On Wed, Apr 9, 2014 at 1:43 PM, Uwe Schindler <[email protected]> wrote:
> Hi,
>> However, I think we should back-port LUCENE-5570 to 4.7.2, especially given
>> that 4.7.2 is the last release Java 1.6 users can use, I do think it's 
>> important to
>> fix bugs there.
>>
>> Uwe, are you very strongly against fixing that issue on 4.7.2?
>
> I am -0 here. This issue just makes debugging harder, but causes no bugs or 
> data corrumption, so we should not add stuff that cannot be solved natively 
> with Java 6. LUCENE-5574 is the real issue, LUCENE-5570 was just confusing to 
> those who tried to understand the bug. But as the bug is fixed, no need to 
> fix the side-effect of 0 byte files.

Wait, are you -0 or -1?  If you are just -0, i.e. you will not veto
Robert's patch to backport to 4.7.2, then I'd like to commit that.

The problem is, this "creates 0 byte file instead of throwing
FNFE/NSFE" bug can mask other (future) Lucene bugs; yes, we fixed
LUCENE-5574, but what about other possible Lucene bugs that fsync the
wrong file?  Hopefully there are none, but if there are, it's much
better to see an exception than silently create 0-byte files which
later manifest as index corruption looking like your filesystem ate
the file not Lucene.

I agree it'd be great if we could do the same change everywhere, but
because 4.7.x is Java 1.6, we can't, and I think the lesser evil is a
slightly different patch (Robert's patch) than just not fixing the bug
at all.

Mike McCandless

http://blog.mikemccandless.com

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to