[PATCH 17/24] gnu: kwidgetsaddons: Fix test failure.

[PATCH 17/24] gnu: kwidgetsaddons: Fix test failure.

[David Craven]

* gnu/packages/kde-frameworks.scm (kwidgetsaddons)[arguments]:
  Enable tests. Set Xvfb pixel depth to 24 bits.
---
 gnu/packages/kde-frameworks.scm | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/gnu/packages/kde-frameworks.scm b/gnu/packages/kde-frameworks.scm
index e08b8cf..113257e 100644
--- a/gnu/packages/kde-frameworks.scm
+++ b/gnu/packages/kde-frameworks.scm
@@ -817,19 +817,19 @@ represented by a QPoint or a QSize.")
     (inputs
      `(("qtbase" ,qtbase)))
     (arguments
-     `(#:tests? #f ; FIXME: libGL error: failed to load driver: swrast.
-       #:phases
+     `(#:phases
         (modify-phases %standard-phases
           (add-before 'check 'check-setup
             (lambda* _
-              (setenv "CTEST_OUTPUT_ON_FAILURE" "1") ; enable debug output
-              (setenv "LIBGL_DEBUG" "verbose") ; enable debug output
               (setenv "DBUS_FATAL_WARNINGS" "0")))
           (add-before 'check 'start-xorg-server
             (lambda* (#:key inputs #:allow-other-keys)
               ;; The test suite requires a running X server.
+              ;; Xvfb doesn't have proper glx support and needs a pixeldepth
+              ;; of 24 bit to avoid "libGL error: failed to load driver: swrast"
+              ;;                    "Could not initialize GLX"
               (system (string-append (assoc-ref inputs "xorg-server")
-                                     "/bin/Xvfb :1 &"))
+                                     "/bin/Xvfb :1 -screen 0 640x480x24 &"))
               (setenv "DISPLAY" ":1")
              #t)))))
     (home-page "https://community.kde.org/Frameworks")
-- 
2.9.0

Re: [PATCH 17/24] gnu: kwidgetsaddons: Fix test failure.

[Danny Milosavljevic]

>                (system (string-append (assoc-ref inputs "xorg-server")
> -                                     "/bin/Xvfb :1 &"))
> +                                     "/bin/Xvfb :1 -screen 0 640x480x24 &"))

Might

             (setenv "QT_QPA_PLATFORM" "offscreen")

also be enough? Starting an Xvfb server and then not stopping it again is kinda ... extreme. Also, who says that :1 is free?

I can see that it was there before, but just to be sure...

Re: [PATCH 17/24] gnu: kwidgetsaddons: Fix test failure.

[David Craven]

> Might (setenv "QT_QPA_PLATFORM" "offscreen") also be enough?

Interesting, that also fixes the problem, but causes a new one...

********* Start testing of KDualActionTest *********
Config: Using QtTest library 5.7.0, Qt 5.7.0
(x86_64-little_endian-lp64 shared (dynamic) release build; by GCC
4.9.3)
PASS   : KDualActionTest::initTestCase()
PASS   : KDualActionTest::testSetGuiItem()
FAIL!  : KDualActionTest::testSetIconForStates() Compared pointers are
not the same
   Actual   (action.inactiveIcon()): (nil)
   Expected (icon)                 : 0x65a4d0
   Loc: [/tmp/guix-build-kwidgetsaddons-5.24.0.drv-0/kwidgetsaddons-5.24.0/autotests/kdualactiontest.cpp(56)]
PASS   : KDualActionTest::testSetActive()
PASS   : KDualActionTest::testTrigger()
PASS   : KDualActionTest::cleanupTestCase()
Totals: 5 passed, 1 failed, 0 skipped, 0 blacklisted, 3ms
********* Finished testing of KDualActionTest *********

>  Starting an Xvfb server and then not stopping it again is kinda ... extreme.

Since the tests are run in a container, I'd expect all processes started inside
the container to receive a sigterm when the container is (shutdown?).

> Also, who says that :1 is free?

Again since the tests are run in a container, I'd expect that there
isn't a xserver
running unless I explicitly start it.

Are my assumptions wrong?
Do you think it's worth tracking down the test failure when my solution works?

Re: [PATCH 17/24] gnu: kwidgetsaddons: Fix test failure.

[David Craven]

I'll withhold this patch pending further investigation.

Thank you!

Re: [PATCH 17/24] gnu: kwidgetsaddons: Fix test failure.

[Danny Milosavljevic]

> Again since the tests are run in a container, I'd expect that there
> isn't a xserver
> running unless I explicitly start it.
> 
> Are my assumptions wrong?

> Do you think it's worth tracking down the test failure when my solution works?

No, if that's the case it's fine to use Xvfb too. It's just that QT_QPA_PLATFORM would be a lot less compute-intensive.