclean now. | 19:47.25 |
tor8 | Robin_Watts: did you do anything about decoding an image twice? | 19:48.43 |
Robin_Watts | tor8: yeah, that's fixed in the latest commit. | 19:48.55 |
| we decode the image at full size while we're pulling in the data. | 19:49.10 |
| If we ever get an inline image that's so large we run out of memory while decoding it, we can worry about that then. | 19:49.38 |
tor8 | the image->tile = decomp_image_from_stream call? | 19:49.38 |
Robin_Watts | in pdf_load_compressed_inline_image yes | 19:50.30 |
tor8 | Robin_Watts: right. looks good to go to me. | 19:50.38 |
Robin_Watts | Ta. | 19:50.42 |
tor8 | I think I'm just about ready to give up and hack the ftoa stuff using printf("%e"), strtod() and then parsing the digits and exponent out of the buffer, and reformatting as in the code you've already seen | 19:51.21 |
| basically, try %.1e, then %.2e etc until we get the same float back | 19:51.52 |
| much slower, but consistently so at least (at most 10 printf's per number we print) | 19:52.31 |
Robin_Watts | tor8: Presumably we can binary search? | 19:53.10 |
tor8 | Robin_Watts: we don't have a good stopping test (or I'm just denser than usual) | 19:53.48 |
Robin_Watts | tor8: Print to the maximum number of figures. convert back. | 19:54.20 |
tor8 | if we get a good result in the middle, we still need to try the previous numbers to see if maybe they're also good | 19:54.21 |
Robin_Watts | tor8: RIght, we need to search until the upper and lower bounds are 1 apart. | 19:55.10 |
| but if our search space is somewhere between 0 and 9 decimal places, worst case we have to test 4. | 19:55.53 |
| test 9, test 5. If the same, we know the success point is between 0 and 5, if not within 6 and 9. | 19:57.58 |
| etc | 19:58.00 |
henrys | marcosw, Robin_Watts can we have email notifications for epage git? | 19:59.38 |
Robin_Watts | I'm sure it can be done. I think possibly it was tor8 that did it for gs etc? | 20:00.18 |
tkamppeter | henrys, have you seen my last message? | 20:01.24 |
henrys | tkamppeter: wow! | 20:02.15 |
Jogux | tkamppeter> awesome! | 20:03.27 |
| henrys / Robin_Watts / marcosw : much as I hate to be the "I'm IMing you to say I've sent you an email" guy, I just send you an email about possible next steps for getting ATS up :-) | 20:04.17 |
| is there a gitweb / gitlab / whatever for the epage git? | 20:05.24 |
Robin_Watts | Jogux: No. | 20:05.41 |
Jogux | aw. | 20:05.59 |
Robin_Watts | git.ghostscript.com shows the public things. | 20:06.02 |
henrys | tkamppeter: with printer discovery and all? | 20:06.34 |
Robin_Watts | We could probably have private-git.ghostscript.com that needed some sort of authentication ? | 20:06.35 |
Jogux | vaguely throws it out there that a gitlab setup would be awesome and we'll wonder how we ever lived without it. :-) | 20:07.00 |
Robin_Watts | Jogux: I would *love* a gitweb view on it. | 20:07.01 |
Jogux | robin> if you love gitweb you'll really like gitlab's webview, ime :-) | 20:07.18 |
Robin_Watts | looks like github. | 20:08.40 |
| I hate github :) | 20:08.45 |
Jogux | yeah, it does a bit. which part of github do you hate? | 20:09.06 |
tkamppeter | henrys, yes, it all worked, even a USB multi-function device with HPLIP, Plug'n'Print (and also Plug'Å'Scan), but a USB printer and HPLIP we have only in level 3. Level 2 (network printers) also worked perfectly with it. | 20:09.18 |
Robin_Watts | just all of it :) | 20:09.26 |
| but I may be being unfair. | 20:09.40 |
henrys | tkamppeter: why not do a release and let folks use it? | 20:10.28 |
tkamppeter | henrys, what do you mean with a release? | 20:11.50 |
henrys | tkamppeter: well a source release with android build instructions for now. | 20:12.32 |
Jogux | robin_watts : the bit I particularly like is the commit/merge tree view - https://gitlab.com/gitlab-org/gitlab-ce/network/master [takes a while to display, and they have a fairly hideous number of merges going on!] | 20:12.35 |
