| Summary: | poky-image-sato-live build failure | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] Meta-yocto | Reporter: | Kevin Tian <kevin.tian> |
| Component: | meta-yocto | Assignee: | Mark Hatle <mark.hatle> |
| Status: | VERIFIED FIXED | QA Contact: | |
| Severity: | major | ||
| Priority: | High | CC: | jessica.zhang, kevin.tian, poky.bs.watcher, poky.watcher, richard.purdie, sgw |
| Version: | 0.9 | ||
| Target Milestone: | 1.0 M1 | ||
| Hardware: | x86 | ||
| OS: | Multiple | ||
| Whiteboard: | |||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Program Management | Documentation change: | --- |
|
Description
Kevin Tian
2010-08-23 22:31:51 UTC
I tried to reproduce the issue locally, but couldn't... 1) I built the emenlow poky-image-sato-live from scratch successfully. 2) I built the emenlow poky-image-sdk-live from scratch successfully, later I built poky-image-minimal-live successfully, and finally I could also build poky-image-sato-live successfully. I used a newer git commit: 8 commits newer than that used by the autobuilder: the issue might have been fixed by the commits?? e.g., commit ca95d02ce0dd7e1ab(package.bbclass: Fix shlibs race) by RP. I'll watch the latest nightly build log about this. The netbook/emenlow poky-image-sato-live builds succeed now... The pesudo related commits might fix the issue. ERROR: Task failed: ('function do_rootfs failed', '/srv/home/pokybuild/poky-slave/milestone-release/build/build/tmp/work/netbook-poky-linux/poky-image-minimal-initramfs-1.0-r0/temp/log.do_rootfs.19169')
NOTE: package poky-image-sato-1.0-r0: task do_setscene: Started
NOTE: Removing shared state for package poky-image-sato
NOTE: package poky-image-sato-1.0-r0: task do_setscene: Succeeded
ERROR: Logfile of failure stored in: /srv/home/pokybuild/poky-slave/milestone-release/build/build/tmp/work/netbook-poky-linux/poky-image-minimal-initramfs-1.0-r0/temp/log.do_rootfs.19169
Log data follows:
| + rm -rf /srv/home/pokybuild/poky-slave/milestone-release/build/build/tmp/work/netbook-poky-linux/poky-image-minimal-initramfs-1.0-r0/rootfs
| + mkdir -p /srv/home/pokybuild/poky-slave/milestone-release/build/build/tmp/work/netbook-poky-linux/poky-image-minimal-initramfs-1.0-r0/rootfs
| + mkdir -p /srv/home/pokybuild/poky-slave/milestone-release/build/build/tmp/deploy/images
| + '[' 0 '!=' 1 ']'
| + for devtable in /srv/home/pokybuild/poky-slave/milestone-release/build/meta/files/device_table-minimal.txt
| + makedevs -r /srv/home/pokybuild/poky-slave/milestone-release/build/build/tmp/work/netbook-poky-linux/poky-image-minimal-initramfs-1.0-r0/rootfs -D /srv/home/pokybuild/poky-slave/milestone-release/build/meta/files/device_table-minimal.txt
| makedevs: /srv/home/pokybuild/poky-slave/milestone-release/build/build/tmp/work/netbook-poky-linux/poky-image-minimal-initramfs-1.0-r0/rootfs/dev/hda: file type does not match specified type!
NOTE: package poky-image-minimal-initramfs-1.0-r0: task do_rootfs: Failed
If we can get another reproducer for this, tar up the tmp/work/... directory where the failure occurred, we need to inspect the filesystem and pseudo log files. So far we've been unable to figure out a cause for this problem. Still looking... I have some further details on this bug: a) I was able to reproduce on an ubuntu maverick 64 bit system b) It seems most likely to happen when there are multiple rootfs tasks running in parallel such as when a -live image is built (the main rootfs and initramfs are built in parallel). c) It only seems to happen every other build. d) It is likely that the rm_work class is triggering the problem, perhaps by cleaning the rootfs directory within WORKDIR. I meant to add: e) The bug is fastest to reproduce if the contents of do_rootfs in image.bbclass is removed after the makedevs code block. Its always the makedevs command that triggers the bug. With my established reproducible failing test case, if I stopped rm_work wiping out ${WORKDIR}/rootfs/, the bug stopped happening. I'm pretty sure this is happening due to the removal of files by rm_work outside of pseudo's knowledge.
verified as autobuilder doesn't show this problem now |