Sugar crash when stopping inactive Etoys

[FGrose]

Observed with Soas-20090305 on Dell 8400 with 3 GByte of RAM

shell.log & webactivity-1.log attached

[FGrose]

shell.log

[FGrose]

webActivity-1.log

[FGrose]

This seemed to occur as the Etoys instance was closing. I hit the Frame|Etoys|Stop panel button at least 2 times as there was no feedback on its status.

[FGrose]

Replying to FGrose:

This seemed to occur as the Etoys instance was closing. I hit the Frame|Etoys|Stop panel button at least 2 times as there was no feedback on its status.

It was the first instance of Etoys opened on this machine. When I tried to reproduce, I noticed that Etoys stopped quickly, and I couldn't reproduce the fault.

[FGrose]

Retesting with Soas1-20090319, and I tried to Stop the Etoys instance from its Frame panel 5 or 6 times with no apparent effect. When I tried to resume it from the same panel, the Frame instance disappeared, so it may have Stopped but not cleared its panel from the Frame. There was no X restart this time. (This was the first instance to run on this installation as well.)

Restarted Etoys to test again, the Frame label and panel behaviour is the same. This time I <Alt><Tab> to the Frame instance and it disappeared as when trying to resume from the panel, so Frame icon and panel seem not to have been dismissed after Stopping.

[FGrose]

Observed with Soas2-200905031329.iso installed with LiveUSB-creator3.6.5 onto this hardware, ​http://www.smolts.org/client/show/pub_ee2f82d0-f863-454f-9d72-b1ee60fb3c65.

(Previously reported Browse involvement was superfluous.)

Steps to reproduce:

1. Launch Etoys
2. Switch to Journal
3. Activate Frame
4. From Frame Etoys panel, select Stop
   • Notice that Etoys icon remains
5. From Frame Etoys panel, select Resume
   • Notice that you are switched to tty1 and then liveuser login

[bert]

The Sugar crash in step 5 is worrying, but I cannot reproduce that crash in a recent SoaS beta running in Virtualbox, and in any case Sugar should not crash even if an activity was buggy. Assigning back to tomeu.

I can confirm the apparent hang in step 4. But once I activate Etoys (wether by clicking resume or pressing alt-tab) it stops normally. The likely cause for that is that the window-close event is not processed while Etoys is deactivated (by the Sugar D-Bus activate() call). Only once Etoys is activated again, the event is received and then the saving and exit is performed. This should eventually be fixed but I'd not consider it critical, stopping an activity while it is inactive is not a too frequent operation. I filed an upstream report nonetheless:
​http://tracker.squeakland.org/browse/SQ-233

[FGrose]

shell log after sugar crash

[FGrose]

Sugar crash observed with soas-beta.iso (14Apr2009) installed in !VirtualBox2.2.2 disc with onboard livecd-iso-to-disk onto this hardware, ​http://www.smolts.org/client/show/pub_f2cf5101-8203-4941-be03-408deedd03e5 on Win7-64 host.

shell log attached.

[FGrose]

Correction to !FGrose: VBox disc came from ​http://www.sugarlabs.org/static/soas/soas-beta-1.zip (20Apr2009), run in !VirtualBox2.2.2 on this hardware, ​http://www.smolts.org/client/show/pub_f2cf5101-8203-4941-be03-408deedd03e5 on Win7-64 host.

[bert]

This is odd. I used the same VDI (created 20 Apr 2009) and VirtualBox 2.2.2, although on a Mac. Does not crash here.

[bert]

shell log with evidence of Sugar crash (BadWindow error)

[bert]

I could repro now - it only crashes when starting Etoys freshly, not when resuming an Etoys project (which is the default now). Sugar crashes with a BadWindow X error, presumably because in that case Etoys exits immediately when it receives the activate(true) call.

[FGrose]

slim.log after sugar crash

[FGrose]

Reproduced also with soas-beta.iso (14Apr2009) booted directly from .iso in !VirtualBox2.2.2 on Jackalope64 host.

slim.log attached

[bert]

FWIW it does not crash in Sugar 0.82 (tested on an XO at build 800).

[tomeu]

Fixed with help from Benjamin.

[sascha_silbe]

Bulk change distribution=SoaS -> component=SoaS