Robin_Watts | Jogux: Ah, right. We have a policy of git pull --rebase | 20:13.03 |
henrys | Jogux: oh god don't start Robin_Watts on merging. | 20:13.22 |
Jogux | henrys : d'oh. sorry. | 20:13.33 |
henrys | ;-) | 20:13.37 |
Robin_Watts | i.e. we never have merges in our tree, unless we have done a long development branch. | 20:13.43 |
Jogux | robin_watts : oo, awesome. although you still have release branch merges presumably | 20:13.58 |
| Paul pointed out we have a rather horrid set of release branches to make sense off in the epage git :-S | 20:14.22 |
Robin_Watts | AIUI we make a release branch when chrisl is ready to release. | 20:14.36 |
| We might then cherry pick fixes from the trunk or merge the trunk into that. | 20:14.58 |
| We never merge release branch back to trunk. | 20:15.04 |
| Hence our trunk stays with a (pretty much) linear history. | 20:15.20 |
Jogux | interesting. I still have much to learn about how you guys do things :-) | 20:15.24 |
Robin_Watts | Partly it's because it's neater, and partly is because it's much less confusing for git luddites. | 20:16.16 |
tkamppeter | henrys, unfortunately, all ARM tests which I described were with the full desktop of Ubuntu natively as OS on the Nexus 7 or the Pandaboard, not under Android. I by myself do not have any experience in developing under Android or iOS. | 20:16.25 |
Robin_Watts | not that we have any of those in the company. | 20:16.28 |
| I have thoroughly brainwashed everyone now I think. | 20:17.01 |
Jogux | gitlab can be great for reviews too - ability to have conversations focussed on a line of code, etc. I presume henrys will now warn me about mentioning reviews too. :) | 20:17.14 |
henrys | tkamppeter: oh this was all done on top of android | 20:17.15 |
| tkamppeter: oh I thought this was all done... | 20:17.40 |
tkamppeter | henrys, no, the devices have natively booted Ubuntu, with no other OS underneath. | 20:17.48 |
henrys | tkamppeter: but at least we know we shouldn't run into arm incompatibles but I really wouldn't expect that anyway. | 20:18.42 |
Robin_Watts | Jogux: gitweb is really useful for reviews. We get people to push to their own repos, and we can look at the commits there. | 20:18.42 |
| Jogux: alias.logg=log --graph --oneline --decorate --all | 20:19.06 |
| Put that in your git config, then do: git logg HEAD~20 | 20:19.24 |
| and you get a text version of that merge thing with nice ascii art. | 20:19.44 |
| and colors, on an appropriate terminal. | 20:19.54 |
Jogux | robin_watts: unless gitweb has got much better recently, gitlab excels for viewing. | 20:19.57 |
tkamppeter | henrys, what is all done (all on source code level) is the dividing into the levels, the support for running the daemons on-demand, the driver-less printing, the network printer discovery and auto-setup, and whatever I have written in the Blueprint. | 20:20.06 |
Jogux | robin_watts> yeah, but I can't clip through to see the commits :-) | 20:20.24 |
Robin_Watts | Jogux: I wouldn't fight a gitlab install. I just wouldn't fight for it :) | 20:20.46 |
tkamppeter | henrys, what is missing is to port this to the Android and iOS operating systems. The architecture is there, the prting still missing. | 20:20.49 |
henrys | Jogux: any clues where I might look to find embedded general docs? Like what printers are supported? | 20:21.14 |
Robin_Watts | henrys: grep the 550Gig of mail? :) | 20:21.27 |
tkamppeter | henrys, main work on the porting is moving over to the Bonjout/DNS-SD environments of the mobile OSes. | 20:21.33 |
Jogux | tkamppeter : so I guess essentially all the platform specific stuff (bluetooth, bonjour, etc) needs rewriting to use the native android features (possibly via ndk), plus all the android packaging work, and probably quite a lot of android ui. | 20:21.34 |
| henrys : like robin says, I guess the 550G of mail. however it's not currently in an easily readable form :-( | 20:21.59 |
tkamppeter | Jogux, yes. | 20:22.02 |
