Quoting Johannes Schauer Marin Rodrigues (2026-07-09 16:26:10) > Quoting Johannes Schauer Marin Rodrigues (2026-07-09 11:10:22) > > I observed that the command which fails (for both ucf as well as openssh > > postinst) is a cp carrying the -Z argument. > > > > If you happen to have a hunch which commit between 2.0-1 and 2.1.1-1 > > could've > > broken cp with SELinux I could try reverting that commit and see if that > > fixes > > it. > > looking at the differences between 2.0-1 and 2.1.1-1 I found that this patch > makes the issue go away: > > --- a/libfakeroot.c > +++ b/libfakeroot.c > @@ -2038,8 +2038,8 @@ ssize_t flistxattr(int fd, char *list, size_t size) > > r = next_flistxattr(fd, list, size); > if (fakeroot_disabled > - || r > 0 > - || (r < 0 && errno != EPERM && errno != EOPNOTSUPP)) > + || r >= 0 > + || (errno != EPERM && errno != EOPNOTSUPP)) > return r; > errno = 0;
the patch can be made a bit smaller even:
--- a/libfakeroot.c
+++ b/libfakeroot.c
@@ -2038,7 +2038,7 @@ ssize_t flistxattr(int fd, char *list, size_t size)
r = next_flistxattr(fd, list, size);
if (fakeroot_disabled
- || r > 0
+ || r >= 0
|| (r < 0 && errno != EPERM && errno != EOPNOTSUPP))
return r;
errno = 0;
You might also want to disregard the conditions necessary for the bug to appear
because it's flaky. While testing this I saw that sometimes the same error
would get triggered even without a foreign arch being involved. It just seems
to be reproducible more reliably when a foreign arch is involved for some
reason.
Just to make sure: I have no idea what I'm doing and hope that you can make
some sense out of this. I'm testing the patch in the dpkg-root-demo salsa CI
pipeline of this MR:
https://salsa.debian.org/helmutg/dpkg-root-demo/-/merge_requests/15
Thanks!
cheers, josch
signature.asc
Description: signature
