| <<<Back 1 day (to 2013/10/16) | 2013/10/17 |
mvrhel_laptop | marcows: you there? | 05:33.48 |
marcosw_ | mvrhel_laptop: yup | 05:33.56 |
mvrhel_laptop | ok. so did you see that I fixed the bug that you told me not to post a solution to? | 05:34.14 |
| I was not sure if I should actually close the bug | 05:34.27 |
| the test file should be added to our regression testing | 05:34.44 |
marcosw_ | you can do that, I didn't give the customer the bug number. | 05:34.57 |
mvrhel_laptop | as fixing it resulted in no diffs | 05:34.58 |
| I think you did | 05:35.04 |
marcosw_ | did I? | 05:35.15 |
mvrhel_laptop | yes | 05:35.25 |
| 694707 | 05:35.29 |
| so I did not close it | 05:35.49 |
marcosw_ | a gmail search doesn't show any emails to the customer with the bug number, and I don't remember doing it (though the former is more reliable than the latter). | 05:37.25 |
mvrhel_laptop | well the number is in bugzilla yes? at least I see it | 05:38.04 |
| the customer number that is | 05:38.11 |
marcosw_ | I don't think the customer checks bugzilla nor are they likely aware of their customer number. Go ahead and close the bug. I'll make sure to add the test file to the repository the next time I add files. | 05:39.35 |
mvrhel_laptop | ok sounds good | 05:39.59 |
| do you want the commit fix hash code in the bug marcosw_? | 05:40.25 |
| I can keep it private | 05:41.03 |
marcosw_ | good idea, make it easier to send a patch when the customer is up to date with their reports. | 05:41.24 |
mvrhel_laptop | yes | 05:41.30 |
Micha` | Any voodoo reason why pdf_dict_puts_drop(group, "K", ...) is done twice in pdf-device.c:815? | 09:16.35 |
Robin_Watts | Micha`: Morning. Let me check. | 11:11.34 |
| Micha`: Stupidity on my part, probably. | 11:13.24 |
| tor7: ping | 11:23.38 |
| tor7: I've forgotten which of the commits on robin/master you've OKd. | 11:24.12 |
tor7 | Robin_Watts: all except the openjpeg and zeniko's | 11:30.07 |
| including the duplicated line one | 11:30.13 |
| I've still got to tweak the openjpeg git repo scripts | 11:30.25 |
| ran out of time last night | 11:30.31 |
Robin_Watts | Ok, I'll push all of them except the openjpeg one and will wait for you to sort that. Thanks. | 11:30.52 |
tor7 | ... and been too busy setting up the chromebook debian | 11:30.55 |
| but now I have mostly everything working just perfectly | 11:31.06 |
Robin_Watts | So you've got it dual booting? | 11:31.23 |
| What's chromeOS like? | 11:31.41 |
tor7 | yup. I have it dualbooting debian wheezy from an SDXC-card | 11:31.55 |
| chromeos isn't all that bad, actually | 11:32.03 |
| *if* you're willing to live in the google ecosystem that's rammed down your throat | 11:32.17 |
Robin_Watts | chrome (the browser) has an extension that lets you do VNC (or similar) to any machine. | 11:32.29 |
tor7 | all the NaCl apps on the chrome webstore thingy work perfectly | 11:32.35 |
Robin_Watts | so if you've got a net connection you're golden. | 11:32.53 |
tor7 | and there is an ssh shell etc | 11:32.57 |
| if you set it in developer mode, you can also get a local linux shell | 11:33.09 |
| apparently you can have it in semi-developer mode, with both the local shell and verified boot crap | 11:33.28 |
Robin_Watts | Given that most of the time I'm using the laptop I'm on a flight (or somewhere with a really crap connection), I'm not sure it'd be such a good idea for me. | 11:33.34 |
tor7 | /usr/local lives in a separate 25G portion along with your home directory | 11:33.42 |
Robin_Watts | Possibly we should install mosh on casper. | 11:34.05 |
tor7 | and there's "crouton" which sets up a chroot debian or ubuntu environment running from there | 11:34.18 |
| but when I use crouton, something is wasting cpu cycles so the fan spins up | 11:34.36 |
| which never happens if I dual boot | 11:34.40 |
