| ltoscano | |
| sandsmark | |
| gateau |
documenttest: Port away from obsolete API.
documenttest: only compile when KDcraw is found
thumbnailprovidertest: Port to QUrl.
thumbnailprovidertest, documenttest: count remote tests as skipped when unavailable, instead of silently ignoring them (also: no more SkipSingle in Qt5)
Depends on D5749
documenttest:
thumbnailprovidertest:
| Automatic diff as part of commit; lint not applicable. |
| Automatic diff as part of commit; unit tests not applicable. |
| tests/auto/documenttest.cpp | ||
|---|---|---|
| 30 | Due to this include, this test should be compiled only if KDcraw is found | |
@ltoscano: Good catch (although nobody complained about it in the KDE4-version).
Diff was updated to:
Also: more detailed test plan.
I do apologize, I thought that I tested this with libkdcraw this time, but I didn't. documenttest is failing for me, and specifically here:
FAIL! : DocumentTest::testMetaInfoBmp() Compared values are not the same
Actual (value) : "200\u00D7100" Expected (expectedValue): "200x100" Loc: [/home/devel/git.kde.org/kde/kdegraphics/gwenview/tests/auto/documenttest.cpp(603)]
Few tests are skipped with " Not running this test: failed to get image. Try running ./fetch_testing_raw.sh in the tests/data directory and then rerun the tests", but this is a separate issues and not a regression, I'm reporting it just for the record but I'd say that it's out of scope for this change.
Works for me™ – just tested again with current gwenview git master + this diff on both of these platforms:
Therefore unlikely a problem with the patch, rather a more serious problem with an underlying library.
What platform are you on? Can you open BMP files with Gwenview? Not sure ATM what library Qt is using for BMP handling, which might explain the difference (in the test, metadata.bmp is created through creating an QImage and saving it to disk). KDE's CI [1] seems to be on Qt 5.7.0.
That's kind of expected. The script (to be run manually after setting up the git repo locally) will download three rather large (~27MB) RAW images, which surely do not belong in the git repo. As I understand it, they are meant as an optional test (similarly to the remote tests). Note: Running the script as ./tests/data/*.sh does not work, it has to be ./*.sh as stated in the warning message.
Uhm, really strange. I had consistently failures. After reapplying the patch, recompiling and retrying, documenttest consistently works!
The environment was/is a current Debian testing (stretch), so Qt 5.7.1.
To be on the safe side (and out of curiosity…), I looked into this some more:
You got "\u00D7" (unicode character for the multiplication sign, aka "×") instead of a normal lower-case letter "x" supposed to be created by the line of gwenview code in [2]. In this regard, the comment about Qt's BMP handling turns out to be more of a red herring. As your Qt version is the same as mine and you probably didn't change your KF5 version, those are also not likely to affect the propagation of the character from the call of i18nc() to the actual test.
Next, let's assume there was no compiler bug (version kept constant between runs) and no hardware issue (large hamming distance between bit representations of both characters, different length too – 2 vs. 1 byte). The only plausible explanation left is direct modification of [2], e.g. a stale patch or a search-and-replace job gone wild.
Anyway, glad you got it sorted out!