lucenenet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Irwin <pir...@feature23.com>
Subject Re: Lucene.net 4.3 failing test case - Merge Background Exception
Date Sun, 06 Oct 2013 23:44:03 GMT
Cool, thanks. Can you push my changes from my branch_4x as well? I've got a
few more commits there.

I'm working with the necessary people to get an apache ID set up so I don't
have to bother you for these commits :-)


On Sun, Oct 6, 2013 at 7:06 PM, mherndon michael <
mherndon@michaelherndon.com> wrote:

> pushed.
>
>
> On Thu, Oct 3, 2013 at 3:51 PM, Paul Irwin <pirwin@feature23.com> wrote:
>
> > Alright cool, thanks. I've got a few extra commits there making progress
> on
> > the QueryParsers assembly that can go too. I'm working on the
> > QueryParsers.Flexible namespace now, might take a few more days for that
> as
> > it's got a lot of classes to port.
> >
> > Paul
> >
> >
> > On Thu, Oct 3, 2013 at 2:49 PM, mherndon michael <
> > mherndon@michaelherndon.com> wrote:
> >
> > > No one has access to the github mirror, thus one still has to apply the
> > > pull request manually.
> > >
> > > Though as others have pointed out, once the pull request is received,
> > > github should acknowledge it.
> > >
> > > I'll merge the commits from your repo this weekend.
> > >
> > >
> > > On Tue, Oct 1, 2013 at 4:58 PM, Paul Irwin <pirwin@feature23.com>
> wrote:
> > >
> > > > I think it would be great to be able to submit a pull request to the
> > > apache
> > > > github mirror, but if that's not possible, someone will have to be
> > > > responsible for the merging. I'll go along with whatever the
> consensus
> > > is.
> > > >
> > > > I think I fixed the FST bug. It was an oversight when porting the
> class
> > > > that there as an ambiguity between a parameter and a field. It now
> > writes
> > > > my full test index correctly. I'll start plugging it into my domain
> app
> > > and
> > > > see if I can really write a full index (Lucene.net 3.0.3 size is
> about
> > > > 400-500GB).
> > > >
> > > > Here's the commit:
> > > >
> > > >
> > >
> >
> https://github.com/paulirwin/lucene.net/commit/8404b339a7018574275f490069b39d9fd4256d14
> > > >
> > > > So I've got 3 commits ready to go into the 4x branch (1 just a minor
> > > > variable name fix for a bad find/replace), once we decide on how to
> do
> > > > them. If I get commit access I suppose I could add the official
> apache
> > > git
> > > > repo as an upstream source and push them in.
> > > >
> > > > Also there's the Chinese work from Athrun Saga, which I haven't yet
> > > merged
> > > > into my repo: https://github.com/paulirwin/lucene.net/pull/2
> > > >
> > > > Paul
> > > >
> > > >
> > > > On Tue, Oct 1, 2013 at 3:32 PM, mherndon michael <
> > > > mherndon@michaelherndon.com> wrote:
> > > >
> > > > > two options unless others can offer a better solution (crosses
> > > fingers):
> > > > >
> > > > > you can clone the official repo, make changes, commit to a repo on
> > > github
> > > > > and I can merge the changes..
> > > > >
> > > > > or
> > > > >
> > > > > you can fork my repo, make changes, send a pull request on github
> > and I
> > > > can
> > > > > merge the changes.
> > > > >
> > > > > I have zero control over the apache mirror.
> > > > >
> > > > > -M
> > > > >
> > > > >
> > > > > On Tue, Oct 1, 2013 at 2:32 PM, Paul Irwin <pirwin@feature23.com>
> > > wrote:
> > > > >
> > > > > > All, It dawned on me how to find the issue and I found it, it
was
> > an
> > > > > order
> > > > > > of operations issue with bit shifting in Packed64. That specific
> > > error
> > > > > now
> > > > > > seems to be fixed with this commit:
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
> https://github.com/paulirwin/lucene.net/commit/279224adcabb4341979ff111216d51ad7ccfc9b7
> > > > > >
> > > > > > So that brings up the question, how should I submit pull
> requests,
> > to
> > > > the
> > > > > > GitHub mirror? It still doesn't have the new branch from the
> apache
> > > > repo.
> > > > > > Let me know what I should do, or someone is welcome to pull
that
> > > commit
> > > > > in
> > > > > > manually.
> > > > > >
> > > > > > The gist I sent below still gives an exception, however, but
> it's a
> > > new
> > > > > > one, an "already finished" exception in FST which is the only
> other
> > > > > > exception I've seen so far. Once we fix that one we should be
in
> > > better
> > > > > > shape.
> > > > > >
> > > > > > Paul
> > > > > >
> > > > > >
> > > > > > On Tue, Oct 1, 2013 at 11:38 AM, Paul Irwin <
> pirwin@feature23.com>
> > > > > wrote:
> > > > > >
> > > > > > > All,
> > > > > > >
> > > > > > > Here's an example of where merge hits a background exception
in
> > the
> > > > 4.3
> > > > > > > branch of lucene.net. I can't figure this one out.
> > > > > > >
> > > > > > > https://gist.github.com/paulirwin/6780136
> > > > > > >
> > > > > > > Here's the exception: background merge hit exception:
> > _4(4.3.1):C1
> > > > > > > _1(4.3.1):C1 _2(4.3.1):C1 _3(4.3.1):C1 _5(4.3.1):C1
> _0(4.3.1):C1
> > > into
> > > > > _6
> > > > > > > [maxNumSegments=1]
> > > > > > >
> > > > > > > I can see that the exception is an IndexOutOfRangeException
> > > happening
> > > > > in
> > > > > > > Lucene.Net.Codecs.Compressing.LZ4.ReadInt, as it's trying
to
> get
> > a
> > > > > value
> > > > > > > out of a byte buffer at an index that is way larger than
the
> > buffer
> > > > > > length.
> > > > > > > I can't yet determine how or why it's getting that very
high
> > value.
> > > > > > >
> > > > > > > You need to build Core and Analyzers to use this code.
I just
> > threw
> > > > it
> > > > > in
> > > > > > > a console app.
> > > > > > >
> > > > > > > Thanks!
> > > > > > > Paul Irwin
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message