| crouton is actually pretty neat. it sets up a separate virtual terminal (ctl+alt+F2) for the X11 server. | 11:35.28 |
| and you can jump back and forth between the chroot debian X and ChromeOS | 11:35.41 |
| sadly, linux toolkits really haven't learned how to do high-DPI yet | 11:36.07 |
| so all the gnome crap is ridiculously tiny | 11:36.15 |
| you can bump up the font size, so you get normal sized text crammed into these ridiculously tiny buttons | 11:36.34 |
| checkboxes the size of a speck of dirt | 11:36.43 |
| firefox copes better than most though, if you set the about:config device pixel scaling to 2 | 11:37.09 |
oznt | hi guys | 11:48.23 |
| can someone give me a quick tipp regarding mupdf? | 11:48.45 |
Robin_Watts | oznt: Ask away. I've got to pop out for 10 mins though. | 11:52.45 |
oznt | I really want to use mupdf inside a gtk application. I know how to use the cairo canvas already, but I dont get how to use mupdf to draw on the canvas | 11:53.53 |
| using poppler it takes only poppler_page_render(page, cr); where cr = gdk_cairo_create(w->window); | 11:55.00 |
| so my question is : what is the equivalent (if any) in mupdf? | 11:55.23 |
tor7 | oznt: render to a fz_pixmap, wrap it in a gdk_pixbuf_new_from_data, create a gtk_image_new_from_pixbuf | 11:59.19 |
| http://git.ghostscript.com/?p=user/tor/mupdf.git;a=blob;f=platform/gtk/muview_gtk.c;h=b571d648599b2cc350dab66ec8dd2f88b07852e4;hb=dfdc944ca72356624ec85a18fc47a27f9abcb3cf#l132 | 11:59.49 |
oznt | tor7, i was thinking about it, that means that each page is a pixmap? can i also do with svg ? is more "expansive"? | 12:01.06 |
tor7 | in the end, a page will always be turned into a pixmap | 12:01.21 |
| even svg | 12:01.24 |
oznt | and a big YAY! this is cool you are putting a GTK example!!! | 12:01.35 |
| this is cool! | 12:01.39 |
tor7 | it's a somewhat working, very incomplete, gtk+ based viewer | 12:01.53 |
| it's on my private repo, and not really fit for general consumption yet | 12:02.07 |
| but feel free to dissect it | 12:02.13 |
oznt | tor7, well, it's good enough for start. the only gtk app i found that uses mupdf was zathura, but they wrapped mupdf in so many levels i could not easily understand it | 12:07.15 |
Robin_Watts | oznt: You were asking about svg... | 12:14.06 |
oznt | Yes, tor7, answered more or less | 12:14.29 |
Robin_Watts | mupdf now has support for outputting to an svg. It's not perfect, but it's not bad. | 12:14.32 |
| But to get mupdf to render to an svg, then to get gtk to render the svg would seem strange. It means you're going through 2 different rendering systems. | 12:15.06 |
| As tor7 says, better to render direct to a pixmap (that happens to really be a gtk image that you can blit). | 12:15.30 |
oznt | tor7, how can I clone your repository? | 12:15.54 |
tor7 | oznt: git clone git://git.ghostscript.com/user/tor/mupdf.git | 12:16.24 |
oznt | thanks | 12:16.28 |
| tor7, i cloned your repository but i don't find the files for gtk, are they in some branch? | 12:22.12 |
Robin_Watts | oznt: "viewer" branch | 12:24.38 |
oznt | ok, sorry for all the noob questions. cloning by default only cloned master | 12:25.41 |
tor7 | oznt: no, you've got all of the branches | 12:25.56 |
| then you need to do "git checkout viewer" | 12:26.02 |
| and also "git submodule update --init" to set up the thirdparty libraries that are needed | 12:26.40 |
oznt | tor7, ok, it seems to compile | 12:28.21 |
| how come the viewer files are not inside the directory platform in my git checkout ? I am getting confused here | 12:28.51 |
tor7 | oznt: what does "git branch" tell you? | 12:29.31 |
oznt | I am on viewer | 12:29.49 |
| but the directory platform does not exists | 12:30.01 |
tor7 | that sounds broken... :( | 12:30.25 |
| which files do you have at the root? | 12:30.32 |
| oznt: "git status" might give some hints on what's happened to your files | 12:31.22 |
