001
From: Kevin McCarthy
Date: Sun, 26 Aug 2018 18:43:20 -0700
Subject: [PATCH] Add additional error handling to safe_rename().
It is apparently possible for link() to return an error but the link
to still be created. Add a double check for that case. If the files
match, unlink the src and
> From f0faf6cf205e0b96c2fb93ea12393458843aa485 Mon Sep 17 00:00:00 2001
> From: Kevin McCarthy
> Date: Sun, 26 Aug 2018 18:43:20 -0700
> Subject: [PATCH] Add additional error handling to safe_rename().
>
> It is apparently possible for link() to return an error but the
On Tue, Aug 28, 2018 at 03:27:51PM -0500, Derek Martin wrote:
> On Sun, Aug 26, 2018 at 06:48:46PM -0700, Kevin J. McCarthy wrote:
> > Here's a first try at a patch to check for the historically documented
> > "link() returning -1 but succeeding" case for NFS.
>
> This seems fine.
Thank you for t
On Tue, Aug 28, 2018 at 06:08:46PM -0700, Kevin J. McCarthy wrote:
> > Though, it seems it leaves in the stat comparison if
> > link() succeeds, which is superfluous (could cause needless
> > performance hit on slow networked filesystems), and violates KISS and
> > "Don't write code you don't need"
On Wed, Aug 29, 2018 at 12:16:14PM -0500, Derek Martin wrote:
> So, I think the post-success stat() calls should should come out, and
> I wouldn't bother using #if to remove it. I think it's clearly just a
> bad idea to do that check there, and we should remove any trace of it
> so that no one is