JDK-8214069 : (JavaFX) Default to xdg-open for document opening on UNIX

[Tristan971]

As per #260, here is a PR to default to xdg-open for document opening on UNIX systems.
It keeps the previous method as fallback in case of failure.

Little implementation note : chose to use try-catch rather than calling which xdg-open as this catches the nonexistence of xdg-open (never seen that for myself though) and also failure of xdg-open itself (in whatever case this might happen).

The only backwards compatibility lost here is if the users/developers were (un)knowingly relying on the lookup order of the browsers and not expecting the default browser to be used (imagining they have multiple browsers installed). I reckon this is acceptable.

---

As a side-note, while there are no tests relating to that part of the code, I still got test failures relating to prism/gtk.
Would be an interesting topic to investigate given #287 maybe ? (although these warnings don't look like they would be related as they're mostly type casting issues)

> Task :graphics:nativeFontFreetype
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-font/pango.c: In function ‘Java_com_sun_javafx_font_freetype_OSPango_FcConfigAppFontAddFile’:
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-font/pango.c:228:73: warning: passing argument 1 of ‘(jboolean (*)(void *, const char *))fp’ makes pointer from integer without a cast [-Wint-conversion]
                 rc = (jboolean)((jboolean (*)(void *, const char *))fp)(arg0, text);
                                                                         ^~~~
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-font/pango.c:228:73: note: expected ‘void *’ but argument is of type ‘jlong’ {aka ‘long int’}
> Task :graphics:ccLinuxFontPango
> Task :graphics:linkLinuxFontPango
> Task :graphics:nativeFontPango
> Task :graphics:ccLinuxGlassGlass
> Task :graphics:linkLinuxGlassGlass
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-glass/gtk/glass_window_ime.cpp: In member function ‘virtual void WindowContextBase::enableOrResetIME()’:
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-glass/gtk/glass_window_ime.cpp:236:66: error: cast between incompatible function types from ‘int (*)(XIM, XPointer, XPointer)’ {aka ‘int (*)(_XIM*, char*, char*)’} to ‘XIMProc’ {aka ‘void (*)(_XIM*, char*, char*)’} [-Werror=cast-function-type]
         XIMCallback startCallback = {(XPointer) jview, (XIMProc) im_preedit_start};
                                                                  ^~~~~~~~~~~~~~~~
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-glass/gtk/glass_dnd.cpp: In function ‘void process_dnd_source_mouse_release(GdkWindow*, GdkEventButton*)’:
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-glass/gtk/glass_dnd.cpp:827:44: error: cast between incompatible function types from ‘gboolean (*)()’ {aka ‘int (*)()’} to ‘GSourceFunc’ {aka ‘int (*)(void*)’} [-Werror=cast-function-type]
         gdk_threads_add_idle((GSourceFunc) dnd_finish_callback, NULL);
                                            ^~~~~~~~~~~~~~~~~~~
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-glass/gtk/glass_dnd.cpp: In function ‘void process_dnd_source_drop_finished(GdkWindow*, GdkEventDND*)’:
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-glass/gtk/glass_dnd.cpp:939:40: error: cast between incompatible function types from ‘gboolean (*)()’ {aka ‘int (*)()’} to ‘GSourceFunc’ {aka ‘int (*)(void*)’} [-Werror=cast-function-type]
     gdk_threads_add_idle((GSourceFunc) dnd_finish_callback, NULL);
                                        ^~~~~~~~~~~~~~~~~~~
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-glass/gtk/glass_dnd.cpp: In static member function ‘static GdkPixbuf* DragView::get_drag_image(gboolean*, gint*, gint*)’:
/home/tristan/perso/openjdk-jfx/modules/javafx.graphics/src/main/native-glass/gtk/glass_dnd.cpp:1116:71: error: cast between incompatible function types from ‘void (*)(gpointer)’ {aka ‘void (*)(void*)’} to ‘GdkPixbufDestroyNotify’ {aka ‘void (*)(unsigned char*, void*)’} [-Werror=cast-function-type]
                                 w, h, w * 4, (GdkPixbufDestroyNotify) g_free, NULL);
                                                                       ^~~~~~
cc1plus: all warnings being treated as errors
cc1plus: all warnings being treated as errors
> Task :graphics:ccLinuxGlassGlassgtk2 FAILED

FAILURE: Build failed with an exception.

* What went wrong:
Execution failed for task ':graphics:ccLinuxGlassGlassgtk2'.
> java.util.concurrent.ExecutionException: org.gradle.process.internal.ExecException: Process 'command 'gcc'' finished with non-zero exit value 1

[kevinrushforth]

The above errors you are getting are compilation errors when trying to build the native code for the javafx.graphics module. Are you running gcc 8 by chance? If so, then you are hitting #226.

[kevinrushforth]

I presume you've tested this (manually, since we don't have an automated test for this feature) ?

[kevinrushforth]

Now that I see what this is doing, wouldn't a simpler solution be to just take the original code and add "xdg-open" to the browsers list as the first item? You might still need a try/catch around the execution, so you could try the next item in the list if you got an exception.

In addition to being less code change, it would unify the approach between the case of using xdg-open and the specific browsers in the list.

[Tristan971]

Indeed that would be a cleaner way to go about it! Will change to it.

[Tristan971]

Changed it. Just slightly reformatted the array declaration as I feel it looks much nicer this way. Let me know if that's fine.

[kevinrushforth]

I'll take a look when I get a chance.

[Tristan971]

(also of course I tried it first)

[kevinrushforth]

Tracked in JBS as JDK-8214069.

[kevinrushforth]

Looks good with one minor request to remove an unrelated addition of a blank line.

You can just push it as a separate commit with just that one line reverted (best not to squash it which will require a force push). I'll merge it and push to jfx-dev/rt tomorrow.

[johanvos]

Looks good to me (apart from the unrelated newline)

[kevinrushforth]

I'll merge it and push to jfx-dev/rt tomorrow.

Or more accurately, I'll merge it shortly after you push the commit to remove the blank line.

[Tristan971]

Removed the linefeed :-)