oznt | tor7, I did git clone http://git.ghostscript.com/user/tor/mupdf.git, and then "git submodule update --init" | 12:32.23 |
| nothing more, for some reason the viewer is not inside platform like in the main repository | 12:32.46 |
tor7 | oznt: then you need to do "git checkout viewer" to switch to the viewer branch | 12:33.06 |
oznt | i did that too, ofcourse | 12:33.20 |
tor7 | rigel ~/src/mupdf $ git log -1 | 12:33.31 |
| commit dfdc944ca72356624ec85a18fc47a27f9abcb3cf | 12:33.31 |
| ... | 12:33.31 |
| do you see the same sha1 checksum? | 12:33.38 |
oznt | no, the latest commit is for some reason 18 jun | 12:34.44 |
| 87d48305e16af7c5d1021fc14a1c8acbd83d5191 | 12:34.53 |
| is the checksome | 12:34.58 |
tor7 | then you've managed to check something else out ... | 12:35.36 |
| git log -1 origin/viewer | 12:36.22 |
oznt | commit 524fc5c1d6ff63ae09b2bc6510fdf78456e589c2 | 12:37.11 |
tor7 | I have no commit with that checksum anywhere in my repository... | 12:37.51 |
| are you sure you're in the right directory? | 12:37.57 |
oznt | yes, i made a clean clone several times already | 12:39.12 |
tor7 | oznt: try "git reset --hard dfdc944ca72356624ec85a18fc47a27f9abcb3cf" | 12:39.36 |
| I have no idea what's going on, but if you run that command you should end up at the latest viewer branch | 12:40.04 |
oznt | i did copy the git url from the webbrowser, now I coppied from the irc window, and it seems to work | 12:40.35 |
| don't know what is going on | 12:40.35 |
tor7 | oznt: "git remote -v" will show you the url it's cloned from | 12:41.17 |
oznt | too late now ... already erased. | 12:41.43 |
| but the snapshot from web built successfully. the gtk window is cool. Still does not display the cover image on the first pdf page | 12:42.12 |
| but you already said it is basic ... but at least i can start from here, with packing other widgets and making it usable | 12:42.51 |
Gigs- | I had a similar problem oznt, erasing the tree and pulling fresh helped | 14:10.44 |
| at least wrt missing submodules | 14:10.54 |
| mupdf should be good enough not to miss entire images btw | 14:12.07 |
Robin_Watts | gah. vs just died and lost me an hours work :( | 14:49.53 |
kens | ?me has had that | 14:50.04 |
Robin_Watts | And it just did it again :( | 14:57.02 |
kens | Yeah I've had that sometimes too | 14:57.16 |
| My only solution is to work differently so as not to trigger it | 14:57.35 |
Robin_Watts | kens: Yeah, it just triggered in exactly the same place. Oh joy. | 14:59.06 |
kens | :-( | 14:59.25 |
chrisl | Try a different version? | 15:00.04 |
Robin_Watts | Let's close the project and delete everything from debugobj etc. | 15:00.12 |
| It's going to be tripping over a corrupt .sbr or something, I bet. | 15:00.38 |
| Still dying. Use emacs instead. | 15:09.11 |
chrisl | OMG! It can't be that bad....!! | 15:09.47 |
ray_laptop | mvrhel_laptop: On GrayDetection -- how did you verify that you get it right (do you have any suggestion for how to automate testing)? And were there some files that were 'tricky' ? | 15:14.01 |
mvrhel_laptop | ray_laptop: I ran it through with various color space tests | 15:14.52 |
| there are only a limited number | 15:15.04 |
| Robin_Watts: I had a question for you | 15:15.35 |
Robin_Watts | uh oh. | 15:15.43 |
mvrhel_laptop | so in my stuff and in the android app, the text search function and the gethtml function both use fz_run_page which uses the doc | 15:16.21 |
Robin_Watts | ok. | 15:16.46 |
mvrhel_laptop | is it safe for that to be happening on another thread while some other thread is doing a get_page on the doc? | 15:16.57 |
Robin_Watts | No. | 15:17.02 |
mvrhel_laptop | my guess would be no | 15:17.04 |
| ok | 15:17.06 |
| I need to keep that from happening | 15:17.12 |
| as thumbnails could be rendering still | 15:17.24 |
