Hello,
>VMMR0.r0 has been built (presumably due to the recent binutils upgrades in >unstable) with the new PLT32 relocation type, which the virtualbox ELF >relocation >code cannot handle at the moment. Having a quick glance at .text >and .rela.text of VMMR0.r0, it seems to me that PC32 and PLT32 can be handled >identically, >similar to what commit b21ebf2fb4cd of the Linux kernel did? I really liked the analysis, the linux commit and the binutils stuff, I crafted a patch based on this comment, and uploaded in unstable If anybody wants to test the "fix", please grab the deb files from there http://debomatic-amd64.debian.net/distribution#unstable/virtualbox/5.2.14-dfsg-2/buildlog or wait some hours for the package to appear on unstable repositories/mirrors. I'm not sure if this causes regressions or whatever else, the assumption "they might behave in the same way" might not apply as it did in the kernel, even if my basic tests confirmed the VM to start. Use at your own risk, like everything that comes from unstable, the urgency is set to low, so I presume we will have 10 days of people testing this before going in buster. thanks you all, I also submitted the patch to upstream mail list and irc for review G.
