lucenenet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Wyatt Barnett <wyatt.barn...@gmail.com>
Subject Re: Board report - updates
Date Thu, 15 Oct 2015 21:50:18 GMT
Sorry to chime in late here -- I got roped into a few things early this
summer and I'm just recovering my life.

One thing I'd suggest in terms of tests is to use nunit's runners rather
than resharper's -- they handle some of the "creative" stuff that might be
going on better and they don't melt down due to too much output in debug
mode.

Regarding setting culture the question I'd have is "does that have any
interesting downstream effects where culture isn't set to
CultureInfo.Invariant?"

As for other updates I'm getting back in the saddle and looking at bringing
home this build process, I need to formualte some things and I'll start a
separate thread for that.

On Thu, Oct 15, 2015 at 2:49 PM Simon Svensson <sisve@devhost.se> wrote:

> Hi,
>
> I'm was running all non-ignored tests from the Lucene.Net.Tests assembly
> which included the long-running ones. For those reading; add
> "LongRunningTest" to Resharper > Options... > Unit Testing > Skip tests
> from categories.
>
> I'm now down to less than 6 minutes.
>
> Anyhow, I'm getting several culture-related failures, some obvious ones
> in TestSort (TestDouble* and TestFloat*). One way to fix these would be
> to modify the FieldCache+AnonymousDoubleParser to use the correct
> overload of Double.Parse; but I get some feeling deep down that it's a
> bad solution, mostly because we do what Java does.
>
> Java:
>
> https://github.com/apache/lucene-solr/blob/lucene_solr_4_8_0/lucene/core/src/java/org/apache/lucene/search/FieldCache.java#L337
> .NET
> <https://github.com/apache/lucene-solr/blob/lucene_solr_4_8_0/lucene/core/src/java/org/apache/lucene/search/FieldCache.java#L337.NET>
> :
>
> https://github.com/apache/lucenenet/blob/master/src/Lucene.Net.Core/Search/FieldCache.cs#L619
>
> We could solve this by setting
> Thread.CurrentThread.(CurrentCulture|CurrentUICulture) =
> CultureInfo.InvariantCulture in LuceneTestCase.SetUp. Does this sound okay?
>
> Doing the above culture hack results in 75 failures for me. That is
> excluding the long-running tests. One of the failures is the
> TestImpersonation that has no special attributes or anything to indicate
> that it should not be run. I'll see if I can debug the errors further...
>
> // Simon
>
> On 14/10/15 19:56, Laimonas Simutis wrote:
> > Simon,
> >
> > To be more specific, I am running the same configuration as in TC
> automated
> > tests. Here is the latest results:
> >
> >
> http://teamcity.codebetter.com/viewLog.html?buildId=197870&tab=buildResultsDiv&buildTypeId=LuceneNet_Core
> >
> > It says 5 tests failed, 4 new. Those 4 appear to be due to some state
> that
> > is not cleaned up properly from other tests in the same run, and one
> > TestDuelingCodecs test fails because it dynamically tries to load codecs
> > from packages that haven't been converted yet.
> >
> > I know there are some long running tests that are skipped in this TC run,
> > they succeed, just take much longer (from what I remember). Perhaps you
> are
> > running those as well?
> >
> >
> >
> > On Wed, Oct 14, 2015 at 1:48 PM, Simon Svensson <sisve@devhost.se>
> wrote:
> >
> >> Hi,
> >>
> >> Running in Release mode helped a lot; the whole test suite (2608 tests)
> >> ran in about 15 minutes. However, I have 99 failing tests and 110
> >> ignored tests. I can get it down to 81 failing tests by hardcoding
> >> LuceneTestCase.SetUp to set CurrentCulture and CurrentUICulture to
> >> InvariantCulture, but some tests fails with
> >> "System.NotSupportedException : this codec can only be used for reading"
> >> which doesn't sound very culture relevant.
> >>
> >> Are you saying that they should all work?
> >>
> >> I could look into the [presumed] culture-related failures; most notably
> >> the ones in the TestSort class where TestDouble and TestFloat fails
> >> using my culture (sv-SE).
> >>
> >> // Simon
> >>
> >>
> >> On 14/10/15 12:19, Laimonas Simutis wrote:
> >>> Simon,
> >>>
> >>> By default the tests run in verbose mode that outputs large amounts of
> >>> logging that slow things down. If you run a release version, the
> >> verbosity
> >>> flag gets turned off. Or you can turn it off via app.config setting,
> the
> >>> code location that controls this flag is here:
> >>>
> >>>
> >>
> https://github.com/apache/lucenenet/blob/master/src/Lucene.Net.TestFramework/Util/LuceneTestCase.cs#L295
> >>>
> >>>
> >>>
> >>> On Wed, Oct 14, 2015 at 2:18 AM, Simon Svensson <sisve@devhost.se>
> >> wrote:
> >>>
> >>>> Hi,
> >>>>
> >>>> Laimonas, when you say that the tests are passing for Lucene Core; how
> >>>> are you executing them?
> >>>>
> >>>> I've got ...
> >>>>  * Windows 7 running in latest Parallells Desktop
> >>>>  * Visual Studio 2015 Community Edition
> >>>>  * I've checked out commit 812e1c5 (from 24 aug 2015)
> >>>>  * I'm running in Debug / Mixed Platforms (which, for some reason,
> >>>> compiles Lucene.Net as x86)
> >>>>  * Lucene.Net.Tests (and all references) build successfully.
> >>>>  * Rightclicking Lcuene.Net.Tests and executing "Run Unit Tests" gives
> >>>> me a toolwindow "Unit Test Sessions - All tests from Lucene.Net.Tests"
> >>>> with a total of 2608 tests.
> >>>>
> >>>> Running all tests takes ... And I'm giving up waiting for a number.
I
> >>>> managed to run 29 of 2608 tests in the ~30 minutes I was waiting.
> >>>>
> >>>> Doing all the above steps in Visual Studio 2013 Professional gives me
> >>>> the same behavior, 29 of 2608 tests executes in 30 minutes before I
> stop
> >>>> the test execution.
> >>>>
> >>>> How should the tests be run?
> >>>>
> >>>> // Simon
> >>>>
> >>>> On 14/10/15 02:29, Laimonas Simutis wrote:
> >>>>> Prescott,
> >>>>>
> >>>>> We are at all tests passing for Lucene Core. Now the next hurdle
is
> to
> >>>>> finish up Analysis port and tests and that's where we got stuck.
> >>>>>
> >>>>> I might start having more cycles in a month or two, but right now
> there
> >>>> is
> >>>>> no activity in the project.
> >>>>>
> >>>>> On Tue, Oct 13, 2015 at 2:38 PM, Prescott Nasser <
> >>>>> prescott.nasser@hotmail.com> wrote:
> >>>>>
> >>>>>> Hey guys, I have been fairly MIA - anyone have a recap for 4.0
> >> release.
> >>>> I
> >>>>>> think its clear we are struggling to get it across the line,
which I
> >>>> will
> >>>>>> report and ask for help.
> >>>>>>
> >>>>>> Sent from my Windows Phone
> >>>>>
> >>>>
> >>>>
> >>>
> >>
> >>
> >
>
>

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