Robin_Watts | I *hope* we're safe from that in the android app, as everything stops while we search. | 15:17.31 |
mvrhel_laptop | when someone starts a text search or does a conversion of a page to html | 15:17.40 |
| I need to do the same in terms of thumbnail and/or upcoming page rendering | 15:18.04 |
| thanks | 15:18.24 |
Robin_Watts | mvrhel_laptop: Possibly you might want to put your own lock around uses of the doc. | 15:18.34 |
mvrhel_laptop | Robin_Watts: yes, I think that would be a good idea | 15:18.51 |
Robin_Watts | That way any search would wait until a thumbnail had finished. | 15:19.06 |
mvrhel_laptop | yes | 15:19.14 |
Robin_Watts | and any thumbnail wouldn't be able to interrupt a search. | 15:19.19 |
henrys | chrisl: I have no idea how you work without emacs do you gdb with the console? Showing one line of code at a time? | 15:19.25 |
chrisl | henrys: I use ddd | 15:19.55 |
ray_laptop | mvrhel_laptop: sorry, I didn't understand that. I understand that there are a limited number of colorspaces, and I know that with images you have to check each pixel, but what about other painting ops -- in particular Pattern colorspaces | 15:20.14 |
henrys | chrisl: ahh | 15:20.28 |
mvrhel_laptop | thanks Robin_Watts. MT rendering with mupdf is finally starting to gel in my brain | 15:20.31 |
| ray_laptop: anytime there is a color conversion and the source color is not gray then we will check | 15:20.58 |
| with a pattern that will happen within the pattern | 15:21.09 |
ray_laptop | do you check the contents of a colored pattern (PaintType 1) | 15:21.10 |
mvrhel_laptop | I guess I don't know what you are asking | 15:21.16 |
ray_laptop | mvrhel_laptop: Oh, I see -- so when the pattern converts the color to paint it. Got it. | 15:21.46 |
mvrhel_laptop | yes. anytime we would be doing a operation with lcms we will check | 15:22.07 |
| unless the source is gray | 15:22.13 |
| lcms does not understand patterns | 15:22.21 |
ray_laptop | mvrhel_laptop: right. | 15:22.35 |
mvrhel_laptop | kens: I will look over your latest fun one later today or this weekend if that is ok | 15:55.50 |
kens | mvrhel_laptop : yes np | 15:55.57 |
| I'll look at fuzzing bugs in the meantime | 15:56.12 |
henrys | the new iPhone comes today I've heard mixed opinions⦠we'll see | 15:57.00 |
kens | 5s or 5c ? | 15:57.13 |
henrys | 5s | 15:57.20 |
kens | suppose to be the good one... | 15:57.29 |
henrys | I've heard very positive and negative stuff about ios7 | 15:58.13 |
kens | yeah that is true | 15:58.25 |
Gigs- | marcosw: it's great you were able to reproduce it, I am surprised but it's definitely a good thing | 16:04.35 |
| Robin_Watts: hmm he gave the bug to you even though Ray put a comment that said he was taking it. Also downgraded the severity, but I'd challenge that a hard segfault on 0.8% of files in two devices is a normal priority bug. | 16:07.26 |
| http://bugs.ghostscript.com/show_bug.cgi?id=694714 | 16:07.31 |
Robin_Watts | Gigs-: Yeah, priorities are for us to set. | 16:07.50 |
Gigs- | I know, I just think it merits more | 16:08.09 |
Robin_Watts | Paying customers get high priority, other people lower. That doesn't mean it won't get looked at. | 16:08.16 |
Gigs- | don't you have paying customers using tiffsep now? | 16:08.29 |
| if anything I'd consider it blocker in the literal sense that I wouldn't make an official release with the bug in it, especially not if you have paying customer relying on tiffsep or psdcmyk | 16:10.41 |
marcosw | I'm going to reboot casper in a few minutes, it shouldn't affect anyone since I'm the only one logged on. | 16:14.19 |
Robin_Watts | Gigs-: Did it ever work? | 16:14.51 |
| I mean, is this a new thing? | 16:14.57 |
Gigs- | I can't be entirely sure, but I believe so. I did a pretty extensive corpus a while back and I didn't see it | 16:15.34 |
| I believe I did that extensive testing a while after planar went in | 16:16.02 |
