With recent deletion of chrome-sandbox from skypeforlinux distribution, skypeforlinux does not start if appropriate options are not enabled in the kernel configuration. Thus install should abort after finding that kernel is not configured properly rather than proceed with installing package that does not work, possibly removing the previously installed and working version.
Which exact kernel options?
I had this problem as well, searched all over for how to solve it by choosing the right kernel options, but what actually fixed it for me was to update to the newest stable kernel. Everything started working again after that (old kernel not working was 4.14.65, update to 5.4.28 fixed it without changing any kernel modules) Hope this helps...
Aborting on missing kernels is generally not an option as it excludes the possibility of people running binary kernels.
(In reply to Jonas Stein from comment #1) > Which exact kernel options? USER_NS on many (older ?) kernels it was not set be default, and often used workaround was to chmod 4755 /opt/skypeforlinux/sandbox instead. Such installations become inoperational after the update, until kernel is recompiled, which may not be immediately possible.
(In reply to David Seifert from comment #3) > Aborting on missing kernels is generally not an option as it excludes the > possibility of people running binary kernels. I understand if ebuild cannot detect what options kernel was compiled with and whether it has the required USER_NS set, which may be the case with binary kernels. But why to proceed with installing the package that will not work when ebuild did it job and has detected that USER_NS is missing ? I do not understand the logic.
(In reply to pogosyan from comment #5) > (In reply to David Seifert from comment #3) > > Aborting on missing kernels is generally not an option as it excludes the > > possibility of people running binary kernels. > > I understand if ebuild cannot detect what options kernel was compiled with > and > whether it has the required USER_NS set, which may be the case with binary > kernels. But why to proceed with installing the package that will not work > when ebuild did it job and has detected that USER_NS is missing ? I do not > understand the logic. Ok so if the ebuild can't detect the kernel options, it should proceed, but if it can detect them, and detect them as missing, it should error out?