henrys | Jogux: I could disassemble the code in less time ;-) | 20:23.09 |
Jogux | henrys : if marcosw can do something with the ats vms, I guess he can also get the mail one up; if I can ssh into it I might be able to find something within 10-15 minutes if the setup is similar to the old picsel mailserver setup. | 20:23.19 |
henrys | Jogux: I am surprised picsel wasn't give some documentation - you don't have the source and you paid a royalty | 20:24.13 |
Jogux | henrys : I'm guessing there were, it just didn't get stored anywhere we can easily get at it. | 20:24.33 |
| henrys : what is it you're hoping to get from the documentation? | 20:24.47 |
henrys | Jogux: a list of printers supported. | 20:25.10 |
Jogux | ahh. yes. that would be nice, wouldn't it. | 20:25.19 |
| henrys : the issue at the moment is really that all this data is on the wrong end of Miles's internet connection, and points has nowhere near enough disc space to be able to do anything with the data. | 20:27.51 |
| are there any other linux servers in mile's office that might have more space? | 20:28.45 |
Robin_Watts | We really want to copy the VMs onto one or more new large discs that we can install in one of the DELL uberservers. | 20:28.50 |
| Jogux: The DELL servers are too loud for Miles' office. | 20:29.02 |
| They are in Marcosw's garage. | 20:29.18 |
henrys | Jogux: what's the total amount of data? | 20:29.44 |
Robin_Watts | The DELL servers can take 2 HDs per node, and they currently have 1 HD per node. | 20:29.59 |
Jogux | henry> 200G of ATS; 550G of mail; few hundred gig of other things we should possibly look into at some point. | 20:30.29 |
Robin_Watts | IIRC they have a 250GB drive per node currently (4 nodes per server) | 20:30.32 |
henrys | Jogux: we roman at google - so 9.99/TB now, does that work? | 20:31.09 |
Robin_Watts | I don't know if each node can easily access the HDs on the other nodes. Presumably they can via network, but that's not ideal. | 20:31.13 |
henrys | s/roman/domain | 20:31.17 |
| per month | 20:31.46 |
Jogux | henrys : not sure. we still have to get the data to google somehow. | 20:31.49 |
Robin_Watts | henrys: We need the VMs unpacked on a server. | 20:32.01 |
Jogux | and we can't easily host VMs from ... yeah, what robin said | 20:32.09 |
henrys | Robin_Watts: oh I thought we needed space. | 20:32.29 |
| only | 20:32.33 |
Jogux | both :-) | 20:32.42 |
| ATS is disc and CPU hungry. | 20:32.47 |
Robin_Watts | henrys: No, we need to be able to run the VMs, and we need to have the data local to those VMs. | 20:33.00 |
| My 4TB HD cost me 115 quid a while ago. | 20:33.30 |
henrys | Robin_Watts: sounds like you need a fast connection too. | 20:33.31 |
| which pushes us to aws | 20:33.45 |
Jogux | henrys : only need a fast connection to get the data onto the VMs. | 20:33.50 |
Robin_Watts | henrys: ATS has been built on the assumption of having a fast connection between nodes. | 20:34.04 |
Jogux | once the VMs are setup, the connection speed requirements aren't too bad. | 20:34.08 |
Robin_Watts | which means that if we move to a server in marcosw's garage, the nodes are all on the same local net, and we're fine. | 20:34.40 |
henrys | jogux:all of our nodes are fairly asymmetric - fast down slow up. | 20:35.00 |
Robin_Watts | We just need a few large HDs put in one of Marcosw's servers, and for them to be preseeded with a copy of the data from the NAS. | 20:35.18 |
Jogux | agrees with what Robin said; that sounds ideal. | 20:36.04 |
Robin_Watts | http://www.amazon.com/Seagate-Desktop-3-5-Inch-Internal-ST4000DM000/dp/B00B99JU4S/ref=sr_1_1?ie=UTF8&qid=1395174961&sr=8-1&keywords=4TB+HD | 20:36.13 |
henrys | Robin_Watts: I was trying to keep the company from folding when marcosw leaves but I guess that can't be avoided ;-) | 20:36.20 |
Robin_Watts | 4 of those would set us back just over $600, and would be ample. | 20:36.37 |
| henrys: marcosw will never leave. He will never be allowed to leave. | 20:36.51 |