| I can't say for sure whether it's in your current release or not though | 16:16.33 |
henrys | marcosw: at the meeting I asked about testing GX_COLOR_INDEX. Do we have test anything that requires recompiling now? | 16:17.06 |
| s/have test/test | 16:17.25 |
marcosw | henrys: do you mean as part of the weekly tests or a clusterpush/commit test? | 16:18.05 |
henrys | yes | 16:18.14 |
| the former | 16:18.29 |
marcosw | many of the weekly tests compile ghostscript with a different set of options, i.e. we do a debug build and a 32 bit build. | 16:18.54 |
henrys | commit testing would be a bit time consuming | 16:18.56 |
marcosw | henrys: yeah, commit testing wouldn't be practical and you can do a clusterpush test easily enough by changing the source. | 16:19.26 |
henrys | so maybe the 32 bit build tests gx_color_index == 32 bit. | 16:19.38 |
marcosw | do we want to add a weekly build with a different value for GX_COLOR_INDEX? Now that we are no longer testing -dDisableFAPI we have a slot available. | 16:23.27 |
mvrhel_laptop | bbiaw | 16:23.43 |
henrys | I believe the makefiles refer to USE_LARGE_COLOR_INDEX or some such | 16:24.27 |
marcosw | are you referring to -DGX_COLOR_INDEX_TYPE="unsigned long int" ? | 16:27.12 |
Gigs- | Robin_Watts: November 2012 is when I filed the bug about non-deterministic tiff output, so that's when I last did a big corpus. Assuming the non-deterministic output isn't actually the same memory corruption, that would place it between November 2012 and September 2013, if that helps | 16:28.36 |
Robin_Watts | Gigs-: If it used to work, and now it doesn't, then you could git bisect it. | 16:29.17 |
ray_laptop | good luck with that | 16:29.55 |
Gigs- | Robin_Watts: let me run a current corpus at 20dpi and see if I can get a reproduction at that resolution. That will be much faster to test than 300dpi. | 16:30.13 |
ray_laptop | I did find that git bisect works, but not quite as documented. | 16:30.16 |
| (for the logs) I found that I have to do the "git bisect start" the git bisect good followed by the git bisect bad, then start testing. If I try git bisect start <bad> <good> then it doesn't work | 16:33.14 |
| marcosw: speak of the devil (or bot) | 16:33.28 |
chrisl | marcosw_: if we're going to test with a GX_COLOR_INDEX=ulong I'm assuming it would make things easier if I put an override in configure? I don't think there is one just now | 16:36.28 |
marcosw | chrisl: yes it would, I was just looking into how to hack that by running sed (or something similar) on the Makefile. | 16:37.03 |
chrisl | marcosw: nah, I'll add the configure option tomorrow, *much* safer, I reckon | 16:37.32 |
marcosw | certainly less likely to break when the Makefile changes. | 16:37.49 |
chrisl | Well, you'll see the commit - I *might* hide the option (so it doesn't show up in the help) as we generally won't want people to fiddle with it | 16:38.47 |
marcosw | henrys: the 32 bit build sets GX_COLOR_INDEX_TYPE to "unsigned long long", so it's not different than the 64 bit build (assuming a 32 bit long long == a 64 bit long int). | 16:40.12 |
henrys | marcosw:oh so we don't test it ⦠don't forger ulong is 64 bits on most of the cluster machines so that won't really work. | 16:41.26 |
chrisl | marcosw: if there's a 64 bit integer variable available, we'll use it - on Linux/Unix that usually "unsigned long long" | 16:41.27 |
henrys | s/forger/forget | 16:41.37 |
| I guess we could use "int" | 16:42.31 |
kens | Goodnight folks | 16:44.13 |
sebras-hotel | kens: good night! | 16:44.24 |
marcosw | henrys: let's use a "char" and really stress the code :-) | 16:44.33 |
henrys | ;-) | 16:45.35 |
chrisl | I was thinking I was allow the type to set at the configure command line: --use_color_index="unsigned int" for example | 16:45.40 |
henrys | chrisl: that sounds good to me. | 16:46.22 |
