Bug 201378

Summary: Sometimes seeing ENOENT for directories in NFS-mounted ports tree
Product: Base System Reporter: david
Component: kernAssignee: freebsd-bugs (Nobody) <bugs>
Status: New ---    
Severity: Affects Only Me    
Priority: ---    
Version: 10.0-STABLE   
Hardware: Any   
OS: Any   

Description david 2015-07-06 19:32:23 UTC
This is a bit vague, for which I apologize -- in large part, it may be viewed as a pleas for guidance as to what other information or testing I might do to help isolate and identify the problem.

I'm filing this under "kern" largely because of my prior experience in isolating the bug that was addressed by http://svnweb.freebsd.org/base?view=revision&revision=189287 -- and which had scarily similar symptoms.

The page at http://www.catwhisker.org/~david/FreeBSD/convert_i386_amd64.html documents some of my activities where I encountered this issue (about mid-way down the page).  Summarizing:

Several years ago, I placed my SVN working copy of the ports tree on the ReadyNAS at home, for use among 3 systems.  One of these systems used svnlite to update the ports tree daily; each of the three used it for building/updating ports in place (using portmaster(8)).  I never had a problem with this.

More recently (ref. the above-cited Web page), I made an initial foray into using poudriere(8) to build custom packages for my installed ports.  After a few mis-steps, I got it to work for about 10 minutes, at which point I started seeing errors such as "make: chdir /usr/ports/devel/gsettings-desktop-schemas: No such file or directory" -- while an "ls" from the same machine showed the directory.

Eventually, I created a new SVN working copy of the ports tree that was locally-mounted... and that did not exhibit the failure.

From what I have seen so far, the effect is 100% reproducible.

That said, it could be a flaw or limitation in the ReadyNAS, or Something Weird Happened with my network, or ... lots of things, I suppose.

But I seem to have access to an environment where the problem can be re-created at will, and the machine where the problem is most evident is not needed for critical activities (and it's reasonably fast), so I'd like to do what I can to nail this bug down and squash it.