From: Alex Riesen <raa.lkml@gmail.com>
Date: Fri, 24 Apr 2009 05:44:18 GMT
Subject: Re: [PATCH] Add an option not to use link(src, dest) && unlink(src) when that is unreliable
Message-ID: <81b0412b0904232244l15cd7347n23702d92b38ab7e5@mail.gmail.com>
In-Reply-To: <alpine.DEB.1.00.0904232358520.10279@pacific.mpi-cbg.de>

2009/4/23 Johannes Schindelin <Johannes.Schindelin@gmx.de>:
> Hi,
>
> On Thu, 23 Apr 2009, Alex Riesen wrote:
>
>> 2009/4/23 Johannes Schindelin <Johannes.Schindelin@gmx.de>:
>> > -       if (link(tmpfile, filename))
>> > +       if (unreliable_hardlinks)
>> > +               ret = ~EEXIST;
>>
>> It is more like "broken_hardlinks" or even "no_hardlinks"!
>
> Wrong.  As I wrote, single-stepping (i.e. leaving enough time between
> link() and unlink()) works as expected.  So it is not even that the
> hardlinks are broken.  Just the serialization between the operations.
>

Since when does link(2) involve file _data_?