Alan and all, So I pulled in the merges from Alan and some other bits I'm working on, and actually had to remerge because of other stuff. I pull commits one at a time these days so I can review each one fully, then diff the end result to see if I missed anything with your merge attempt since you were kind enough to have a nice merge branch for me (except you'll notice I had a few other fixes, free in different order, etc.). Anyway. The reason I haven't pushed dev to master yet is attached. The tests are falling over because opendir() happily succeeds on the modprobe2.conf *file* in the testsuite and then modprobe gets very unhappy trying to read from it. The same is not true on an F12 box running 2.6.33-rc8. So something horrible is broken upstream :) Assuming I'm not told "oh yea, we're breaking everything you ever knew about opendir() semantics" though some weird file directory streams thingy I should push dev to master later. Jon.
--- Begin Message ---
- Subject: opendir() on a file???
- From: Jon Masters <jonathan@xxxxxxxxxxxxxx>
- Date: Tue, 09 Mar 2010 04:27:00 -0500
- Organization: World Organi[sz]ation of Broken Dreams
Folks, Now I might be missing something, and I know I'm behind on LKML[0], but the following isn't supposed to work in my book: /* * Weird kernel test */ #include <sys/types.h> #include <dirent.h> #include <stdio.h> int main(int argc, char **argv) { DIR *dir; dir = opendir("foo.conf"); if (dir) printf("WTF?\n"); return 0; } This is on an ext4 filesystem, whereas on a box with an older kernel this test correctly does not print "WTF?". I know some filesystems experiment with streams and treating files as directories, etc. but I wasn't aware that anything particular had changed recently? The box is running almost an upstream kernel, and I can poke if I'm told this not intended: 2.6.34-0.8.rc0.git11.fc14.x86_64. What am I missing? Jon. [0] The podcast isn't dead, I'm just suffering from a cold and will be taking a day off to recover and catch up with that sometime today.
--- End Message ---
