Display a message when an activity fails to start

[wadeb]

The attached patch displays a simple error message when an activity's process ends while the activity is still flagged as launching.

[wadeb]

Note: These patches should be considered a prototype for further development. I get occasional segfaults when pressing the Close button, and I would like to improve the looks further.

Also, if ​http://wiki.sugarlabs.org/go/Features/Problem_Reports is accepted, there needs to be a button to open the Report a problem control panel.

[mstone]

These patches look promising to me and I'll be happy to help you iron out the wrinkles some other morning or evening. One suggestion in the mean-time: in addition to popping up a window, it's really important to the overall interaction design that we make a journal entry containing the activity's log file when activity launches fail because we can then bring existing tools into play in order to get that log file home. Might you be willing to write up a third patch containing this extension for me to test while I've got this stuff paged in to my head?

Lastly, regarding segfaults: take a look at #1123. I mention it because alsroot's patches in it caught my eye when I looked at the history of the activityfactory code that you're modifying.

[wadeb]

#1123 seems very relevant to this bug. Here is the backtrace from the segfault for reference.

#0  PyErr_Fetch (p_type=0xbf93d39c, p_value=0xbf93d398, p_traceback=0xbf93d394) at Python/errors.c:235
#1  0x00b1254f in dbus_py_variant_level_clear (self=0x9b2df80) at abstract.c:143
#2  0x00b125f2 in DBusPyStrBase_tp_dealloc (self=0x9b2df80) at abstract.c:443
#3  0x02064e34 in dict_dealloc (mp=0x9b341c4) at Objects/dictobject.c:911
#4  0x02088155 in subtype_dealloc (self=0x9b2df6c) at Objects/typeobject.c:1006
#5  0x02064e34 in dict_dealloc (mp=0x9b3402c) at Objects/dictobject.c:911
#6  0x02088155 in subtype_dealloc (self=0x9a75b2c) at Objects/typeobject.c:1006
#7  0x0207daa7 in tupledealloc (op=0x968e92c) at Objects/tupleobject.c:170
#8  0x00df7121 in pyglib_error_check () from /usr/lib/python2.6/site-packages/gtk-2.0/glib/_glib.so
#9  0x002cd984 in g_source_callback_unref (cb_data=0x9a283c0) at gmain.c:941
#10 0x002cdf4e in g_source_destroy_internal (source=0x9adbbd0, context=<value optimized out>, have_lock=1) at gmain.c:720
#11 0x002ce35b in g_main_dispatch (context=<value optimized out>) at gmain.c:1849
#12 IA__g_main_context_dispatch (context=<value optimized out>) at gmain.c:2377
#13 0x002d19e0 in g_main_context_iterate (context=0x90a2f80, block=<value optimized out>, dispatch=1, self=0x90a4e98)
    at gmain.c:2455
#14 0x002d1e4f in IA__g_main_loop_run (loop=0x93f4e78) at gmain.c:2663
#15 0x070e7c09 in IA__gtk_main () at gtkmain.c:1205

[wadeb]

Better patch for GIL lock issue

[wadeb]

Ok, this patch appears to fix the problem.

I basically punted and moved all the processing into an idle callback. This seems like the safest way to work around the GIL issue reported in #1123; this patch should be considered a more complete fix for that bug.

[wadeb]

Replying to wadeb:

Doh, looks like I uploaded the wrong patch last night. I'm going to make some UI improvements as well, so removing r? for now.

[wadeb]

Update to GIL fix; remove debug prints

[wadeb]

New version of the Sugar patch

[wadeb]

I posted a brand new patch for Sugar.

• Improves the layout so the icon doesn't move when the error appears
• Uses Stop as suggested by Gary
• Delays for a few seconds between pulses to give the user something to count and potentially improve performance.

[bernie]

Ping? I've tested this patch for a while, and I think it's a huge usability improvement. Can we review it and possibly merge it?

[garycmartin]

Replying to bernie:

Ping? I've tested this patch for a while, and I think it's a huge usability improvement. Can we review it and possibly merge it?

A large +1 from me, it is a _HUGE_ improvement in behaviour. Now, FWIW I've seen a few spurious activity launches that leave duplicate top frame activity icons in the toolbar, but I've not been convinced that it is the result of this patch, yet.

[bernie]

YAY!

[bernie]

Screenshot of the UI change introduced by this patch

[tomeu]

missed feature freeze

[alsroot]

I guess everybody forgot about it..

In my mind patch could be less invasive e.g. w/o introduction new new dbus method and most invasive part would be "Stop" button.

I'm preparing another patch that at least could useful for 0.84 branch and if everybody will agree, for 0.88

[alsroot]

sugar-toolkit patch is reusing NotifyLaunchFailure dbus method to notify shell about not successful activity termination (activity could already show main window) but sugar part will proper handle this. At the end wadeb's new dbus method will make code cleaner, I just not sure can we introduce new dbus API changes for 0.88 and for 0.84 (for XO-1.5).

sugar patch is based on wabeb's with just removing NotifyActivityEnded dbus call. If GUI part is a blocker for inclusion to 0.88, I guess it could be simplified to just showing "Stop" button.

[alsroot]

..if last patch could be considered as less invasive
it also adds new gettext string since w/o message stop button could be confusing

[alsroot]

r+ by tomeu and erikos on #sugar

​http://git.sugarlabs.org/projects/sugar/repos/mainline/commits/3c6cbc2ff7d7fc1ba59862cdaa3a86bcb4e1cba0
​http://git.sugarlabs.org/projects/sugar-toolkit/repos/mainline/commits/161aec771e058416d672c36b3129cc4a1a11c7e2

[bernie]

Thank you, this is very useful!

[bernie]

Sayamindu, can we backport this patch to 0.84?

[dnarvaez]

Milestone 0.88 deleted