[RFC] 4950302: (fs spec) Random write produces different results on Linux vs Windows from same .class

[David Lloyd]

On Wed, Nov 20, 2019 at 5:09 PM Brian Burkhalter
<brian.burkhalter at oracle.com> wrote:
>
>
> On Nov 19, 2019, at 4:06 PM, Brian Burkhalter <brian.burkhalter at oracle.com> wrote:
>
> One possible fix is [3] where if O_APPEND is set, it is unset to make the pwrite() call and then reset. This of course could be problematic if another thread were writing to the same file descriptor simultaneously: not all uses of IOUtil.write() use exclusion locks.
>
>
>
> On Nov 20, 2019, at 7:01 AM, Alan Bateman <Alan.Bateman at oracle.com> wrote:
>
> Toggling the mode should be okay if a workaround is really needed but it has to be coordinated in the FileChannel implementation as that it can be coordinated with other operations. That will also avoid putting a potentially costly check in the positional write method too.
>
>
> I’ve been looking at this some more, and while it might be possible to coordinate accesses to accommodate toggling the mode, it’s not clear that it is worth introducing so much more complexity and ugliness for this edge case. Not sure yet how difficult it would be to introduce per-write append with pwritev2 + RWF_APPEND. As mentioned in the thread this would have to fall back to existing behavior where it is not yet supported.

Would it be a good idea to mention that in the API doc then, as an
impl. note or something?  I think this is something that people should
be aware of before trying it.  A cursory search of StackOverflow
indicates that people tend to discover this "the hard way",
unfortunately.
-- 
- DML