| slangasek | stgraber: so is that latter sysvinit change cleaning up after previous incomplete sysvinit migration code? | 00:01 |
|---|---|---|
| slangasek | (I suppose that's why it might be in sysvinit instead of lxc?) | 00:01 |
| slangasek | stgraber: also, can you please put a version guard on that upgrade code? | 00:02 |
| stgraber | slangasek: yeah, and lxc doesn't have any package inside the containers, so we don't have a way of fixing them other than SRUing initscripts for that. In theory the check I uploaded shouldn't even match a setup that's not broken but I should have indeed added a version guard on top of that if someome actually wants the "broken" setup. | 00:54 |
| slangasek | stgraber: <nod> the risk of a false positive is small, but I'm paranoid about upgrade code not having version checks; so please fix that in quantal and in the SRU :) | 01:00 |
| === apachelogger___ is now known as apachelogger | ||
| stgraber | slangasek: please reject the SRU then, I'll prepare an updated postinst | 02:01 |
| slangasek | stgraber: rejected, thanks | 02:23 |
| stgraber | slangasek: re-uploaded to -proposed, testing the one to quantal to make sure I didn't mess up anything in the version check | 02:25 |
| === mthaddon` is now known as mthaddon | ||
| psusi | so I'm trying to figure out how to bootstrap an x32 system by practicing building i386 packages from amd64... if I run debuild -ai386, it seems to build an i386 binary package ( though dh_strip tries to run i686-linux-strip instead of strip, so I had to make a symlink ), but I can't figure out how to get debuild to pass gcc -march=i386 | 03:36 |
| slangasek | psusi: so debuild -ai386 will set DEB_HOST_ARCH and friends; packages that know about cross-building will take this as a hint to *do* cross-building, which means they'll look for a cross-compiler | 03:38 |
| slangasek | psusi: so they will look for i686-linux-gnu-gcc - provide this and have it do the right thing | 03:38 |
| psusi | slangasek, right... I guess that's why dh_strip was trying to run i686-linux-strip instead of just strip.... | 03:38 |
| psusi | but the actual build didn't complain about not being able to find i686-linux-gnu-gcc | 03:39 |
| psusi | it seemed to run gcc and build just fine... just barfed at dh_strip, which I got around with a symlink | 03:39 |
| slangasek | ideally i686-linux-gnu-gcc could be implemented by gcc itself just knowing what to do; but at a push, you can make it a short wrapper script that calls exec gcc -m32 $@ | 03:39 |
| slangasek | right, then your test package seems to not be cross-build-friendly | 03:39 |
| slangasek | (which package?) | 03:39 |
| psusi | but... while it built an i386 binary deb, the actual binaries are still amd64 since it isn't telling gcc the right -m | 03:40 |
| scientes | slangasek, yes i tested using multi-arch for that, it works great! | 03:40 |
| slangasek | depending on the package, it may be using gcc as a fallback | 03:40 |
| infinity | psusi: Yeah, it's obviously hardcoding the compiler, which is not cross-friendly. | 03:40 |
| psusi | hrm.... that could be | 03:40 |
| slangasek | and maybe creating i686-linux-gnu-gcc will be enough to make it happy | 03:40 |
| scientes | just using a -m32/-m64 wrapper for -afoo that can be satified by multilib | 03:40 |
| slangasek | again, which package? | 03:40 |
| infinity | Or, I suppose it could be cross-friendly but falling back in the absense of triplet-cc, yeah. | 03:40 |
| scientes | slangasek, you need symlinks for binutils, and gcc/cpp/etc, and then it works | 03:41 |
| slangasek | yes | 03:41 |
| psusi | so... you are saying to get it to work, I have to write a wrapper gcc script to run gcc -m32? can't get dh to be happy just running gcc and adding -m32 to the CFLAGS? | 03:41 |
| scientes | but yeah i had that working great | 03:41 |
| scientes | paulliu, yep, i'll go find what i was using | 03:41 |
| scientes | paulliu, slacker_nl this works: http://paste.ubuntu.com/1102847/ | 03:42 |
| scientes | i called it "gcc-32", and you could easily make a similar gcc-64 | 03:43 |
| psusi | it seems kind of silly to have to invoke all of the cross compiling machinery when the native gcc supports the "cross" arch, you just need to add a flag | 03:43 |
| psusi | but you're saying that's just how you have to do it? | 03:43 |
| scientes | psusi, the "c99" "c89" commands work exactly the same | 03:43 |
| scientes | of course this is called by the full qualified name | 03:43 |
| scientes | i got blocked on trying to wrap my head around how the packaging for gcc-defaults-cross might work | 03:44 |
| scientes | i think it should be packaged like this IMHO | 03:45 |
| psusi | so if you have to pretend you are doing a full cross compile... why doesn't our gcc package come with the cross compiler wrapper script to add -m32? | 03:45 |
| scientes | psusi, it could either be part of (gcc|g++|gobjc)-multilib packages, or a seperate package | 03:46 |
| infinity | psusi: You could certainly file a bug to that effect. | 03:48 |
| psusi | how do you smack bzr bd into passing options to debuild? it pukes on -ai386 | 03:58 |
| infinity | -- perhaps? | 03:58 |
| infinity | (I don't use it, just a wild guess) | 03:58 |
| ScottK | psusi: -- $OPTIONS | 03:59 |
| psusi | good guess... I was thinking that but the man page doesn't show it ;) | 03:59 |
| ScottK | Yes. I've written bugs about it. | 03:59 |
| ScottK | It's not a guess. I knew because I couldn't find it either and had to ask. | 03:59 |
| ScottK | "inconsitent interface" is on my reasons to avoid UDD kinds of things. | 04:00 |
| ScottK | Who'd have thought -S -- -sa -us -uc was a good plan? | 04:00 |
| psusi | hrm... when I add -ai386, it says /usr/include/linux/errno.h can't find asm/errno.h | 04:02 |
| scientes | psusi, install libc6-dev:i386 ? | 04:04 |
| psusi | ohh... I guess gcc expects a separate 32bit headers directory | 04:04 |
| scientes | psusi, or rather apt-get build-dep -ai386 <package> | 04:05 |
| slangasek | ScottK: it's equally valid to call it as -- -S -sa -us -uc, if that makes you happier ;P | 04:12 |
| psusi | hrm.. so why does apt-get want to remove the native amd64 -dev packages to install the i386 ones? | 04:13 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!