| Summary: | [Autobuilder] Some toolchains are built twice with 20130604-5 build | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] OE-Core | Reporter: | Yi Zhao <yi.zhao> |
| Component: | devtools / tool chain | Assignee: | Beth Flanagan <elizabeth.flanagan> |
| Status: | VERIFIED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | Medium | CC: | elizabeth.flanagan, infras.ab.watcher, Infras.watcher, jessica.zhang, ke.zou, meta.mr.watcher, meta.watcher, ning.zhang, otavio, sgw, yi.zhao |
| Version: | 1.5 | ||
| Target Milestone: | 1.5 M2 | ||
| Hardware: | x86 | ||
| OS: | Multiple | ||
| Whiteboard: | |||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Documentation change: | Don't know | |
|
Description
Yi Zhao
2013-06-06 09:50:56 UTC
The tarball/archives should be named differently to differentiate the different target architecture toolchains For the 20130618 build, there are 2 powerpc toolchains in http://autobuilder.yoctoproject.org/pub/nightly/20130618-2/toolchain/ . Here is the root cause for this issue: Currently on autobuilder, the toolchain tarballs are also built in nightly-fsl-arm and nightly-fsl-ppc. Although they have same archs with nightly-arm and nightly-ppc, the target machines are different. e.g. for arm, the target machines are armv5te (qemuarm on nightly-arm) and armv7a-vfp-neon (imx53qsb on nightly-fsl-arm). For ppc, the target machines are ppc603e (qemuppc on nightly-ppc) and ppce500v2 (p1022ds on nightly-fsl-ppc). Unfortunately, for each arch, the toolchain tarballs have same name, e.g. poky-eglibc-x86_64-powerpc-toolchain-gmae-1.4+snapshot-DATE.sh. If the DATE is different, there will be 2 toolchains tarball in public directory. If the DATE is same (the 2 builds finish on the same day), the one will cover another one when publishing artifacts. We don't perform test with toolchains for nightly-fsl-ppc and nightly-fsl-arm now, so I suggest we can disable toolchain build on nightly-fsl-arm and nigthly-fsl-ppc as a quick fix. Disabling them is only a temp fix. The real solution here is to fix
SDK_NAME = "${SDK_NAME_PREFIX}-${SDK_ARCH}-${TUNE_PKGARCH}"
to something a little bit more descriptive like:
conf/bitbake.conf:SDK_NAME = "${SDK_NAME_PREFIX}-${SDK_ARCH}-${MACHINE}-${TUNE_PKGARCH}"
This however I'm sure will have some cascading impacts. For now, I'm going to disable the fsl toolchain builds to keep them from tromping on the sdk, but this needs a better fix than that.
The arm target arch is still incorrect in 20130710-1 build. It should be "armv5te", but now it is "armv7a-vfp-neon". Toolchain sdk location: http://autobuilder.yoctoproject.org/pub/nightly/20130710-1/toolchain/ Hi, This issue still happens on recent builds. Do we have any solution ? Yes, there is a patch in testing that changes SDK_NAME. This is resolved with commit: http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=b2d8f69089ad1b243b69645f07eeeea8d5a271b7 Verified with 1.5 M4 RC2. Git rev: master/5745e45b18e5099e94b4d5a73bc97dc6d4cdc91f |