Page 1 of 1

System clock bug (crash)

Posted: Tue Nov 17, 2015 5:08 pm
by xenic
I keep a digital clock (Sys:Utilities/Clock) on my Workbench toolbar and just discovered that if I turn off compositing in GUI prefs, the clock crashes when Workbench attempts to close and reopen.
To reproduce:

EDIT: Switch to 'kernal.debug' in Kickstart/Kicklayout and cold boot.
Turn on compositing in GUI prefs if it isn't already set.
Start SYS:Utilities/Clock.
Turn off compositing in GUI prefs and select "USE".
Clock crashes when Workbench attempts to reset.
There is no crash when using the normal kernal.

Can a beta tester confirm the problem and add it to the bug tracker?

Re: System clock bug (crash)

Posted: Tue Nov 17, 2015 5:17 pm
by Raziel
No betatester here, but it also doesn't crash.

Clock window gets reset to the top left underneath the WB bar.

No crash, nothing.
Treid with and without date, even "Saved settings".

Must be something else on your setup

Re: System clock bug (crash)

Posted: Tue Nov 17, 2015 6:18 pm
by xenic
Raziel wrote:No betatester here, but it also doesn't crash.

Clock window gets reset to the top left underneath the WB bar.

No crash, nothing.
Treid with and without date, even "Saved settings".

Must be something else on your setup
I'll need to investigate further. I'm using an X1000 with OS4.1FE. What hardware and OS version are you using?

EDIT: O.K. I forgot that I had switched to the debug kernal (Kickstart/kernal.debug) to debug another program. I tested with the normal kernal and there's no crash. Now I'm beginning to wonder if the program I'm working on is buggy when using the debug kernal or if there is something wrong with the debug kernal. If system programs like the clock crash with the debug kernal but not with the normal kernal then something is wrong somewhere.

Re: System clock bug (crash)

Posted: Wed Nov 18, 2015 6:58 am
by Raziel
I had a similar problem with a third party program crashing while the debug.kernel was in charge.

It was indeed a bug in the program which never showed up under normal conditions due to the normal kernel being more forgiving.
So, it is indeed worthy to get this into the bugtracker, as programs shouldn't crash under debug and mask that crash under normal.

Re: System clock bug (crash)

Posted: Wed Nov 18, 2015 12:22 pm
by tonyw
Take a look at the debug log or the GR output. Is the DEAR something like 0xCCCCCCCC? Is there a register value something like that?
If so, it is probably caused by the clock utility trying to remove a node twice from a list (which will only crash when using the debug kernel).

If I can reproduce it, I'll post a BZ for you.

[edit]
Yep, it's an easy one. I'll fix it myself (famous last words)...
[/edit]

[edit]
Bug #9385, fixed in version 53.2 (19.11.2015).
[/edit]

Re: System clock bug (crash)

Posted: Wed Nov 18, 2015 3:31 pm
by tboeckel
xenic wrote:I keep a digital clock (Sys:Utilities/Clock) on my Workbench toolbar and just discovered that if I turn off compositing in GUI prefs, the clock crashes when Workbench attempts to close and reopen.
No offence, but a simple bug report like "it crashes" is worth nothing. It is even more worthless without the corresponding crashlog. Lucky you Tony was able to reproduce this himself so quickly and knows how to fix it. But in general the crashlog is really essential! Why do developers have to beg for it again and again each time?

Re: System clock bug (crash)

Posted: Wed Nov 18, 2015 4:02 pm
by xenic
@tonyw
Yes, as a matter of fact register 'r0' contains 0xCCCCCCCC. Thanks for confirming the problem. Hopefully it is an easy fix.

Re: System clock bug (crash)

Posted: Fri Nov 20, 2015 10:01 am
by samo79
@tonyw

As you are here can you add an iconify gadget into the clock ? ;)

Re: System clock bug (crash)

Posted: Fri Nov 20, 2015 12:11 pm
by tonyw
No, the code is just too old, it would need a rewrite.