Jogux | robin_watts : hrm, is the server SATA? | 20:36.57 |
Robin_Watts | henrys: Well, the server, once set up, could live anywhere. | 20:37.37 |
| Jogu: Checking. | 20:37.58 |
tor8 | Robin_Watts: okay, new version on tor/master that should be 100% lossless for floats. | 20:38.11 |
henrys | Robin_Watts: are they quiet enough for Miles office? | 20:38.34 |
Robin_Watts | http://www.theserverstore.com/content/dell-poweredge-c6100-servers | 20:38.36 |
| henrys: Nooooo. | 20:38.40 |
tor8 | but we may get 50 digit floats out of it... | 20:38.40 |
Jogux | anyway, if there's 250G per node, that sounds workable for now, anyway, assuming there's a few nodes with large amount free :-) | 20:38.44 |
tor8 | Robin_Watts: and if you can figure out the binary search loop, be my guest. my brain is too fried right now. | 20:39.08 |
Jogux | robin : does sound like it's SATA. awesome. | 20:39.18 |
Robin_Watts | Jogux: I suspect each node has a checkout of our testfiles on, which is 25Gig (possibly twice cos of SVN) | 20:39.19 |
Jogux | robin_watts : I think we can work with that. | 20:39.39 |
Robin_Watts | We really need marcosw to weigh in here. | 20:39.56 |
Jogux | the biggest VM is 100G; the others average 10G-ish. | 20:40.02 |
henrys | Robin_Watts: so they really can't go anywhere other than Marcos' garage. | 20:40.17 |
| unless we go outside the company | 20:40.32 |
Robin_Watts | henrys: Do you have a garage? or does michael? :) | 20:40.46 |
henrys | Robin_Watts: actually power price wise marcos is the worst choice | 20:41.40 |
Robin_Watts | If we can find a self-storage place near one of us that offers power, and we can have cable put into, we'd be golden :) | 20:41.51 |
henrys | well maybe you guys in the uk would be not sure. | 20:41.58 |
Jogux | hehe | 20:42.09 |
| we pay 10p-ish per kWh here. | 20:42.20 |
| which is about 17 cents USD | 20:42.39 |
Robin_Watts | Jogu: higher than that I think. | 20:42.42 |
| And it'll be much higher when Putin turns the gas off. | 20:43.07 |
| He has a new extension to heat, apparently. | 20:43.20 |
Jogux | robin_waitts: 10.207p on the last bill I can find :-) | 20:43.30 |
| admittedly I think it's gone up since then | 20:43.51 |
Robin_Watts | 16.11p day rate, 5.86p night rate for me. | 20:44.07 |
henrys | Jogux: california is about the same - almost half that in Oregon and Washington, sounds like mvrhel_laptop is our man ;-) | 20:45.00 |
| hydro !!! | 20:45.18 |
| Jogux: should this all stay together - it seems the email could be put somewhere else. We don't need fast access to that. | 20:46.30 |
Jogux | henrys : yeah, the email could go anywhere. | 20:47.21 |
| could shove an extra disc in points perhaps? | 20:47.46 |
Robin_Watts | The person who would shove an extra disc in any of these servers is marcosw. | 20:48.22 |
henrys | well we'll wait for marcosw I'll send him email and tell him to look at the logs. | 20:48.27 |
Jogux | robin_waitts: seems I'm on 10.925p/unit now (I'm on a single rate) | 20:48.32 |
Robin_Watts | Jogux: I should change to a single rate. | 20:48.47 |
Jogux | henrys : fwiw, the picsel web pages say it supports: | 20:51.25 |
| PCL6, PCL5 and PCL3GUI, three common versions of the industry standard printer language from Hewlett Packard. | 20:51.26 |
| Epson ESC/P-R, the most widely used printer language from Epson. | 20:51.26 |
| Samsung SPL, for Samsung printers | 20:51.26 |
| I'm still very, very puzzled as to why it doesn't see either of my HPs at home in the app store build. | 20:52.17 |
henrys | Jogux: you can run wireshark. I built a little app with their library and watched the snmp stuff go back and forth between the printer but it ended up failing. It is an old printer though. | 20:54.09 |
Jogux | henrys : yeah, I need to investigate, just not had time yet :-( | 20:55.03 |
| I'm certain there use to be a list of supported printers, and that is has since vanished. I suspect it may have vanished for a reason (eg. was incorrect). | 20:56.07 |
