/srv/irclogs.ubuntu.com/2012/04/02/#ubuntu-arm.txt

=== Lopi_ is now known as Lopi
=== etienne51__ is now known as etienne51
=== doko_ is now known as doko
jadahlwhen I try to install libc6:armel (on my x86_64-install) it says it depends on tzdata:armel. there is only tzdata:all, and it seems not being able to install because of this. to install libc and other libraries for cross compilation to ARM, is there other way of doing it? using xapt keeps forcing me to apt-get -f install removing g++ so not sure how to move forward with that tool12:36
tedglool, Morning, heard that you might know something about valgrind on ARM? :-)13:43
tedglool, We've got it failing to link on the dbusmenu build and I'm trying to figure out what's up.13:43
tedghttps://launchpadlibrarian.net/98928961/buildlog_ubuntu-precise-armhf.libdbusmenu_0.5.96-0ubuntu1_FAILEDTOBUILD.txt.gz13:43
tedglool, Sorry, xchat crashed.  If you responded, I missed it :-/13:45
gildeantedg: he didn't13:54
looltedg: oy13:58
tedglool, Heh, might thoughts exactly :-)13:58
tedgmy13:58
looltedg: I find it weird that it's mentioning /usr/lib/debug/usr/lib/arm-linux-gnueabihf14:01
loolld shouldn't be poking there?!14:01
tedglool, Could it be something funky with the pc file?  Do you have an ARM system handy?14:01
tedgHere's the x86 pkg-config14:02
tedg$ pkg-config --cflags --libs valgrind14:02
tedg-I/usr/include/valgrind  -L/usr/lib/valgrind -lcoregrind-amd64-linux -lvex-amd64-linux -lgcc14:02
looltedg: Why do you actually need to link to valgrind?14:02
loolI dont see any valgrind include in the testapp14:02
tedglool, It has a couple macros so that the startup/shutdown doesn't end up in the output.14:02
tedglool, It's just including callgrind.h which is all one lib14:03
looltedg: I'm not familiar with linking to valgrind, but from what I can tell, the build is failing to link tools/testapp/ which doesn't include callgrind.h or any valgrind header14:04
loolI've only run valgrind as a standalone tool myself14:04
tedglool, Well, it sets up the pkgconfig setup for all the tests, they aren't done independently.14:05
tedglool, So one of the tests uses it, but all of them get the link because the configure.ac is lazy (or it's authors, perhaps :-) )14:05
loolah14:05
looltedg: It's widespread practice to define the libs only once and then have all objects linked to them; it means you get superfluous lib deps on your binaries which get cleaned up after the fact by --as-needed; but that's unrelated, I guess this works on x86 and you expect it to work on ARM14:06
tedglool, If it's helpful, the C file that uses the lib is test-json-server.c14:07
looltedg: You could open a bug against valgrind in Ubuntu and/or upstream with a reduced testcase and poke linaro-toolchain@lists.linaro.org about it; the main upstream maintainer of valgrind is also on IRC and did the ARMv7 port himself, maybe we will help14:07
looltedg: "sewardj"; I see he is connected to another IRC network at the moment14:08
looltedg: I don't think the problem is with the use of valgrind that you have; it seems to be with valgrind object files on ARM when used for useless linking with random binaries (testapp doesn't use any valgrind foo)14:09
tedglool, Okay, I'm just worried about hard freeze at this point.  I need to figure out if we need to drop the test in the packaging or we can fix it.14:09
looltedg: You could try building the testcase and see whether *that* passes, if it does a workaround is to split valgrind CFLAGS and LIBS into their own VALGRIND_CFLAGS and VALGRIND_LIBS and only list them in LIBADD/LDADD where it actually matters (test-json-server.c)14:10
looltedg: Or if you don't have time to try that out, make valgrind optional and don't list that test and that package on ARM (detect valgrind availability and pass --disable-valgrind if DEB_HOST_GNU_CPU is arm)14:12
looltedg: I don't have good understanding of ELF on ARM or other valgrind subtelties to resolve this, but linaro-toolchain@ would likely be able to advise14:12
tedglool, What's the name of the arm porter box?  Do we still have one?14:12
tedglool, I can try a couple things there.14:13
looltedg: We do have one; I believe there are now DNS aliases to connect to them14:13
looltedg: porter-armhf.canonical.com14:13
looltedg: it seems down though, ping IS?  I can telnet other porter boxes on port 22, but not the armel/armhf ones14:16
tedglool, Okay, I will, sewardj might have a fix here...14:19
=== yofel_ is now known as yofel
=== zyga is now known as zyga-food
=== zyga-food is now known as zyga
=== calculu5 is now known as calculus
=== Ursinha` is now known as Ursinha
=== Ursinha is now known as Guest60849
=== Guest60849 is now known as Ursula___
=== Ursula___ is now known as Ursinha
=== prpplague^3 is now known as prpplague
ppisatiGrueMaster: you around?18:18
GrueMasterFor the rest of the week.  sup?18:19
ppisatiGrueMaster: cool, wait, i have a kernel for you18:19
GrueMasterFor...what platform?18:19
ppisatipanda18:20
GrueMasterI'm not actively testing panda any more.18:20
ppisatiah18:20
GrueMasterMy focus atm is arm server only.18:21
ppisatik18:21
GrueMaster(and I'm not sure how much longer that will be).18:21
ppisatiGrueMaster: but is you panda connected? i need a boot only test18:22
ppisatiGrueMaster: or can you tell me if anyone is doing any testing at all on panda?18:22
GrueMasterppisati: I was told all arm QA was being taken over by the QA team.  ask them.18:23
=== zyga is now known as zyga-afk
MartynHas anyone seen David Mandala in the last day?20:16
GrueMasterMartyn: Try "ping davidm".20:19
=== stgraber_ is now known as stgrber
=== stgrber is now known as stgraber
=== nslu2-log_ is now known as nslu2-log
=== infinity1 is now known as infinity

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!