> targetroot: directory block of inode 0xd3eb extends across I/O block
boundary, which shouldn't happen!
I see the same messages in the logs on the TNF test system, on both
amd64 and i386. They appeared quite recently, with the following commit:
Yes, I added that message. I believe it is harmless, but I put it in
to replace a comment from dholland that something shouldn't happen.
It could be that I made a mistake in my refactoring of ufs_rename, or
it could be that what shouldn't happen does happen and both dholland
and I need to look closer at ufs_direnter.
The quickest way to analyze this, I suppose, would be for me to revert
my changes to ufs_rename, just change the comment to a printf, and see
whether I can still elicit the message -- if so, it's almost certainly
harmless. I will try to find the time to do this.
In the mean time, does the message cause a problem? If so, I can just
change it back to a comment until we've analyzed this.