Bug#910489: glibc: fakechroot needs glibc 2.28 for renameat2 usage by mv in coreutils

2018-12-04 Thread Aurelien Jarno
Version: 2.28-1

On 2018-10-07 09:30, Johannes 'josch' Schauer wrote:
> Source: glibc
> Severity: normal
> Control: block #909612 by -1
> 
> Hi,
> 
> currently, it is not possible to use fakechroot with mv from coreutils
> in a Debian unstable chroot. Solving this issue requires an update of
> glibc in Debian to at least 2.28.
> 

glibc 2.28 is now in unstable. Closing the bug.

-- 
Aurelien Jarno  GPG: 4096R/1DDD8C9B
aurel...@aurel32.net http://www.aurel32.net



Bug#910489: glibc: fakechroot needs glibc 2.28 for renameat2 usage by mv in coreutils

2018-10-07 Thread Johannes 'josch' Schauer
Source: glibc
Severity: normal
Control: block #909612 by -1

Hi,

currently, it is not possible to use fakechroot with mv from coreutils
in a Debian unstable chroot. Solving this issue requires an update of
glibc in Debian to at least 2.28.

Details:

If the system has the renameat2 syscall (since Linux 3.15), then 'mv'
from coreutils (since 8.30) will directly use syscall(SYS_renameat2,
...) to make use of that system call. This is bad news for fakechroot
because fakechroot cannot intercept directly made syscalls but relies on
glibc wrappers. Since glibc 2.28 such a wrapper was added and there is a
patch against coreutils [1] that makes use of the glibc function instead
of the direct syscall, if available. For this to work, we need glibc
2.28 or otherwise, that change in coreutils will have no effect in
Debian.

Thus, I'm filing this bug to track the situation, blocking the according
bug in fakechroot.

Thanks!

cheers, josch

[1] https://debbugs.gnu.org/cgi/bugreport.cgi?bug=32796