Closed asfimport closed 7 years ago
Trejkaz (migrated from JIRA)
I see this when running our unit tests also. Evidently somewhere up the call chain there is some code we don't trust calling us. We then call Lucene, which ultimately calls this stuff.
(We're bitten by #7981 as well, because we are still on 5.4.)
Aaron Madlon-Kay (migrated from JIRA)
Since there has been no action on this issue, we were forced to publish our patched version:
compile 'org.omegat.lucene:lucene-core:5.2.1'
Michael McCandless (@mikemccand) (migrated from JIRA)
Is OS_ARCH.contains("64"))
really a safe way to determine if we are running in a 64 bit JVM? Maybe we should only do this on fallback, if the security manager doesn't let us do System.getProperty("sun.arch.data.model")
?
Aaron Madlon-Kay (migrated from JIRA)
> Maybe we should only do this on fallback
That's precisely what my patch does.
Michael McCandless (@mikemccand) (migrated from JIRA)
> That's precisely what my patch does.
Woops, sorry, I did not look closely enough: you're right!
OK I think this is a good approach ... I'll push soon. Thanks Aaron Madlon-Kay!
Michael McCandless (@mikemccand) (migrated from JIRA)
> That's precisely what my patch does.
Woops, sorry, I did not look closely enough: you're right!
OK I think this is a good approach ... I'll push soon. Thanks Aaron Madlon-Kay!
ASF subversion and git services (migrated from JIRA)
Commit 813b6855656ecd50a7a28376822bd7b65154cee8 in lucene-solr's branch refs/heads/master from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=813b685
LUCENE-7135: work around security manager when checking for 32/64 bit JVM
ASF subversion and git services (migrated from JIRA)
Commit 96372be35c7c7d937561ce5d526d688ab447cd86 in lucene-solr's branch refs/heads/branch_6x from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=96372be
LUCENE-7135: work around security manager when checking for 32/64 bit JVM
Michael McCandless (@mikemccand) (migrated from JIRA)
Thanks Aaron Madlon-Kay!
Aaron Madlon-Kay (migrated from JIRA)
Thanks very much for your help, @mikemccand.
Is there any hope for a backport to 5? I suspect it will take a long time for our dependencies to move up.
Michael McCandless (@mikemccand) (migrated from JIRA)
I think it's unlikely we'll do anothe 5.x release Aaron Madlon-Kay ... but I'll mark this as 5.5.4 just in case we do.
Aaron Madlon-Kay (migrated from JIRA)
Understood. Thanks!
Michael McCandless (@mikemccand) (migrated from JIRA)
Reopen for possible eventual 5.5.4 backport.
ASF subversion and git services (migrated from JIRA)
Commit 2baad4c22d05a1fcc4a09044eae868b6a5bfe1cf in lucene-solr's branch refs/heads/master from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=2baad4c
LUCENE-7135: add issue number in CHANGES.txt
ASF subversion and git services (migrated from JIRA)
Commit f9e2f0c5b65b389e330a16657396a922e47fce1d in lucene-solr's branch refs/heads/branch_6x from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=f9e2f0c
LUCENE-7135: add issue number in CHANGES.txt
Shalin Shekhar Mangar (@shalinmangar) (migrated from JIRA)
This issue is marked for 6.3 but it was not backported to branch_6_3. So the fix version should be 6.4
Michael McCandless (@mikemccand) (migrated from JIRA)
Woops, thanks @shalinmangar, I fixed it.
ASF subversion and git services (migrated from JIRA)
Commit 08526f581edef098cd94e899beef2a0fff95535c in lucene-solr's branch refs/heads/branch_6_3 from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=08526f5
LUCENE-7135: work around security manager when checking for 32/64 bit JVM
ASF subversion and git services (migrated from JIRA)
Commit 3366219c6edbb8c7791e8e8bab08ede5b485e2d1 in lucene-solr's branch refs/heads/branch_6_3 from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=3366219
LUCENE-7135: add issue number in CHANGES.txt
Michael McCandless (@mikemccand) (migrated from JIRA)
OK I backported this for 6.3.0
Michael McCandless (@mikemccand) (migrated from JIRA)
There is a bug in the logic here: it marks a 64 bit JVM as 32 bit!
ASF subversion and git services (migrated from JIRA)
Commit 92f56ea9dd72d935c21aadae6a20ee23fa9c3cf8 in lucene-solr's branch refs/heads/master from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=92f56ea
LUCENE-7135: only use OS_ARCH if we couldn't access sun.arch.data.model
ASF subversion and git services (migrated from JIRA)
Commit 97c97838022efcac8ccbcffb5edf5735605fdf17 in lucene-solr's branch refs/heads/branch_6x from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=97c9783
LUCENE-7135: only use OS_ARCH if we couldn't access sun.arch.data.model
ASF subversion and git services (migrated from JIRA)
Commit a66a44513ee8191e25b477372094bfa846450316 in lucene-solr's branch refs/heads/branch_6_3 from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=a66a445
LUCENE-7135: only use OS_ARCH if we couldn't access sun.arch.data.model
Aaron Madlon-Kay (migrated from JIRA)
Very sorry about that!
Michael McCandless (@mikemccand) (migrated from JIRA)
No worries Aaron Madlon-Kay, stuff happens :) I'm just glad we caught it before 6.3.0 was out.
Michael McCandless (@mikemccand) (migrated from JIRA)
One nice side effect of this 2 day accident was it caused e.g. our nightly geo benchmark https://people.apache.org/\~mikemccand/geobench.html to use NIOFSDirectory
instead of MMapDirectory
, so we get to see at least for this benchmark how much better MMapDirectory
is!
I'm working on nightly benchmarks for the sparse case, preview here: https://people.apache.org/\~mikemccand/lucenebench/sparseResults.html#search_sort_qps
There's an even bigger drop in that benchmark, which is what made me think there might be an issue with this change in the first place. Benchmarking is important :)
ASF subversion and git services (migrated from JIRA)
Commit 92f56ea9dd72d935c21aadae6a20ee23fa9c3cf8 in lucene-solr's branch refs/heads/apiv2 from Mike McCandless https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=92f56ea
LUCENE-7135: only use OS_ARCH if we couldn't access sun.arch.data.model
Shalin Shekhar Mangar (@shalinmangar) (migrated from JIRA)
Closing after 6.3.0 release.
I have an app that I deploy via WebStart that uses Lucene 5.2.1 (we are locked to 5.2.1 because that's what LanguageTool uses).
When running under the WebStart security manager, there are two locations where exceptions are thrown and prevent pretty much all Lucene classes from initializing. This is true even when we sign everything and specify
<security><all-permissions/></security>
.RamUsageEstimator
, fixed by #7981In
Constants
, caused by the callSystem.getProperty("sun.arch.data.model")
(stack trace below).The latter is still present in the latest version. My patch illustrates one solution that appears to be working for us.
(This patch, together with a backport of the fix to #7981, seems to fix the issue for our purposes. However if you really wanted to make my day you could put out a maintenance release of 5.2 with both fixes included.)
Migrated from LUCENE-7135 by Aaron Madlon-Kay, 1 vote, resolved Nov 04 2016 Environment:
Attachments: LUCENE-7135.diff