henrys | Jogux: yes I did see the PDL list was hoping for a bit more detail. | 20:56.43 |
marcosw | Robin_Watts henrys and Jogux: I'm back, just read the logs. | 20:58.22 |
Jogux | marcosw : worth reading the email I sent you too (subject Picsel VMs or something) | 20:59.07 |
henrys | marcosw: I think you have the most job security at this point. | 20:59.18 |
marcosw | henrys: ironic, since I'm part time :-) | 20:59.33 |
| how did you guys find the VM backups data on the nas? I spent some time browsing it and missed that. | 21:00.04 |
Robin_Watts | marcosw: Jogux takes the credit for that. He insisted it had to be there, and went looking for it. | 21:01.00 |
Jogux | marcosw> showmount -e 10.1.10.20 from points :-) | 21:05.36 |
| marcosw : tbf, I had the advantage that I knew we should have VM images somewhere | 21:06.22 |
| because I know the guy that prepared the VM images for the administrators ;-) | 21:06.46 |
marcosw | Jogux: the reason I ask is that there is about 1TB of stuff on the nas that I can't find (i.e. 2.9TB are used but the /mnt/picsel stuff only adds up to 1.8TB). I'll mount all of the showmount volumes, back them up, and then you guys can poke around. | 21:09.22 |
Robin_Watts | marcosw: How are the DELL uber servers currently set up? | 21:10.26 |
Jogux | marcosw> cool, thanks. afaict, there's nothing on the other volumes. | 21:10.30 |
Robin_Watts | Can each node quickly get to it's 'co-nodes' harddrives? | 21:10.48 |
Robin_Watts | isn't sure of terminology here. | 21:11.07 |
marcosw | robin_watts: yes, they are all one a switch and stuff moves between them at ~100MB/s | 21:11.28 |
Robin_Watts | HOw much free HD space do they each have? | 21:11.58 |
marcosw | they each have one 250GB drive, which is about one half full. | 21:12.12 |
| but all have at least two extra drive slots, and drives are very cheap. | 21:12.24 |
Robin_Watts | marcosw: OK, probably enough to be going on with. | 21:13.16 |
| The issues then would be getting the data to the servers. | 21:13.31 |
| Ah, you'd backed up onto a USB drive, right? Is there room to back up the VMs on that too? | 21:13.58 |
marcosw | btw, I pay about $0.10 per kw at night and twice that during the day. The servers cost about $200/month in power (at the current utilization). | 21:14.04 |
Robin_Watts | Then that backup could be plugged into one of the servers? | 21:14.12 |
Jogux | I think at least one extra huge harddrive would be useful (so we can have a look through the mailserver too) | 21:14.25 |
marcosw | yes. It's a 4TB USB drive and so there is plenty of space. And yes, driving to Miles' office is likely the fastest way to move the data. | 21:14.38 |
Jogux | marcosw : ouch! That's only going to go up once we start running ATS :-( | 21:14.41 |
Robin_Watts | What runs the VMs? Virtualbox? | 21:15.16 |
Jogux | I think it's a good option in the shortterm regardless. | 21:15.19 |
marcosw | at the staff meeting I played around with backup the NAS to amazon glacier. I gave up when I realized that it would take 47 days to upload it all. | 21:15.22 |
Jogux | robin_watts : kvm | 21:15.22 |
| marcosw : :-) | 21:15.31 |
| robin_watts : http://www.linux-kvm.org/page/Main_Page | 21:15.51 |
Robin_Watts | That should work on ubuntu, right? | 21:16.02 |
Jogux | yes. I'm hoping MarcosW has come across it before. If not I would reckon mace has. | 21:16.14 |
marcosw | doesn't kvm run at a lower level than the os? | 21:16.29 |
Jogux | marcos> yes, but under the host os I believe. | 21:17.19 |
| in theory, sudo apt-get install qemu-kvm libvirt-bin ubuntu-vm-builder bridge-utils :-) | 21:17.30 |
marcosw | hey, I found the control panel screen that I was looking for this morning. The VM Backups is 1.04 TB in since, none of the other volumes contain anything (i.e. the Download folder is 12 KB). | 21:17.33 |
Robin_Watts | marcosw: Yeah, on the web front end? That's where I was fiddling. | 21:17.58 |
Jogux | marcows : I found that, then lost it again :-) | 21:18.03 |
