Hi Vincent,
does the image lock up when you execute Exception signal? The two functions you see looping are part of the implementation of primitives 195 & 196, findNextUnwindContextUpTo: (primitiveFindNextUnwindContext) & terminateTo: (primitiveTerminateTo). So if the image locks up the question is, is what is creating a circular structure of contexts, the VM or the image? I am willing to take a look at an image that contains the case. If you can create an image which enters the lop on start-up without my having to do anything I'm happy to take a look.
On Wed, Aug 19, 2015 at 3:22 PM, Vincent BLONDEAU < vincent.blondeau@polytech-lille.net> wrote:
Hi,
On last Moose image, I debugged the VM and executed “Exception signal”.
I paused the VM and saw that 2 methods are looping together in gcc3x-cointerp.c:
static sqInt findMethodWithPrimitiveFromContextUpToContext(sqInt primitive, sqInt senderContext, sqInt homeContext);
static sqInt findMethodWithPrimitiveFromFPUpToContext(sqInt primitive, char *startFP, sqInt homeContext);
The image is looping only when the debugger is launch on Error (not by launching it by the script below).
Maybe someone on the Pharo mailing list knows what is happening?
Thanks in advance,
Cheers,
Vincent
*De :* moose-dev-bounces@iam.unibe.ch [mailto: moose-dev-bounces@iam.unibe.ch] *De la part de* Andrei Chis *Envoyé :* mardi 18 août 2015 10:53 *À :* Moose-related development *Objet :* [Moose-dev] Re: Object new foo
Hi Vincent,
You example will also kill the Pharo debugger :)
The issue is that you pass 'Processor activeProcess' which gets terminated when you close the debugger.
When you open the debugger you need to make sure the process
on which it is opened is suspended and can get terminated with no problem.
A different way to redo the example would be:
context := Context
sender: nil receiver: {1 . 2 . 2} method: (OrderedCollection>>#add:) arguments: #(10).
process := Process
forContext: context priority: Processor userInterruptPriority.
GTGenericStackDebugger openOn: (DebugSession process: process context: context).
There is a basic test for opening the debugger in GTGenericStackDebuggerSmokeTest but more could help.
Cheers,
Andrei
On Mon, Aug 17, 2015 at 11:53 PM, Vincent BLONDEAU < vincent.blondeau@polytech-lille.net> wrote:
Indeed. Maybe some tests has to be added..
But for now, it seems that the debugger is crashing the image… Not just by launching it:
GTGenericStackDebugger openOn: (DebugSession process: Processor activeProcess context: (Context
sender: nil receiver: {1 . 2 . 2} method: (OrderedCollection>>#add: ) arguments: #(10)))
But when an exception is thrown, its seems that we fall into an infinite loop around the primitive 199. Maybe the context is not well set?
Cheers,
Vincent
*De :* moose-dev-bounces@iam.unibe.ch [mailto: moose-dev-bounces@iam.unibe.ch] *De la part de* Alexandre Bergel *Envoyé :* lundi 17 août 2015 13:26 *À :* Moose-related development *Objet :* [Moose-dev] Re: Object new foo
This is weird that such bug can be introduced and no alarm is raised
Alexandre
Le 17 août 2015 à 08:03, Jan Kurš kurs@iam.unibe.ch a écrit :
Yep I had this problem (falled back to previous version), I suspected debugger being the reason.
Cheers Jan
On Sun, 16 Aug 2015 03:05 Alexandre Bergel alexandre.bergel@me.com wrote:
in Moose 6.0 simply freezes. Anyone else sees this?
Alexandre
_,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
vm-dev@lists.squeakfoundation.org