| Summary: | Native builds mix headers from | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] OE-Core | Reporter: | Thomas Köller <thomas.koeller> | ||||
| Component: | core | Assignee: | Thomas Köller <thomas.koeller> | ||||
| Status: | RESOLVED FIXED | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | Undecided | CC: | meta.mr.watcher, meta.watcher, randy.macleod, richard.purdie | ||||
| Version: | unspecified | ||||||
| Target Milestone: | --- | ||||||
| Hardware: | x86 | ||||||
| OS: | x86_64 | ||||||
| Whiteboard: | |||||||
| OS type for building Yocto: | --- | Type of Regression: | --- | ||||
| Verified: | Documentation change: | No (bug/feature does not impact docs) | |||||
| Attachments: |
|
||||||
Thomas, can you please provide exact steps to reproduce the problem? Thanks (In reply to comment #1) > Thomas, can you please provide exact steps to reproduce the problem? > Thanks Ok. As I said, I am building on a Fedora 27/x86_64 host (a VM running in VirtualBox under Widows 10, actually): [tkoeller@fedora-vb-tkoeller build-rpb]$ uname -a Linux fedora-vb-tkoeller 4.13.15-300.fc27.x86_64 #1 SMP Tue Nov 21 21:10:22 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux [tkoeller@fedora-vb-tkoeller build-rpb]$ gcc --version gcc (GCC) 7.2.1 20170915 (Red Hat 7.2.1-2) Copyright (C) 2017 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. I am following the build instructions given in https://github.com/96boards/documentation/wiki/DragonBoard-820c-Getting-Started-With-Linux#installing-an-open-embedded-based-image and https://github.com/Linaro/documentation/blob/master/Reference-Platform/CECommon/OE.md I create a new directory named linaro-build, cd into it and do 'repo init' / 'repo sync': repo init -u https://github.com/96boards/oe-rpb-manifest.git -b morty repo sync Next step: [tkoeller@fedora-vb-tkoeller linaro-build]$ MACHINE=dragonboard-820c DISTRO=rpb source setup-environment Welcome to OpenEmbedded Reference Platform Build (OE RPB) For more information about OpenEmbedded see their website: http://www.openembedded.org/ Your build environemnt has been configured with: MACHINE = dragonboard-820c SDKMACHINE = x86_64 DISTRO = rpb You can now run 'bitbake <target>' Some of common targets are: rpb-console-image rpb-desktop-image rpb-weston-image core-image-base core-image-minimal Then, I add a couple of lines to the autogenerated 'conf/auto.conf' file, to control bitbake and make parallel execution, and to enable use of an existing shared download cache. The result looks as follows: [tkoeller@fedora-vb-tkoeller build-rpb]$ cat conf/auto.conf DISTRO ?= "rpb" MACHINE ?= "dragonboard-820c" SDKMACHINE ?= "x86_64" # Extra options that can be changed by the user INHERIT += "rm_work" # +++ enable yocto shared download cache +++ SOURCE_MIRROR_URL ?= "file:///media/sf_ModBizISP/yocto_download_cache" INHERIT += "own-mirrors" BB_GENERATE_MIRROR_TARBALLS = "1" BB_NUMBER_THREADS = "16" PARALLEL_MAKE = "-j8" Finally, I invoke bitbake: [tkoeller@fedora-vb-tkoeller build-rpb]$ bitbake rpb-console-image which then results in the error reported. Thanks for the extra info, that helped a lot. We ran into this issue with morty on Ubuntu1710 as well. This has been fixed with: http://git.yoctoproject.org/cgit.cgi/poky/commit/?h=morty&id=8c84d07063ea0ebe8f335fafc2e619555bdab812 |
Created attachment 4129 [details] Log file of failed compile task I experienced build errors when building native tools on my Fedora 27 x86_64 host, see attached log file. The reason turned out to be that part of the header files used were taken from the directories specified with '-isystem', while others were taken from the build system's standard include path '/usr/include'. This produced an inconsistent mix of header files. Obviously, dependencies were wrong. I then added the following line to my local.conf: BUILD_CPPFLAGS += "-nostdinc" and found that this produced even more errors, already in the 'configure' task of the m4-native recipe. I think that no build step should reference headers from the build system outside oe, and using '-nostdinc' everywhere may be an easy way to detect violations of this rule. Although I did not check, I suspect there may be similar problems for libraries used in the build.