marcosw | Jogux: I'll read up on it. I've spent time with Virtual Box and Vmware Fusion so how hard can yet another virtual thing be :-) | 21:18.37 |
Jogux | hehe | 21:18.43 |
| should we talk to whoever needs to authorise mace to work on this? | 21:19.16 |
| as I suspect once we've got the VMs running, getting ATS to work will require further bashing. | 21:19.48 |
| and mace knows these systems very well | 21:21.31 |
| given he architected the whole thing ;-) | 21:21.52 |
Robin_Watts | foods. | 21:22.02 |
marcosw | robin_watts: I suspect I'm now doing all the same things you did trying to mount VM Backups :-) | 21:22.10 |
Jogux | :-) | 21:22.25 |
henrys | Jogux: he's authorized just send the invoice to miles | 21:22.43 |
Jogux | henrys : awesome, thanks :-) | 21:23.19 |
| marcosw : it's really just the vhost2 directory we need most urgently - any rough guess as to when you might be able to get that onto the server at your place? | 21:25.02 |
| it's about 200G I think. can be (and probably should be!) split across multiple nodes. | 21:25.27 |
marcosw | Jogux: 200G is going to be hard to send out from Miles' office via internets, so that means a drive to San Rafael (about an hour from my home). | 21:26.23 |
Jogux | marcosw : yeah, sorry. :-( hope it's at least a scenic drive or something :-S | 21:27.00 |
marcosw | unfortunately not so much scenic. but miles usually buys me lunch when I get there, so I have that going for me, which is nice. | 21:27.36 |
marcosw | is confounded by why the VM Backups woun't mount over nfs... | 21:28.02 |
Jogux | cool :-) | 21:28.02 |
| marcosw : to be perfectly honest, I'm slightly confused as to why File Server /does/ mount. | 21:29.54 |
marcosw | Jogux: how big is vhost2? 200 GB? I can just use the nas gui to save it locally on one of the cluster nodes at miles' office :-) | 21:30.06 |
Jogux | marcosw : I just found the setting! | 21:30.22 |
| marcos : 200 I think | 21:30.25 |
marcosw | Jogux: where ?!?! | 21:30.32 |
Jogux | marcosw: can you try mounting again please? | 21:30.45 |
| marcosw : I suspect where you were looking (the nfs permissions for the mount), but in the "select permission type" dropdown, select 'nfs host access'. | 21:31.14 |
marcosw | worked | 21:31.16 |
Jogux | it was set to access right: 'deny access', it's now "no limit" | 21:31.41 |
marcosw | is that anywhere near Control Panel -> Shared Folder? | 21:31.52 |
Jogux | it is - see the dropdown menu at the top of that? | 21:32.09 |
marcosw | no... | 21:32.46 |
Jogux | https://www.dropbox.com/s/rhmyukaqw31fr1s/Screenshot%202014-03-18%2021.33.01.png - that menu :) | 21:33.13 |
marcosw | oh, right. Now I feel stupid. | 21:33.42 |
Jogux | ah, you have to click on the middle icon on the 'action' column to get into that | 21:33.43 |
| marcosw : I missed it the first time too :-( | 21:33.56 |
| and the second time, tbh | 21:34.02 |
marcosw | Preusmably I did this for the File Server a couple of weeks ago :-) NFS sharing was turned off on the NAS when I got it. | 21:34.38 |
Jogux | okay, that is slightly more embarrassing :-) | 21:34.52 |
marcosw | when I had kids 16 years ago my memory started failing. when i turned 50 it started failing faster ... | 21:35.30 |
| okay, now to backup the VM Backups (assuming I can remember how to do that :-) ) | 21:36.09 |
Jogux | :-) | 21:36.17 |
| and the good news is, it's actually only 144G (I think the UI has some dodgy figures) | 21:36.35 |
| and the mail server is "only" 342G. | 21:37.39 |
marcosw | got the backup rsync started on the first try. little victories is all I have left. | 21:37.50 |
Jogux | hehe | 21:38.06 |
| there's a couple of things here that I'm not 100% sure what they are. eg. license1 is puzzling me. it's possible that might have some arm ADS licenses on it (which are worth a few thousand dollars each iirc) | 21:39.22 |
marcosw | so I'm getting 900KB/s transfering files from Miles' office to my house. I believe that means ~44 hours for 144GB. | 21:46.31 |
henrys | marcosw: that's got to compress no? | 21:47.05 |
