2017-01-22 5:16 GMT+03:00 Slávek Banko slavek.banko@axis.cz:
On Wednesday 21 of December 2016 00:19:58 Fat-Zer wrote: In short: great, good work!
Long: I have some problems with tests in tdelibs.
- On Debian/Ubuntu is used WITH_GCC_VISIBILITY and tests
tdewallet/backend/testbf and testsha causes FTBFS - see patch bp0016-fix-tdewallet-tests-with-hidden-visibility.diff
- In dcop/tests/run-tests.sh is used "source ./shell.generated..."
where "source" is a bash specific => test may fail with other shells (dash on Debian/Ubuntu). Fixed in bp0017-fix-dcop-test-bashism.diff
- Test tdeio/kmimetypetest expects mimetype definitions in tdesycoca
cache => fails if it is done prior to installation. Proposed patch bp0018-fix-tdeio-kmimetype-test.diff contains a workaround for the test during the compilation on a clean build environment.
- Test tdeabc/vcardparser/testread2 fail, probably because of some
erroneous utf8 conversion. For now I've had not examined it in detail == patch not included.
- Test tdeabc/testldapclient expects dcop server, tdelauncher and access
to a foreign ldap server on the Internet => fails if it is done prior to installation and depends on many factors which can not be influenced. Therefore, I suggest not to perform this test during compilation - see patch bp0019-fix-tdeabc-ldapclient-test.diff
Slavek, thanks for reviewing and debugging all the errors.
I agree with and approve 4 patches out of 5: all except bp0018-fix-tdeio-kmimetype-test.diff.
IMO the script is potentially error-prone, e.g. it may slightly break some-ones installation if .trinity/share/mimelnk is already a symlink. And in general it's quite kludgy workround... All in all I would prefer just to disable the test instead of fixing it that way...
Should I push the patches to the repository? or are you going to merge the patches soon?