| marcosw: ti would be useful for us to compare performance also, the timings are easy once you set up the tests right? | 16:47.03 |
marcosw | henrys: yes, the timings are automatically generated. I have to compare them to the benchmark, but as long as I remember to do it that's easy. | 16:47.55 |
henrys | chrisl: ail you change the pcl configure as well? | 16:48.54 |
| or should I do that? | 16:49.02 |
| s/ail/will | 16:49.17 |
chrisl | henrys: I'm okay doing that - it probably needs makefile changes, too, as I suspect it currently isn't setup to allow the type to set like the gs one is | 16:49.47 |
marcosw | oops, just realized that it's almost 10:00 and have to run to uni. my advisor is out of town so I'm running one of his classes today. there is a guest lecturer, so I don't get to do much, but I can't be late :-) Be back this afternoon. | 16:50.31 |
henrys | just USE_LARGE_COLOR_INDEX=0 | 16:50.32 |
| written to config.mak no? | 16:50.46 |
chrisl | henrys: that's for VS, not gcc, IIRC | 16:51.14 |
henrys | right you are they really should be the same though. | 16:51.52 |
| maybe i should bite the bullet and fix PCL so it doesn't need the damn setting | 16:52.30 |
chrisl | henrys: why does PCL need the setting? It's not immediately obvious | 16:53.07 |
henrys | it shouldn't I agree | 16:54.39 |
| let me set it up to break in the debugger and I'll try and fix it. | 16:55.08 |
chrisl | henrys: I assume for PCL we want to default to (the equivalent of) USE_LARGE_COLOR_INDEX=0 since it generally just needs RGB rendering? | 16:55.55 |
henrys | chrisl: correct | 16:56.23 |
chrisl | henrys: okay, thanks. | 16:56.53 |
henrys | actually without recursive make on the library it's hard to break it ! | 16:57.30 |
chrisl | henrys: BTW, on the font stuff: it'll take a bit more time - I had a problem with the simulator build (I always do!), and the other wrinkle is I can't access a list of glyph names in the MT fonts, so I need to do that differently. It shouldn't take much longer, though. | 16:58.56 |
henrys | chrisl: is there any evidence they make use of the cjk fonts in the ufst? | 17:08.14 |
| that would certainly be a shortcoming for us, if so. | 17:08.31 |
chrisl | henrys: they do, I had to add stuff to support that | 17:08.56 |
henrys | okay so that's a big one. There is some free font stuff about that might fill in some gaps. | 17:09.55 |
chrisl | IIRC, the only include two. They might have decided that we do with DroidSans is enough | 17:10.39 |
| henrys: they only include a gothic and a mincho for Japanese - IIRC, there are government sponsored free equivalents we could use | 17:26.03 |
henrys | chrisl: okay great news | 17:27.42 |
chrisl | I'll need to have a search for them | 17:28.27 |
henrys | is there some ways to stop this damned "devise busy" bullshit in windows when I delete a file gawd... | 17:33.20 |
Robin_Watts | henrys: What's the problem? | 17:34.12 |
henrys | if I try and delete a file an application is using it says device busy | 17:34.51 |
Robin_Watts | henrys: Yeah, windows won't let you delete stuff that's open. I'm not sure that's an entirely bad thing :) | 17:37.48 |
Gigs- | if you use a samba server, you can log in to the linux side and delete such files | 17:39.54 |
| the problem robin is that it says files are in use when they aren't sometimes | 17:40.25 |
henrys | no big deal just not accustomed to it. thanks | 17:40.47 |
Robin_Watts | Gigs-: It's relatively rare for windows to get it wrong. | 17:41.05 |
| Samba getting it wrong is another thing though :) | 17:41.13 |
chrisl | henrys: we can use the Gothic and Mincho fonts from IPA - the problem we'll likely have is size: the Monotype fonts 532 are currently using are the "special" compressed ones. | 17:41.27 |
henrys | chrisl: certainly not a problem for large printers | 17:42.59 |
| or high end | 17:43.07 |