marcosw | I'm assuming the data is already compressed (as a test I'm transferring a debian .iso file which I'm sure is). | 21:47.59 |
| Jogux: so once I get the the contents of vhost2 to my house and get kvm running (which appears straightfoward) what then? | 21:49.25 |
| bbiaw | 21:51.47 |
Jogux | marcosw> I suggest then just set mace loose on it | 22:11.16 |
| he'll need to figure out what is where etc, probably tweak hostnames in various config files, etc, etc. | 22:12.02 |
| marcosw : that sounds like an equally good solution. I suggest transferring win8-64-ats01 last; it's the largest and possibly, currently, the least interesting. | 22:13.04 |
| interestingly I'm getting 900 kB/sec now; wasn't that high earlier. | 22:13.39 |
| (I'm grabbing win2003_ats01.img, as I hope that might be a pre-setup useful build environment) | 22:14.08 |
| marcosw: I reckon if you can sustain the 900 kilobytes/sec, and exclude win8, everything should be there round about this time tomorrow-ish. | 22:15.31 |
marcosw | so I'll want to transfer the five *ats01/*.img files? what's the ats02 file? | 22:33.41 |
Jogux | marcosw : could you try to mount /mnt/vmbackups/vhost1/mail/mail-picsel.img and /mnt/vmbackups/vhost1/intranet2/intranet2.raw on points - ie. losetup, kpartx, mount ( http://dgc.uchicago.edu/20130530/mounting-a-kvm-disk-image-without-kvm/ ) (sorry, I seem to be asking you to do lots... if you want to give me root I'm fine with that too :-) ) | 22:33.52 |
marcosw | Jogux: I don't mind, but I can give you root as well. | 22:34.20 |
Jogux | marcosw : either, both, I'm fine either way. :-) | 22:34.44 |
| marcows : the ats02; not entirely sure; I think it's a second linux build or test client. best if we can have it. | 22:35.05 |
| marcosw: ats_master and db-server are very important too | 22:35.24 |
marcosw | each ats*.img runs on one machine or a cluster of machines? | 22:35.34 |
Jogux | in the original, real, ats, each .img was on a physical machine. | 22:35.54 |
| one machine per .img. | 22:36.02 |
| ideally if you could spread across as many machines as practical that would be good | 22:36.47 |
| possibly I mean "nodes" in this situation rather than "machines. | 22:37.18 |
| it's hard to know the exact setup, but I suspect debian32-ats01 and debian32-ats02 should definitely be on as seperate machines as possible. | 22:38.01 |
| as I think they'll burn through the most cpu. | 22:38.18 |
| (but in it's final days, smartoffice tech were running the whole system on a single machine I believe.) | 22:39.13 |
marcosw | I've mounted the first partition of the ./mail-picsel.img on /mnt/mail1, unfortunately it's not interesting... | 22:40.24 |
Jogux | agreed :-) | 22:41.02 |
marcosw | the second partition appears to be swap | 22:41.04 |
| I can't mount partition 5 (the third partition): mount: unknown filesystem type 'LVM2_member' | 22:41.30 |
Jogux | ah, darn, that's a LVM blah. | 22:42.10 |
| http://www.brandonhutchinson.com/Mounting_a_Linux_LVM_volume.html may help | 22:42.33 |
marcosw | this was more helpful: http://www.itbox4vn.com/2011/06/fixed-unknown-filesystem-type.html | 22:47.58 |
| I've mounted the partition /mnt/mail2 | 22:48.24 |
Jogux | awesome | 22:48.46 |
Jogux | has a look to see if I can find the docs henrys was after | 22:48.55 |
marcosw | of course I've already forgotten what I did, so don't ask me to do it again :-) | 22:49.18 |
Jogux | ah, actually i'm going to run into permissions issues - if you don't mind giving me root that's probably the easiest solution | 22:49.53 |
marcosw | done. | 22:50.19 |
Jogux | ta! | 22:50.26 |
marcosw | thinks he probably should have backed up points before giving Jogux root :-) | 22:53.48 |
Jogux | if I were you, I'd have backed it up before giving Robin_Watts root ;-) | 22:55.34 |
marcosw | good point... | 22:55.48 |
Jogux | henrys : got access to the mail store. grepping for mails from embedded-general; hopefully it'll finish by morning... | 22:58.53 |
