Last update broke browsers...

Ok,

I started with a fresh image (Squeak3.10.2-Trunk-091128.zip) and
updated. Then I reinstalled some packages (FFI, OSProcess, OSProcessIO,
CommandShell, Balloon3D) and the problem disapeared.

Last image had installed VMMaker, Speech and other packages. I'll try to
reinstall everything in order to see where the problem arises).

On Mon, Dec 28, 2009 at 01:22:05AM -0200, Casimiro de Almeida Barreto wrote:
> Ok,
>
> I started with a fresh image (Squeak3.10.2-Trunk-091128.zip) and
> updated. Then I reinstalled some packages (FFI, OSProcess, OSProcessIO,
> CommandShell, Balloon3D) and the problem disapeared.
>
> Last image had installed VMMaker, Speech and other packages. I'll try to
> reinstall everything in order to see where the problem arises).

There is definitely a problem somewhere. I have a broken image with the
same symptoms that you are reporting. In my case, it happened when I was
trying to save an update to the squeak trunk, so I was expecting that it
was something related to those changes (the large changes file updates).
But maybe that was just a coincidence.

In any case, I can confirm that this same problem has happened one time
to me. Unfortunately that I cannot reproduce it in other images, and it
is hard to figure out what is wrong with the broken image because all the
debugger windows have the same "red X" failure.

Dave

I saw some code trapping Exception in the MultiDisplayScanner family.
That might be a source of problems.

2009/12/28 David T. Lewis <[hidden email]>:

David T. Lewis wrote:
> In any case, I can confirm that this same problem has happened one time
> to me. Unfortunately that I cannot reproduce it in other images, and it
> is hard to figure out what is wrong with the broken image because all the
> debugger windows have the same "red X" failure.

Try moving .sources and .changes out of the way. If it's in any way
related to source file handling (I'm not convinced of that), this should
avoid using them.

Cheers,
   - Andreas

On Mon, Dec 28, 2009 at 02:03:12PM +0100, Andreas Raab wrote:
> David T. Lewis wrote:
> >In any case, I can confirm that this same problem has happened one time
> >to me. Unfortunately that I cannot reproduce it in other images, and it
> >is hard to figure out what is wrong with the broken image because all the
> >debugger windows have the same "red X" failure.
>
> Try moving .sources and .changes out of the way. If it's in any way
> related to source file handling (I'm not convinced of that), this should
> avoid using them.

I tried running the broken image with .sources and .changes out of
the way, and there is not change to the symptoms.

On Mon, Dec 28, 2009 at 11:10:13AM +0100, Nicolas Cellier wrote:
> I saw some code trapping Exception in the MultiDisplayScanner family.
> That might be a source of problems.

Aha!

The SqueakDebug.log points at MultiDisplayScanner, so I took a look
at recent changes there. Some methods were out of sync with the
trunk repository, but should not have been. I reloaded the package,
and the image is no longer broken.

So I think that I must have corrupted my changes file (and some
compiled methods) for this image during my earlier work on the large
changes files, and this produced the red X problem later on.

My only remaining concern is that Casimiro de Almeida Barreto reports
similar sounding symptoms in his broken image. Hopefully it's just
a coincidence.

Dave

Em 28-12-2009 14:51, David T. Lewis escreveu: return Boolean...

MultiDisplayScanner(Object)>>error: (NonBooleanReceiver: procceed for truth)
[] in WorldState>>displayWorldSafely
BlockClosure>>valueWithPossibleArgs:
[] in BlockClosure>>ifError:
BlockClosure>>valueWithPossibleArgs:
[] in MethodContext(ContextPart)>>handleSignal:
BlockClosure>>ensure:
MethodContext(ContextPart)>>handleSignal:
NonBooleanReceiver(Exception)>>signal
NonBooleanReceiver(Exception)>>signal:
MultiDisplayScanner(Object)>>mustBeBooleanIn:
MultiDisplayScanner(Object)>>mustBeBoolean
MultiDisplayScanner>>displayLine:offset:leftInRun:
MultiNewParagraph>>displayOn:using:at:
FormCanvas>>paragraph:bounds:color:
TextMorphForEditView(TextMorph)>>drawOn:
FormCanvas(Canvas)>>draw:
FormCanvas(Canvas)>>drawMorph:
[] in TextMorphForEditView(TextMorph)>>fullDrawOn:
FromCanvas>>roundCornersOf:in:during:

2009/12/28 Casimiro de Almeida Barreto <[hidden email]>:
The only candidate I see is:
    done := self perform: stopCondition.
If performing the stopCondition does not return a Boolean, you'll
trigger the Error...
But stopConditions are made of #endOfRun #embeddedObject #crossedX #cr
#tab #paddedSpace (#space not used in MultiCharacterScanner).
Plus #columnBreak.
And all these messages return true or false in MultiCharacterScanner
hierarchy, so I don't see any obvious reason...

Nicolas