build.log and output of emerge --info attached. qemu version: 0.12.1, set up as documented on http://slonopotamus.org/gentoo-on-n8x0/qemu I'd attach coredump but have no idea where it is created.
Created attachment 216188 [details] emerge --info
Created attachment 216190 [details] build.log
CC-ing arm on this one, armin76 any thoughts?
Well no clue...could be qemu's fault, could be sandbox... With FEATURES="-sandbox" the user says it works...he also said he's going to try different qemu versions, so we'll see...
Breaks same way under 0.12.2. Tried 0.11.1, got something different: malloc() ../../sandbox-1.6/libsandbox/libsandbox.c:resolve_path():185: malloc(8192) failed: Unknown error 4096 qemu: uncaught target signal 6 (Aborted) - core dumped rsync: writefd_unbuffered failed to write 4 bytes to socket [sender]: Broken pipe (32) rsync: connection unexpectedly closed (815320 bytes received so far) [sender] rsync error: error in rsync protocol data stream (code 12) at io.c(600) [sender=3.0.6
This may be a sandbox or qemu problem.
chromium needs to run qemu to build ? seems kind of lame. does it only fail with the live scm version ?
(In reply to comment #7) > chromium needs to run qemu to build ? seems kind of lame. > > does it only fail with the live scm version ? > Maybe i wasn't specific enough... I'm running armv6j gentoo stage in qemu on x86_64 box. Only live fails because other versions do not use rsync. Disabling sandbox allows chromium to build. That means one of: 1. qemu is broken 2. rsync is broken 3. sandbox is broken 4. other I happily built lots of other stuff ( http://slonopotamus.org/Packages ) in that setup (and regularly sync portage tree) without any problems. Tried to manually run rsync inside sandbox (sandbox rsync -r /usr/portage/distfiles/svn-src/chromium/src/ /tmp/chromium/), got following (two attempts): ISE rename(printing/.svn/text-base/.printing_context_linux.cc.svn-base.vFnXvG): Bad address abs_path: (null) res_path: /tmp/chromium/printing/.svn/text-base/.printing_context_linux.cc.svn-base.vFnXvG /usr/lib/libsandbox.so[0x420952d8] qemu: uncaught target signal 6 (Aborted) - core dumped rsync: writefd_unbuffered failed to write 4 bytes to socket [sender]: Broken pipe (32) rsync: connection unexpectedly closed (853301 bytes received so far) [sender] rsync error: error in rsync protocol data stream (code 12) at io.c(600) [sender=3.0.6] More additional info: it fails on same file with fixed set of files (i.e. without doing svn up). And more: when running in sandbox, rsync process slowly grows in size up to ~220MB VIRT / 30MB RES, while rsync without sandbox runs in almost constant memory at ~190MB VIRT / 8MB RES.
Kicking this to sandbox. If you don't have any idea what causes the problem, please kick to qemu.
Marat, it would be interesting to see if that can be reproduced with sys-apps/sandbox-2.2. Could you try?
Marat, please re-open if sandbox-2.2 doesn't solve the problem.
Sorry for a long delay, was kinda busy last time :) # equery l sandbox [ Searching for package 'sandbox' in all categories among: ] * installed packages [I--] [ ] sys-apps/sandbox-2.4 (0) Found even easier way to reproduce - install qemu, download gentoo armv6j stage3, setup chroot as documented at https://github.com/slonopotamus/n8x0-overlay/wiki/qemu , try `emerge gcc` (fails during src_unpack, but happily compiles gcc with FEATURES=-sandbox, 100% reproduceable): >>> Unpacking source... >>> Unpacking gcc-4.4.5.tar.bz2 to /var/tmp/portage/sys-devel/gcc-4.4.5/work malloc() ../../sandbox-2.4/libsandbox/libsandbox.c:resolve_path():263: malloc(8192) failed: Cannot allocate memory qemu: uncaught target signal 6 (Aborted) - core dumped /usr/lib/portage/bin/ebuild.sh: line 363: 3612 Broken pipe $1 -c -- "$srcdir$x" 3613 Aborted | tar xof - * ERROR: sys-devel/gcc-4.4.5 failed: * failure unpacking gcc-4.4.5.tar.bz2 Tried different qemu versions (qemu-0.11.1 qemu-0.12.0 qemu-0.12.4 qemu-0.12.5 qemu-0.13.0 qemu-0.14.0)
(In reply to comment #12) > Found even easier way to reproduce - install qemu, download gentoo armv6j > stage3, setup chroot as documented at > https://github.com/slonopotamus/n8x0-overlay/wiki/qemu , try `emerge gcc` > (fails during src_unpack, but happily compiles gcc with FEATURES=-sandbox, 100% > reproduceable) Okay, Chromium team has nothing to do here then. Thank you for finding shorter steps to reproduce.
Updated summary
i'm not going to investigate this at all considering it works on real hardware
I've a very similar problem emerging chromium on x86 (this is a thinkpad x200, real hw ;) build.log and output of emerge --info attached. When malloc() fails I get this on dmesg: make[2813] general protection ip:b761bc75 sp:bfe3b6b8 error:0 in libc-2.11.3.so[b75f0000+141000] Emerging chromium with FEATURES="-sandbox" works fine. I must also say that I'm starting to suspect that sandbox is triggering a kernel bug because I'm suffering from hard lockups (only!) while compiling with sandbox enabled with both vanilla 2.6.37 and 2.6.38. It's not due to overheating or high system load, it freezes even using a single core at the lowest speed with MAKEOPTS="-j1"
Created attachment 272877 [details] build.log of chromium
Created attachment 272879 [details] emerge --info
(In reply to comment #16) > I've a very similar problem emerging chromium on x86 (this is a thinkpad x200, > real hw ;) build.log and output of emerge --info attached. When malloc() fails > I get this on dmesg: > > make[2813] general protection ip:b761bc75 sp:bfe3b6b8 error:0 in > libc-2.11.3.so[b75f0000+141000] > > Emerging chromium with FEATURES="-sandbox" works fine. > > [...] Just an additional information: I've tried different vanilla sources from 2.6.34 to 2.6.38, always getting the same error
(In reply to comment #19) > > [...] > > Just an additional information: I've tried different vanilla sources from > 2.6.34 to 2.6.38, always getting the same error More infos, I tried every version of sandbox in portage. The chromium problem occurs with sandbox-2.3-r1, sandbox-2.4 and sandbox-2.5. Everything works fine using sandbox-1.6-r2.
*** Bug 648516 has been marked as a duplicate of this bug. ***
qemu & sandbox have changed significantly in the last decade. see if it still fails nowadays.