impossible to believe, but I tracked this change down to be responible for the llvm3_8 build failure; go figure.

as a summary: due to this change, we now pull in systemd-mini into the buldroot (system-group-trusted:systemd); this then triggers via tmpfiles.d to createa link in /etc/os-release -> ../usr/lib/os-release which in turn makes llvm stumble over an os-release that 'exists' but can't be opened; is is happy to ignore it's absence, but is not happy when it's there but can't be read

@namtrac Can you please see the rest of the comments on this submission? I think the correct thing would be for llvm to handle the case of /etc/os-release being a dangling link the same way as /etc/os-release being absent completely (it does not mind the file being missing, but minds it being a dangling symlink at the moment)

Dominique Leuenberger (dimstar) - over 7 years ago

reviewer

An easy way out would be: BuildRequires: distribution-release (will bring you dummy-release) and thus a real /etc/os-release file to inspect

so i can keep it as-is in permissions? ;)

Dominique Leuenberger (dimstar) - over 7 years ago

reviewer

zes, I don't think a change in permissions makes any sense - llvm3_8 would be falling flat on the nose for whatever change that happens to pull in systemd into it's build root at any moment when there is no /etc/os-release (come to think: that should break on SLE all the time, as there systemd is in every buildroot - but likely also a package adding a /etc/os-release then)

---

Ludwig Nussel (lnussel) - over 7 years ago

source maintainer

we need BuildIgnore here. permissions is in ring0 but system-users are not

---

Max Lin (mlin7442) - over 7 years ago

good catch!

---

Dominique Leuenberger (dimstar) - over 7 years ago

reviewer

sr#501680