Hi,
On Sun, 13 Jan 2019, Adrian Bunk wrote:
> Test cases that passed in patchelf 0.8 fail since 0.9,
> and segmentation fault on things like setting rpath
> might be close enough to "entirely broken".
In that case, it would certainly help upstream if someone
(maintainer/porter) could try to "git bisect" the issue so that we can
better identify the issue.
And maybe we want to revert to the former patchelf in Debian
(i.e. 0.9+really0.8) if it lasts too long.
The failing tests in 0.9 are new, in that the test is not the same as in 0.8. As I commented in the arch-RM bug, I don't think this is a regression (but I haven't checked at all). James Cowgill mentioned a change in glibc as the possible cause[1]. One way to test would be to run the test from 0.9 with the binary from 0.8.
That said, I certainly welcome help. I'm completely overloaded right now so I can't look at this.