Bug 581792

Summary:	dev-db/sqlite fails tests (fts5simple-11.2 shell1-5.0)
Product:	Gentoo Linux	Reporter:	Matt Turner <mattst88>
Component:	Current packages	Assignee:	Arfrever Frehtes Taifersar Arahesis <arfrever.fta>
Status:	RESOLVED TEST-REQUEST
Severity:	normal	CC:	etienne.buira, kingjon3377, proxy-maint
Priority:	Normal
Version:	unspecified
Hardware:	All
OS:	Linux
Whiteboard:
Package list:		Runtime testing required:	---
Attachments:	emerge --info sqlite-3.12.2-build.log

Description Matt Turner gentoo-dev

2016-05-02 04:14:53 UTC

When testing sqlite for bug 578940, I noticed that it failed its test suite:


SQLite 2016-04-18 17:30:31 92dc59fd5ad66f646666042eb04195e3a61a9e8e
2 errors out of 179025 tests on macbook Linux 32-bit little-endian
!Failures on these tests: fts5simple-11.2 shell1-5.0

On amd64:

dev-db/sqlite-3.11.1/temp/build.log:!Failures on these tests: fts5simple-11.2
dev-db/sqlite-3.12.0/temp/build.log:!Failures on these tests: fts5simple-11.2 shell1-5.0
dev-db/sqlite-3.12.2/temp/build.log:!Failures on these tests: fts5simple-11.2 shell1-5.0

I would much rather stabilize something that passes its own test suite...

Comment 1 Arfrever Frehtes Taifersar Arahesis 2016-05-02 22:12:59 UTC

(Test suite passed for me.)

Attach full build log and output of `emerge --info`.

Comment 2 Matt Turner gentoo-dev

2016-05-02 22:55:36 UTC

Created attachment 432996 [details]
emerge --info

Comment 3 Matt Turner gentoo-dev

2016-05-02 22:56:23 UTC

Created attachment 432998 [details]
sqlite-3.12.2-build.log

Comment 4 gentoo_usr 2016-05-17 20:38:41 UTC

I can confirm the failures for dev-db/sqlite-3.12.2 on my system.
It's also amd64 but profile hardened/linux/amd64/selinux.

Some google results:

The shell1-5.0 "E0" error is present in SQLite 3.10 and ignored[0,1]. However, the failure was on a x86 system running BSD. In SQLite 3.12 the error is gone[2].
The error is also known in SQLite 3.8.10 and also reported on the upstream mailing list[3]. Sadly the only response was 'Cannot reproduce'.
Later it was reported again for the successor version 3.8.11[4] with unclear consequences.

I've found nothing for the fts5simple-11.2 error (except for this bugreport).

[0] https://www.sqlite.org/checklists/3100000/index
[1] https://www.sqlite.org/checklists/3100000/history?id=12a
[2] https://www.sqlite.org/checklists/3120000/index
[3] https://marc.info/?l=sqlite-dev&m=144565682925857&w=2
[4] http://sqlite.1065341.n5.nabble.com/SQLite-version-3-8-11-coming-soon-td83112.html

Comment 5 Arfrever Frehtes Taifersar Arahesis 2016-09-23 11:52:08 UTC

Failure in shell1-5.0 should be wrought around by https://sqlite.org/src/info/544c990afd3b6406 (included in SQLite >=3.14.0).

Please test >=dev-db/sqlite-3.14.2.

Comment 6 Étienne Buira 2016-11-29 16:05:13 UTC

fts5simple failure is a test error that have been fixed in https://sqlite.org/src/info/a495f8e77e142411

With both fixes applied on 3.13.0 (once epatch_user was added to ebuild), the testsuite ran flawlessly.

Comment 7 gentoo_usr 2016-11-30 18:17:20 UTC

I can confirm that the two failures are gone in dev-db/sqlite-3.14.2 and dev-db/sqlite-3.15.1. (Both versions seem to have the above mentioned upstream fixes).