On Fri, Apr 25, 2014 at 4:54 AM, Honza Horak <[email protected]> wrote:
> Hi guys,
>
> in Fedora we got plenty of crash reports from users running strictly 32bit
> arch build of sqlite-3.8.4 or later [1]. The problem can be reproduced
> easily on 32bit, but is not seen on 64bit nor 32bit builds until
> sqlite-3.8.3, which seems to be important fact for me.
>
Thanks for the bug report, Honza. A few quick notes:
(1) Apparently this bugs has been tracked and worked on by RedHat for over
a month. If you had sent a bug report to this mailing list back near the
middle of March when you first noticed the problem, we probably could have
fixed it for you within a few hours.
(2) Thanks for including the SQL that causes the crash. However, the query
statement is still useless to us without a schema. Please also consider
doing "sqlite3 $database .schema" and sending the output with future bug
reports.
>
> I found that this issue was introduced by commit
> dca1945aeb3fb005263f9be00ee8e72b966ae303 [2] and when applying the
> attached patch (which reverts part of the commit above), it works fine for
> me, so we'll use it as a temporary fix.
>
> The problem we see in nautilus is that sqlite crashes during running the
> following select query (grabbed from GDB, so some values are missing and it
> is usable in sqlite console just like this):
>
> SELECT DISTINCT
> (SELECT "nie:url"
> FROM "nie:DataObject"
> WHERE ID = "1_u") COLLATE TRACKER AS "2_u",
> CAST ("urn_u_rank" AS TEXT) AS "3_u",
> COALESCE(SparqlFormatTime (
> (SELECT "nfo:fileLastModified"
> FROM "nfo:FileDataObject"
> WHERE ID = "1_u")), SparqlFormatTime (
> (SELECT "nie:contentLastModified"
> FROM "nie:InformationElement"
> WHERE ID = "1_u"))) AS "4_u",
> COALESCE(SparqlFormatTime (
> (SELECT "nfo:fileLastAccessed"
> FROM "nfo:FileDataObject"
> WHERE ID = "1_u")), SparqlFormatTime (
> (SELECT "nie:contentAccessed"
> FROM "nie:InformationElement"
> WHERE ID = "1_u"))) AS "5_u"
> FROM
> (SELECT "nfo:FileDataObject1"."ID" AS "1_u",
> 1,
> "fts3"."docid" AS "ID",
> tracker_rank(matchinfo("fts3"."fts",
> 'cl'),fts_column_weights()) AS "urn_u_rank"
> FROM "nfo:FileDataObject" AS "nfo:FileDataObject1",
> "nie:DataObject" AS "nie:DataObject2",
> "fts" AS "fts3"
> WHERE "nfo:FileDataObject1"."ID" = "nie:DataObject2"."ID"
> AND "nie:DataObject2"."ID" = "fts3"."docid"
> AND "nie:DataObject2"."tracker:available" = ?
> AND "fts3"."fts" MATCH '"status*"'
> AND (SparqlUriIsDescendant(? COLLATE TRACKER,
> (SELECT "nie:url"
> FROM "nie:DataObject"
> WHERE ID = "1_u") COLLATE TRACKER)
> AND (SparqlLowerCase (
> (SELECT "nfo:fileName"
> FROM "nfo:FileDataObject"
> WHERE ID = "1_u") COLLATE TRACKER) GLOB ?)))
> ORDER BY "urn_u_rank" DESC
>
> For full reproducer in 32bit Fedora 20, please, see the original bug
> report [1].
>
> [1] https://bugzilla.redhat.com/show_bug.cgi?id=1075889
> [2] http://www.sqlite.org/src/vinfo/dca1945aeb3fb005263f9be00ee8e7
> 2b966ae303?sbs=0
>
> Regards,
> Honza
>
> _______________________________________________
> sqlite-users mailing list
> [email protected]
> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
>
>
--
D. Richard Hipp
[email protected]
_______________________________________________
sqlite-users mailing list
[email protected]
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
