| Summary: | v86d fails to build on arm64 host | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] OE-Core | Reporter: | Ross Burton <ross.burton> |
| Component: | core | Assignee: | Unassigned <unassigned> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | enhancement | ||
| Priority: | Low | CC: | fathi.boudra, koen, meta.mr.watcher, meta.watcher, ndec13, sgw |
| Version: | 2.1 | ||
| Target Milestone: | Future | ||
| Hardware: | x86 | ||
| OS: | arm64 | ||
| Whiteboard: | |||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Documentation change: | No (bug/feature does not impact docs) | |
|
Description
Ross Burton
2016-07-07 10:08:39 UTC
This is the v86d do_configure log: DEBUG: Executing shell function do_configure It looks like your architecture 'aarch64' isn't supported by this version of v86d. Unrecognized setting x86emu. looking at the configure script for v86d, it uses 'uname -m' to check for the target arch.. in other words, it doesn't cope very well with cross compilation..
in fact, it looks like even today building on x86 or x86-64 will yield to different results.. independently of the x86 MACHINE we are building for..
e.g.
copt_x86emu_test()
{
local m=`uname -m`
if [ "$m" = "i686" -o "$m" = "i586" -o "$m" = "i486" -o "$m" = "i386" ]; then
echo "n";
elif [ "$m" = "x86_64" ]; then
echo "y";
else
echo "It looks like your architecture '$m' isn't supported by this version of v86d." >&2
exit 1
fi
}
Yes, v86d's default behaviour and cross compiling is terrible and is a clear example of why autotools are great. Luckily we bypass the autodetection logic and force-enable x86emu, so this autodetection logic just has to execute without aborting and nothing else. Khem has already sent a patch for this, changing state. Merged in oe-core 709c603dec19690bcf21034867b815b5679b921a. |