Bug 226429 (sqlite)

https://www.redhat.com/archives/fedora-extras-commits/2005-April/msg01144.html

Well, this was a pre-Bugzilla review.

Package review:

- rpmlint checks return:

    sqlite-3.5.4-2.fc9.src.rpm -

sqlite.src:14: W: unversioned-explicit-obsoletes sqlite3
sqlite.src:14: W: unversioned-explicit-obsoletes sqlite3-devel

sqlite3 last appeared (AFAICT) in FC-3's extras.  Since it's more than 3
releases old, we can drop this obsoletes.  Also, the matching 'provides' are
missing.

sqlite.src: W: mixed-use-of-spaces-and-tabs (spaces: line 57, tab: line 11)

Easy to fix.

  sqlite-devel-3.5.4-2.fc9.i386.rpm -

sqlite-devel.i386: W: summary-ended-with-dot Development tools for the sqlite3
embeddable SQL database engine.

* both of the above are fixed in the attached patch to the spec file. *

  sqlite-tcl-3.5.4-2.fc9.i386.rpm -

sqlite-tcl.i386: W: no-documentation

* Can probably be ignored

sqlite-tcl.i386: W: summary-ended-with-dot Tcl module for the sqlite3 embeddable
SQL database engine.

* fixed in attached patch

sqlite-tcl.i386: E: arch-dependent-file-in-usr-share
/usr/share/tcl8.5/sqlite3/libtclsqlite3.so
sqlite-tcl.i386: W: unstripped-binary-or-object
/usr/share/tcl8.5/sqlite3/libtclsqlite3.so

* This is obviously wrong, they should be in /usr/lib - I'm not sure what to do
to fix this properly, however.

- package meets naming guidelines - Ok
- license ( ) OK, text in %doc, matches source - Ok - Public Domain
- spec file legible, in am. english - Ok
- source matches upstream - Ok
- package compiles on devel (x86) - Ok
- no missing BR - Ok
- no unnecessary BR - Ok
- no locales - Ok
- not relocatable - Ok
- owns all directories that it creates - Ok
- no duplicate files - Ok
- permissions ok - Ok
- %clean - Ok
- macro use consistent - Ok
- code, not content - Ok
- no need for -docs - Ok
- nothing in %doc affects runtime - Ok
- no need for .desktop file - Ok
- devel package ok - Ok
- no .la files - removed in %install section
- post/postun ldconfig ok - Ok
- devel requires base package n-v-r - Ok


- Tests are not run by default.

* This is probably because several tests fail.  I've enabled tests in the
attached specfile patch, and included patches which fix or comment out the tests
that fail.  Someone needs to check to see why the failing tests are failing, and
either fix the tests upstream, fix the bug upstream, or figure out why the tests
fail for us.

Created attachment 291445 [details]
patch to specfile to fix some package review issues

this patch to the specfile fixes some of the package review issues, and enables
make check

Created attachment 291446 [details]
use .so instead of .la files when running tests

Created attachment 291447 [details]
fix the bind test - NaN doesn't seem to be valid for a double column

Created attachment 291448 [details]
fix io test - but I don't know why it was broken to begin with, so investigation is warrented.

Created attachment 291449 [details]
printf formatting is broken for some reason

Created attachment 291450 [details]
the format of this tcl error message changed.  the test should be fixed, but I don't know tcl syntax enough to fix it.

Robin, the Fedora-review flag should be set to "-" only if a package has
fundamental flaws and the review should deny it.
As long as a package is under review, the flag should be "?"

Yes, you're right, of course - I fail at following the process. :-)

Thanks.

Panu, this seems to be stuck for some time due to not reply from package maintainers, can you follow to Robin's proposals & review?

Lets just say that it might help somewhat that this bug has been brought to my attention :)

I'll look into this as soon as I've dealt with some other more pressing matters.

Okay, I fixed the rpmlint nitpickery over tab-vs-space etc + obsoletes + the io tests (the difference comes from SQLITE_DISABLE_DIRSYNC build-switch so its ok) in devel. With that, as of sqlite 3.6.14, the test-suite is error-free on ix86 + x86_64, the other testsuite-patches here dont seem to be relevant (at least anymore). Except, hmm, what's with the "use .so instead of .la" patch?

PPC (both 32 and 64bit) is not quite there however, this nan-breakage:
http://www.sqlite.org/cvstrac/tktview?tn=3404 and then a pile of rtree-tests fails (possibly due to the nan-issue, haven't looked that closely).

Thanks, Panu!

--

Robin: Can you continue with the review, please? The packages are here: http://koji.fedoraproject.org/koji/packageinfo?packageID=485

Yes, I'll get on it.  I probably won't be able to get to it until tomorrow, though.

Sorry, one more: rpmlint points out that the -rpath issue is present here.  Ref:

https://fedoraproject.org/wiki/Packaging/Guidelines#Beware_of_Rpath

The suggested method using sed to strip out the hardcoded paths seems to work for me at first glance.

rpath is disabled starting with 3.6.14-2

Thanks! So close to review+ :), right?

Yup!  Passes review.

Thanks all.

Wohoo :) Thanks.