Robin_Watts | ray_laptop: I have a first cut at this process_page thing here. | 17:45.12 |
| I need to pick a suitable device to test it with. | 17:45.22 |
| One without the downscaler :) | 17:45.33 |
chrisl | henrys: just something to remember - they were *very* keen on sharing the font data between our PDF and their PS, so it was a factor | 17:45.51 |
henrys | chrisl: yup ⦠I think cost may trump that but maybe not. | 17:47.02 |
chrisl | I gotta go - back to these headaches tomorrow....... | 17:57.42 |
ray_laptop | grrr... gitpush.sh doesn't work. The STOOPID host key for casper got changed. Why'd you do that marcosw ? | 20:17.49 |
| darn, and even fixing ~/.ssh/known_hosts on msysgit doesn't fix it. I can ssh to it, and can get to it with putty (after it confirms that I want to replace the host key) but the "remote: syncing step still fails) | 20:26.54 |
| has anyone been able to run a regression since marcos messed with (rebooted) casper ? | 20:27.36 |
Robin_Watts | ray_laptop: For the logs: Try now. | 23:06.57 |
marcosw | no idea why the host key for casper changed. I'll change it back. | 23:28.39 |
| okay, host keys restored from backup. I'll try to figure out why the changed, the only thing I did was an apt-get upgrade and a reboot. | 23:33.01 |
| ray_laptop: I put the old host keys back, so everything should work again. | 23:41.28 |
ray_laptop | marcosw: I don't know what's going on. I get a different host key, but I still get an error from gitpush.sh | 23:42.08 |
| The fingerprint for the RSA key sent by the remote host is | 23:42.24 |
| f1:d8:04:95:43:23:9c:02:79:d9:ca:18:9b:5f:c5:a5. | 23:42.26 |
| oops. That's the one I used to get. | 23:42.41 |
| Now I get: | 23:43.02 |
| The fingerprint for the RSA key sent by the remote host is | 23:43.04 |
| 31:f6:2f:bd:29:2c:98:ac:45:c3:b1:21:61:c2:6e:91. | 23:43.05 |
| followed (as before) by: | 23:43.32 |
| Add correct host key in /home/regression/.ssh/known_hosts to get rid of this message. | 23:43.33 |
| Offending RSA key in /home/regression/.ssh/known_hosts:50 | 23:43.35 |
| remove with: ssh-keygen -f "/home/regression/.ssh/known_hosts" -R casper.ghostscript.com | 23:43.36 |
| RSA host key for casper.ghostscript.com has changed and you have requested strict checking. | 23:43.38 |
| Host key verification failed. | 23:43.39 |
| rsync: connection unexpectedly closed (0 bytes received so far) [sender] | 23:43.41 |
| rsync error: unexplained error (code 255) at io.c(605) [sender=3.0.9] | 23:43.42 |
| I can do git pull from casper just fine. | 23:45.11 |
| and I can ssh to casper (after having deleted casper from the known_hosts and agreeing to accept the host key) | 23:45.50 |
marcosw | try the clusterpush again, I think Robin_Watts fiddled with the ~regression known_hosts file. | 23:48.08 |
ray_laptop | marcosw: better! THANKS!!! | 23:50.55 |
marcosw | great, sorry about this. still don't understand why the host keys changed. The apt-get I did before the reboot doesn't show anything. The only package that was updated at all related to encryption was gnupg, but it doesn't seem reasonable that it would change the keys. | 23:52.28 |
ray_laptop | marcosw: I just noticed that Robin_Watts had said "Try now." in the logs. | 23:53.01 |
marcosw | and in fact the apt-get upgrade finished at 16:18 gmt and the host keys changed at 16:31 gmt, so that couldn't have been it. | 23:54.16 |
ray_laptop | marcosw: I suspect that you are correct, since the earlier tries (with the f1.d8... host key) complained about line 19. After you restored the host key, it complained about line 50 | 23:54.40 |
| marcosw: when was the reboot ? (it may have generated a new host key after the reboot) | 23:55.45 |
marcosw | right you are. The reboot was ~15 minutes after the update was finished. I'll check the boot log. | 23:58.20 |
| Forward 1 day (to 2013/10/18)>>> | |