Messages in this thread | | | Date | Wed, 24 Mar 2010 18:06:22 +0000 | From | Al Viro <> | Subject | Re: [pnfs] [GIT BISECT] first bad commit: 1f36f774 Switch !O_CREAT case to use of do_last() |
| |
On Wed, Mar 24, 2010 at 07:58:00PM +0200, Boaz Harrosh wrote: > Al hi > > Would you like to attempt a revert of this patch (or group of patches) > Just to get rid of the thought that git bisect was just peeking the > wrong guy. Maybe it's just something else? Can you understand the > relevance of all this?
If you see breakage at that commit and do not see it on its parent, we do have the right guy...
As for reverting, try reverting 781b16775ba0bb55fac0e1757bf0bd87c8879632 first, then this commit.
How consistent are the effects you are seeing from test to test on the same kernel? This one was very interesting, since it seemed to fail with -EISDIR while opening .git/objects/pack. Which is a directory and which should fail with -EISDIR if and only if we pass O_CREAT to open(). And passing O_CREAT on that one is probably not an intended behaviour of git...
Does anybody else see NFS breakage starting at that commit, BTW? Other testcases would be useful...
| |