segmentation fault on startup with no stack printed.... any way to recover the image?

My laptop battery ran down when I was not paying attention.  I saved the
image at about 6PM last night.  The modification date on the image and
changes file is 3:15AM.  I'm guessing that is when the battery died.
When I attempt to start the image (with cog or the interpreter vm) I get

Segmentation fault (core dumped)


It does not print anything to the PharoDebug.log.  Is there a process to
follow to attempt to recover the image?  If I drag the changes file onto
another clean Pharo 1.4 image, I'm not able to go back as far as I'd
like.  Ideally I would be able to get all the changes between the base
Pharo 1.4 image and the image I lost.


Thanks for any guidance you can provide

Hi paul

did you try to receover the changes?
we should improve this part too
did you try the following:
        - take a fresh 1.4
        - rename your change file as the change of the new version
        - and do recover ?

Stef
On Mar 3, 2013, at 3:36 PM, Paul DeBruicker <[hidden email]> wrote:

Hi Stef,

If I use a base Pharo-1.4 image I can recover the changes so thanks for
pointing that out.  Is there a way to skip over the doIts that I
might've done in an 'explore' window during the 'recover changes' process?



If I use the broken image's changes file and a more recent image of mine
rather than the base Pharo 1.4 image I get this notice:

"this image has never been saved since changes were compressed"

Not sure what to make of that.   I thought I might be able to save some
work by using an older version of the broken image from a backup of my
laptop.


I don't have a debugging VM but this is what gdb outputs for the crash:

Program received signal SIGSEGV, Segmentation fault.
0x0000000000419835 in initializeInterpreter ()


Thanks again

Paul


On 03/03/2013 06:43 AM, stephane ducasse wrote:

On 03/03/2013 06:57 AM, Paul DeBruicker wrote:
> If I use the broken image's changes file and a more recent image of mine
> rather than the base Pharo 1.4 image I get this notice:
>
> "this image has never been saved since changes were compressed"
>
> Not sure what to make of that.   I thought I might be able to save some
> work by using an older version of the broken image from a backup of my
> laptop.


If I use the next older image of mine this message does not appear.  The
image with that notice had had 'cleanUpForRelease' run in it prior to
the its most recent save.

On Mar 3, 2013, at 3:57 PM, Paul DeBruicker <[hidden email]> wrote:

> Hi Stef,
>
> If I use a base Pharo-1.4 image I can recover the changes so thanks for
> pointing that out.  

ok even better.

> Is there a way to skip over the doIts that I
> might've done in an 'explore' window during the 'recover changes' process?

Normally you can mark (filter) the entries (now pay attention since class def are doit (not good) but we hope that the new
change model will help us addressing this change.
>
>
>
> If I use the broken image's changes file and a more recent image of mine
> rather than the base Pharo 1.4 image I get this notice:
>
> "this image has never been saved since changes were compressed"

Ok may be my approach was not that good :).
In VW we could receover from another change file and we should add that behavior to Pharo too.


> Not sure what to make of that.   I thought I might be able to save some
> work by using an older version of the broken image from a backup of my
> laptop.
>
>
> I don't have a debugging VM but this is what gdb outputs for the crash:
>
> Program received signal SIGSEGV, Segmentation fault.
> 0x0000000000419835 in initializeInterpreter ()

strange.
It should not crash.

Now eliot fixed several bugs related to become: and they were bringing a lot of instability to the system.
I hope that we will get to a stable point soon.

Stef

On Mar 3, 2013, at 4:10 PM, Paul DeBruicker <[hidden email]> wrote:

ok
for the release we reset a lot of behavior. So probably saving it once can help.