| marcosw : feel free to kill / nice whatever my grep if it's affecting anyone else... | 22:59.09 |
henrys | thank Jogux | 22:59.48 |
| thanks | 22:59.54 |
Jogux | no problem | 23:00.17 |
| marcosw : can you try similar mount tricks with /mnt/vmbackups/vhost1/intranet2/intranet2.raw please? hopefully I can find the wiki data... | 23:00.54 |
| oh, I can do it. d'oh. ignore me ;) | 23:01.05 |
marcosw | Jogux: no worries about using to much cpu, points is the backup bugzilla machine, so unless picas (the actual bugzilla machine) fails it is completely idle. | 23:03.18 |
Jogux | marcosw : ah, great. | 23:03.45 |
marcosw | i'm going to umount mail1, might as well keep things somewhat tidy. | 23:04.08 |
Jogux | sure | 23:04.19 |
marcosw | btw, the vmbackup backup is going to take a while, so don't damage anything :-) | 23:05.29 |
Jogux | er. hm. yeah. I'd thought it was ro mounted, but I now see it's not :-) | 23:07.09 |
Jogux | remounts /mnt/mail2 etc as ro :-) | 23:07.47 |
marcosw | if you stop your grep I'll fix taht. | 23:07.50 |
| never mind, | 23:07.56 |
Jogux | tries to remount /mnt/vmbackups as ro too, but it just ignores me | 23:08.37 |
| but... yeah. I'll be careful. thanks for reminder :-) | 23:08.53 |
Jogux | finds the wiki :-) | 23:09.52 |
| I'm on a roll today | 23:09.55 |
marcosw | I have to run, but I think I've outlived my usefulness in any case. I'll drive up to see Miles tomorrow and transfer the various kvm images to an external drive and bring them home. If there is anything I should grab in additon to the contents of vhost2 send me an email. | 23:12.45 |
Jogux | marcosw : great, thanks. vhost2 should be the most pressing; anything else can wait, or we can grab any missed really important bits with loopback points as that seems to work pretty well :-) | 23:16.08 |
| henrys : http://www.embeddedgeneral.com/Printer_support_List.pdf | 23:21.20 |
| I utterly boggle that isn't publically linked or in google. | 23:21.32 |
| I can't entirely believe it's correct either; picsel's press release claimed "thousands" of printers I'm sure. | 23:22.07 |
henrys | Jogux I don't I think EG has some IP issues | 23:22.20 |
| and they keep a low profile | 23:23.41 |
Jogux | ah right | 23:25.07 |
| I see references to some bugzilla bugs about printing support; we may want to try looking at those I guess. | 23:25.27 |
henrys | Jogux: yeah the focus should be layout | 23:26.17 |
| Jogux: we're going to work on the printing and leave you with the hardcore SOT stuff. | 23:27.40 |
Jogux | henrys : pedro is busy looking into layout (now he's got a windows build to use), trying to eval the problems to figure out what to tackle first | 23:27.44 |
| henrys : cool | 23:27.50 |
| henrys : you're happy enough for me to keep pushing at ATS till we get a working system too? hopefully with other people doing most of the work :-) | 23:28.34 |
henrys | well the printing doesn't really need a lot of SOT expertise and we do have pretty good printing know how ⦠Yeah the ATS is critical | 23:29.13 |
| can't do a release without a regression ! | 23:29.25 |
Jogux | yep :-) | 23:29.33 |
| henrys : want some API docs too? | 23:38.45 |
henrys | Jogux:sure | 23:39.00 |
| Jogux: although I've pretty much figured it out from building a small program. | 23:39.30 |
Jogux | henrys : email winging it's way to you. pretty sure they're not the latest; the latest is probably in there somewhere if it's worth the time for someone to dig further. | 23:41.07 |
henrys | Jogux thanks heading out for a bit probably see you tomorrow, getting late over there isn't it? | 23:42.54 |
Jogux | yeah, er, kinda. | 23:43.07 |
Jogux | is one of those "I'm making progress so keep at it" guys :) | 23:43.17 |
| but, will be stopping before too much longer :-) | 23:43.39 |
| picsel twiki is on casper - /home/joseph/picsel-twiki/data, hunt for .txt files for the moment (unless/until we setup a proper install) | 23:44.40 |
| Forward 1 day (to 2014/03/19)>>> | |