Description of problem: After updating nfs-utils, the service does not start anymore, because bz 1115179 introduces an error in the application logic of exportfs. (Failins when reading an empty /etc/exports - which is perfectly ok) Version-Release number of selected component (if applicable): How reproducible: always Steps to Reproduce: 1. 2. 3. Actual results: nfs-server fails to start, because ExecStartPre=/usr/sbin/exportfs -r fails with exit-code 1 Expected results: nfs-server should start even if one of the exports files is empty. Additional info: When looking at the diff of bz 1115179, it becomes clear that the newly added lines: ++ int volumes = 0; and ++ if (volumes == 0) ++ xlog(L_ERROR, "No file systems exported!"); introduce a new (wrong) behavior (failing with exit code 1 if an exports file does not result in any exports, because it is either empty or only consists of commented lines) - Please note: Invoking xlog(L_ERROR,...) does not only log but also results in an exit(1) because it sets export_errno to 1! However: An empty or commented exports file is perfectly legal! This situation happens for example, if one adheres to the recommended use by *not* adding export definitions to the default file /etc/exports (provided by setup-2.8.71-2.fc20.noarch) but adding new .exports files below /etc/exports.d/. => Instead of invoking xlog(L_ERROR,...) this should probably reduced to a warning (which does not result in an exit of the tool). Furthermore the log message is *not* very helful unless the file name in question is reported as well... CU -Fritz
Created attachment 957698 [details] Patch to fix the behavior (Use as patch004 on top of nfs-utils-1.3.0-exportfs-nofail.patch)
BTW: The commit which I'm talking about in my original report is this one: http://pkgs.fedoraproject.org/cgit/nfs-utils.git/commit/?h=f20&id=2e682e7bcc96558fa2b3e0a611d6093a630b87e3
*** This bug has been marked as a duplicate of bug 1164477 ***