| Summary: | Using sstate prebuilds on another host fails with eglibc-initial do_install error | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] Meta-yocto | Reporter: | Scott Garman <scott.a.garman> | ||||||
| Component: | meta-yocto | Assignee: | Richard Purdie <richard.purdie> | ||||||
| Status: | RESOLVED FIXED | QA Contact: | |||||||
| Severity: | normal | ||||||||
| Priority: | Medium | CC: | poky.bs.watcher, poky.watcher, sgw | ||||||
| Version: | 0.9 | ||||||||
| Target Milestone: | 1.0 PointRelease | ||||||||
| Hardware: | x86 | ||||||||
| OS: | Multiple | ||||||||
| Whiteboard: | |||||||||
| OS type for building Yocto: | --- | Type of Regression: | --- | ||||||
| Verified: | Documentation change: | --- | |||||||
| Attachments: |
|
||||||||
|
Description
Scott Garman
2011-03-02 16:20:59 UTC
Created attachment 102 [details]
do_install log for eglibc-initial
Created attachment 103 [details]
local.conf used during build
I am preserving this build in its current state in case I can provide more information. I encountered this problem a week or so ago and seem to recall that cleaning out the toolchain and rebuilding still resulted in failures. I can re-verify that if needed. A quick brain dump on where I'm at with sstate issues. I've pushed two fixes which vastly improve this situation: http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=16e1f1051eeb51e88244e550bfcf14a1a6ce50d1 http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=273cfeccb7715890be645112966b6a7c5f35ec07 however, if I run a build, "bitbake eglibc -c compile", then move tmp to tmp-moved (i.e. rename it out the way), change to a different build directory, copy the sstate over and run "bitbake eglibc -c compile", then run "bitbake poky-image-minimal" I see a failure in eglibc do_install with "i586-poky-linux-gcc not found". Originally, gcc-cross-intermediate was installed from sstate. As some of its tasks are missing (license task?), it gets rerun, its do_configure uninstalls the populate_sysroot output and this happens whilst the eglibc do_install is running, hence the missing compiler error. Why? Some logic in bitbake allowing the populate_sysroot stamp to have effect, even when the task was going to get rerun anyway? Sounds like we're missing some logic somewhere... I have a fix for this in the paule/sstate-fix2a contrib branch. I think it needs some review and more testing but it does seem to fix the issue. Fixed in master in http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=b4268c08c350a7928a0b1a041b04ffe5a44e77b4 (thanks to Paul!) |