On 07/21/2017 06:56 AM, Kevin Wolf wrote: >> >> Convert to runtime check to cope with that. > > Fair enough in this specific case because we still support older Linux > kernels and we want to fail gracefully if the binary was built against > a newer kernel.
Or, more likely: if we are built against a newer glibc (that has the constants) but an older kernel (that lacks support for the constants). > > However, I think the real problem here is with the WSL ecosystem if qemu > is routinely built against a real Linux while WSL doesn't provide the > same functionality. WSL should provide kernel headers that match what > it can provide (i.e. either remove the unimplemnted constants or > implement them). In other words, you're arguing that binaries built for WSL should be cross-compiled rather than native compiled (similar to how mingw binaries are built in a Cygwin environment), such that the cross-compiler picks up the correct altered headers for WSL limitations. I agree - but I have no influence on how likely that is to come about. > > So for the future, I'm not sure that we should add workarounds for WSL > shortcomings when no real Linux is affected. > > This is even more true when the reporter is a Microsoft employee. > > Kevin > -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
signature.asc
Description: OpenPGP digital signature
