slangasek | Noskcaj: looking | 00:02 |
---|---|---|
slangasek | Noskcaj: it's not correct that it only fixes a CVE; there are changes to thread support handling, which I'm in no position to vet at the moment | 00:03 |
Noskcaj | oops, ok | 00:05 |
sarnold | I believe there's more work to be done to fix the security issues: http://www.openwall.com/lists/oss-security/2013/10/11/8 | 00:10 |
=== Noskcaj10 is now known as Noskcaj | ||
infinity | pitti: I just vomited a little at that udev/hv patch. | 10:49 |
infinity | pitti: How do these people not have a kernel driver that's doing the right thing here? Letting userspace manage cpu/memory hotplug sounds like madness. | 10:50 |
infinity | pitti: Aaanyhow, now that that's out of my system, closing my eyes and pressing "accept". | 10:50 |
c_korn | hello, a question regarding gensymbols. when building a package there is a warning: "warning: debian/liboyranos0/DEBIAN/symbols doesn't match completely debian/liboyranos0.symbols.amd64" but running "dpkg-gensymbols -pliboyranos0 -Pdebian/liboyranos0" does not do anything. | 12:16 |
mdeslaur | yesterday's iso gets stuck at the timezone selection screen...is that a known issue? | 12:48 |
sladen | mdeslaur: file it | 12:50 |
sladen | mdeslaur: dups can be sorted out later | 12:50 |
mdeslaur | sladen: thanks | 12:50 |
mdeslaur | bug 1239127 for the curious | 12:55 |
ubottu | bug 1239127 in ubiquity (Ubuntu) "20131011.1 iso gets stuck at timezone selection screen" [Undecided,New] https://launchpad.net/bugs/1239127 | 12:55 |
=== em is now known as qman | ||
=== qman is now known as em | ||
Laney | wgrant: did you forward the gst-base change? | 17:27 |
reaby | hello everybody, i'm new to this, and found out that it's very challenging to create a new launcher (shortcut) to desktop | 17:39 |
reaby | could it be possible for next release to have easier way ? | 17:39 |
reaby | thanks. | 17:40 |
reaby | native linux apps are fairly easy since you can just make a link, but a wine app or launcher for minecraft is impossible to do | 17:42 |
reaby | what i tried for minecraft was to do a bash script to launch it, but even if i put executable flag for the file, it still wants to open the file at gedit | 17:43 |
reaby | and when i try to change the file opener, there is no way to "just open" | 17:43 |
reaby | and if i try the search button from internet is just crashes | 17:44 |
=== doko_ is now known as doko | ||
doko | Laney, pitti: any chance to have a look at the gmime test failure? | 17:50 |
doko | looking at 2.6.18 | 17:58 |
doko | same failure with this version | 18:00 |
Laney | doko: probably https://git.gnome.org/browse/gmime/commit/?id=bb1c6094e04ce54de52289cd5ba7f9b73694ef1f | 18:04 |
Laney | let me try this | 18:04 |
Laney | seems to work | 18:11 |
doko | Laney, please upload | 18:17 |
Laney | sec | 18:17 |
Laney | just being killed on hitman | 18:17 |
Laney | ok, I died :( /me uploads | 18:19 |
Laney | done, it is | 18:25 |
halfie | hi, do you guys turn on "-Werror=format-security" by default in Ubuntu? | 20:24 |
halfie | cjwatson, are you around? | 20:27 |
halfie | what do you guys think about "prelink" in general? is there demand for it from Ubuntu users? | 20:27 |
Noskcaj | halfie, I don't think we do because Werror has many false positives. One of the more experienced devs might actually know. | 20:27 |
halfie | Noskcaj, oh I see. I was told that it had very very few false positives. weird. | 20:28 |
halfie | if it has known false positives then I can't "fight" to get it enabled. | 20:28 |
Noskcaj | Well, when Werror get false positives, it makes things FTBFS | 20:29 |
halfie | yes ;( | 20:30 |
slangasek | prelink defeats ASLR and modifies files owned by the package system; it should be avoided. | 20:31 |
