glibc-2.19.90-15.fc21 had broken the build root on aarch64 The following errors happen when setting up the buildroot when glibc-2.19.90-15.fc21 is tagged in: http://arm.koji.fedoraproject.org//work/tasks/7464/2317464/root.log DEBUG util.py:282: Transaction Summary DEBUG util.py:282: ================================================================================ DEBUG util.py:282: Install 24 Packages (+135 Dependent packages) DEBUG util.py:282: Total download size: 106 M DEBUG util.py:282: Installed size: 410 M DEBUG util.py:282: warning: /etc/hosts created as /etc/hosts.rpmnew DEBUG util.py:282: warning: %post(shared-mime-info-1.2-2.fc21.aarch64) scriptlet failed, exit status 1 DEBUG util.py:282: Non-fatal POSTIN scriptlet failure in rpm package shared-mime-info-1.2-2.fc21.aarch64 DEBUG util.py:282: /usr/bin/grep: write error DEBUG util.py:282: basename: write error DEBUG util.py:282: tr: write error DEBUG util.py:282: sort: close failed: /tmp/sortqXmKgP DEBUG util.py:282: groupadd: failure while writing changes to /etc/group DEBUG util.py:282: groupadd: failure while writing changes to /etc/group DEBUG util.py:282: warning: group utempter does not exist - using root DEBUG util.py:282: warning: group utmp does not exist - using root DEBUG util.py:282: Installed: DEBUG util.py:282: bash.aarch64 0:4.3.11-1.fc21 bzip2.aarch64 0:1.0.6-11.fc21 DEBUG util.py:282: coreutils.aarch64 0:8.22-14.fc21 cpio.aarch64 0:2.11-26.fc21 DEBUG util.py:282: diffutils.aarch64 0:3.3-6.fc21 fedora-release.noarch 0:21-0.6 DEBUG util.py:282: findutils.aarch64 1:4.5.12-2.fc21 gawk.aarch64 0:4.1.1-2.fc21 DEBUG util.py:282: gcc.aarch64 0:4.9.0-4.fc21 gcc-c++.aarch64 0:4.9.0-4.fc21 DEBUG util.py:282: grep.aarch64 0:2.18-1.fc21 gzip.aarch64 0:1.6-2.fc21 DEBUG util.py:282: info.aarch64 0:5.2-2.fc21 make.aarch64 1:3.82-19.fc21 DEBUG util.py:282: patch.aarch64 0:2.7.1-7.fc21 redhat-rpm-config.noarch 0:23-1.fc21 DEBUG util.py:282: rpm-build.aarch64 0:4.11.2-10.fc21 sed.aarch64 0:4.2.2-6.fc21 DEBUG util.py:282: shadow-utils.aarch64 2:4.1.5.1-10.fc21 tar.aarch64 2:1.27.1-3.fc21 DEBUG util.py:282: unzip.aarch64 0:6.0-12.fc21 util-linux.aarch64 0:2.24.2-4.fc21 DEBUG util.py:282: which.aarch64 0:2.20-6.fc21 xz.aarch64 0:5.1.2-8alpha.fc21 DEBUG util.py:282: Dependency Installed:
(In reply to Peter Robinson from comment #0) > glibc-2.19.90-15.fc21 had broken the build root on aarch64 The -15 update is a pull of new upstream changes between April 29th and May 15th. Can you confirm that -14 works? Is the mock chroot completely broken or can you login to it? Anything you can do to narrow down the failure helps.
-14 is fine and once we untagged -15 all is back to building OK. I don't have access to the builders to be able to login to chroot, but as chroot creation doesn't complete I'm not sure how usable it would be
(In reply to Peter Robinson from comment #2) > -14 is fine and once we untagged -15 all is back to building OK. > > I don't have access to the builders to be able to login to chroot, but as > chroot creation doesn't complete I'm not sure how usable it would be Alright I'm building the latest upstream glibc and I'll see what the testsuite does.
I think this is fixed, but I have no idea why... testing now.
I'm unable to reproduce this after backporting a patch from master. Building now (and it should go away in the next bi-weekly rebase?) --Kyle
This turns out to have been caused by: commit 2fbdf5339aaeaa423bd661524f7ec7af41fa2606 Author: Roland McGrath <roland.com> Date: Wed May 14 11:06:36 2014 -0700 Consolidate not-cancel.h files. which resulted in some changed syscall definitions in libio, resulting in this weird bug. Fun. Anyway, fixing the notcancel syscall prototype resolves it. (I also did a -17 build which has some upstream-candidate TLS fixes as well.) regards, Kyle
I've rebased rawhide now.