slangasek | Noskcaj, halfie: we absolutely do turn on -Werror=format-security by default | 20:32 |
Noskcaj | slangasek, Thank you for knowing far more than i do about this stuff | 20:32 |
slangasek | Noskcaj: that's not the same as turning on -Werror generally, it's about treating a specific class of warnings as errors | 20:32 |
Noskcaj | ok | 20:32 |
halfie | slangasek, how do I verify / see the default flags? (I am a Fedora guy ;( ) | 20:33 |
cjwatson | they always show up in gcc -dumpspecs, although reading that takes practice | 20:33 |
slangasek | halfie: 'dpkg-buildflags' tells you the defaults; build logs for individual packages should show how these flags are actually applied, though many packages misbehave and hide their compiler commandlines in the output | 20:33 |
halfie | slangasek, yes, prelink is the first thing I remove. I am trying to get Fedora to disable it and give the choice to the user instead. | 20:33 |
slangasek | Fedora uses prelink? oy | 20:34 |
halfie | slangasek, BY DEFAULT! | 20:34 |
cjwatson | (my answer is the default for the compiler; slangasek's answer is the default for package builds, at least those that honour dpkg-buildflags) | 20:34 |
halfie | :-( | 20:34 |
halfie | cjwatson, both answers are useful for me. thanks! | 20:34 |
halfie | so, I have two action items. 1) do *not* use prelink by default 2) enable "-Werror=format-security" by default. | 20:35 |
slangasek | man, that sounds like a lot of work - maybe you should just switch to Ubuntu instead ;) | 20:36 |
halfie | slangasek, I did run Ubuntu for some years :-) ... I get paid to work on Fedora these days ;) have to be "loyal", right? :D | 20:37 |
halfie | (but I am trying to pick-up Ubuntu packaging stuff) | 20:38 |
halfie | Ubuntu also has "-Bsymbolic-functions" turned on. I have read that "it is permitted for users of the shared library to override a symbol (function name or global variable) with their own versions." ... how is this overriding actually done? | 20:44 |
soren | I'm confused. I'm looking at an ELF binary that doesn't seem to have an RPATH set ("objdump -x `which binary` | grep RPATH" is silent), says it needs libwhateverclient.so. libwhateverclient.so resides in /usr/lib/whatever, which isn't in /etc/ld.so.conf. Somehow it manages to find it anyway. | 20:46 |
soren | "ldconfig -v | grep libwhatever" is also silent. | 20:46 |
soren | Oh, hang on. Now it doesn't work anymore. | 20:47 |
soren | what the.. | 20:47 |
soren | Eeeek. | 20:48 |
soren | postinst has: "ldconfig /usr/lib/whatever" | 20:48 |
soren | Running "ldconfig -v" restored normalcy. | 20:49 |
soren | That's a first. | 20:49 |
soren | (and hopefully last) | 20:50 |
mdeslaur | soren: wow | 21:11 |
infinity | soren: Eww, seriously? | 21:24 |
infinity | soren: Is this packaged IN THE ARCHIVE? | 21:25 |
soren | infinity: Gosh, no. | 21:30 |
soren | infinity: It just has a built-in facility for producing (some sort of) debian packages. | 21:30 |
infinity | soren: Some sort of, indeed. | 21:31 |
soren | Ok, so according to http://lintian.debian.org/tags/binary-or-shlib-defines-rpath.html, it's fine for this package (let's call it "whatever") to have binaries specify /usr/lib/whatever as an RPATH. | 21:31 |
soren | Nonsense. | 21:32 |
soren | Or is it? Man, my packaging-fu is really rusty. | 21:33 |
infinity | soren: Yeah, a self-contained RPATH like that is fine. | 21:33 |
soren | No, my initial statement seems accurate... but how does that work with multiarch? | 21:34 |
infinity | soren: With multiarch, the rpath should also be multiarch? | 21:34 |
soren | infinity: Let's pretend that I'm an idiot... | 21:34 |
infinity | soren: lib in /usr/lib/$(triplet)/package | 21:34 |
infinity | soren: And RPATH of same in binary. | 21:34 |
infinity | soren: Not that it would matter desperately if it was /usr/lib, since your binary is /usr/bin (which isn't multiarch already), and the exlicit allowance here is for private libraries *in the same package*. | 21:35 |
infinity | (If you choose to read that as "from the same source package", which might be valid, then yeah, you could multiarch the private libs package, and use a multiarch RPATH). | 21:36 |
soren | infinity: Ah, ok. Yeah, it was the bit about the binary being in a non-multiarch location that got me confused. | 21:36 |
soren | Does one of the dh_* things remove rpath automatically? | 21:37 |
infinity | Not that I know of, unless someone made dh_strip more clever. | 21:37 |
infinity | I used to build-dep on chrpath and strip it myself. | 21:37 |
soren | I just wonder why my more proper debian packaging spits out binaries without this rpath set. | 21:37 |
soren | Now I know better where to look. Thanks. | 21:37 |
infinity | Oh, is the RPATH set in upstream's binaries, but stripped in yours? | 21:38 |
infinity | Curious. | 21:38 |
soren | Yeah. | 21:38 |
soren | This thing is a treasure trove of curiosities. | 21:38 |
infinity | soren: The only debhelper/rpath mention I see is: | 21:41 |
infinity | debhelper (7.3.9) unstable; urgency=low | 21:41 |
infinity | * cmake: Avoid forcing rpath off as this can break some test suites. | 21:41 |
infinity | It gets stripped by cmake at install time. Closes: #538977 | 21:41 |
infinity | -- Joey Hess <joeyh@debian.org> Sat, 01 Aug 2009 15:59:07 -0400 | 21:41 |
soren | Yeah, saw that. Maybe I'm full of it. The only RPATH I see mentioned now is pointing to my dev area (presumably, as Joey also mentions, for testing). | 21:42 |
soren | I wonder how this is mean to work anywhere. | 21:43 |
infinity | Well, it's meant to "work" due to that postinst snippet and a funamental misunderstanding that subsequent ldconfig calls will make it explode. :P | 21:43 |
infinity | It's not the first upstream who doesn't understand ldconfig. | 21:44 |
infinity | I've been in a long and heated debate with McAffee over their insistence that it's not a bug for one of their random crap libraries to claim an SONAME of ld-linux.so.2 | 21:45 |
infinity | You can guess what happens to user's systems on the next ldconfig run after installing their software. | 21:45 |
mlankhorst | well | 21:46 |
mlankhorst | not much any more, I guess | 21:46 |
infinity | *rimshot* | 21:46 |
soren | infinity: whuh? | 21:47 |
* soren falls asleep, confused | 21:50 | |
=== sam113101 is now known as sam113101_afk | ||
slangasek | halfie: overriding the symbol is done by simply providing a symbol of the same name in your application, which then takes precedence in the linker resolution | 22:13 |
Noskcaj | What is the best way to see what things on the iso still need porting to python 3? | 22:15 |
=== lionel_ is now known as lionel | ||
=== sam113101_afk is now known as sam113101 | ||
infinity | Noskcaj: germinate, or walking package dependencies would be the "right" way, but fiddly if you're not sure how. The simple way would be to install ubuntu-desktop and ubuntu-live in a chroot and then "apt-get remove libpython2.7-minimal" and see what goes with it. :P | 22:36 |
cjwatson | soren: man-db does that very thing (although with RUNPATH, not RPATH - which means it remains overrideable using LD_LIBRARY_PATH) for its private libraries. | 22:48 |
cjwatson | (Not sure my last attempt at that got through.) | 22:48 |
ersi | Noskcaj: If you start pulling in that mess, feel *very* free to write up findings. I got "Help with porting to python3" on my TODO (albeit not at the top spots) | 23:07 |
=== freeflying is now known as freeflying_away | ||
=== freeflying_away is now known as freeflying |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!