Smalltalk › Squeak › Squeak - Dev

Re: [Pharo-dev] Sign of Impending Iceberg Doom on Pharo 6.1

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

4 messages Options

Eliot Miranda-2

Re: [Pharo-dev] Sign of Impending Iceberg Doom on Pharo 6.1

Hi Sean,

On Fri, Aug 24, 2018 at 10:03 AM Sean P. DeNigris <[hidden email]> wrote:

Sometimes while working in Iceberg on 6.1, I see a progress bar(s) like the
following:
<http://forum.world.st/file/t128965/Screenshot_2018-08-24_12.jpeg>

This seems to usually signal that a crash is not far away. It's doubly
concerning because there is no clear way to stop the troubled process
(interrupts often bring up a debugger on something else) and because, since
one is now unable to save any code, it's unclear how to start over easily in
a new image. In an extreme scenario, I recently panicked during such an
experience and accidentally clicked "Save" instead of "Save As…" resulting
in an image that took hours worth of code (I know maybe not a great
practice) with it to the grave.

Any idea why this happens or what we might be able to do anything about it?

No. Yes. Indirectly I hope to help debug this. I am nearly finished with an error handling facility in the VM for FFI calls that will fail an FFI callout primitive with an error code indicating the kind of exception and the pc. This should at least stop the system from crashing and give us an improved ability to make sense of the exception; at least we'll be able to examine the arguments to the call. More details as I roll out the facility. It will require integrating some support code, rewriting FFI callout signatures to include an error code, etc.

p.s. I was so concerned that I immediately resolved to move to Pharo 7, but
that turned out not to be an option do to the recently reported bug with
extension methods (that don't exactly match the package name)

-----
Cheers,
Sean
--
Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html

_,,,^..^,,,_

best, Eliot

Sean P. DeNigris

Re: [Pharo-dev] Sign of Impending Iceberg Doom on Pharo 6.1

Administrator

Eliot Miranda-2 wrote
> I am nearly finished with an error handling facility...

Cool! Thanks for the heads up. I'll try not to kill any more images until
then ;-)

-----
Cheers,
Sean
--
Sent from: http://forum.world.st/Squeak-Dev-f45488.html

Cheers,
Sean

Ben Coman

Re: [Vm-dev] [Pharo-dev] Sign of Impending Iceberg Doom on Pharo 6.1

In reply to this post by Eliot Miranda-2

On Sat, 25 Aug 2018 at 01:40, Eliot Miranda <[hidden email]> wrote:

Hi Sean,
On Fri, Aug 24, 2018 at 10:03 AM Sean P. DeNigris <[hidden email]> wrote:
Sometimes while working in Iceberg on 6.1, I see a progress bar(s) like the
following:
<http://forum.world.st/file/t128965/Screenshot_2018-08-24_12.jpeg>

This seems to usually signal that a crash is not far away.

Do you mean seconds or minutes or more away from a crash?

It's doubly concerning because there is no clear way to stop the troubled process
(interrupts often bring up a debugger on something else) and because, since
one is now unable to save any code, it's unclear how to start over easily in
a new image.

Just some general ideas (you may have already considered)

The string "Diffing..." appears in only one location in MCPatch>>initializeWithBase:target:

I'm not clear on the behaviour of #on:displayProgress: whether its possible some inside loop

can make that string appear multiple times, or whether it indicates #initializeWithBase:target:

is being called multiple times.

A specific check there for an escape key-combo might help debug this.

Or a class variable reset from Iceberg that halts when it exceeds a limit on the number of loops.

Or log to file five or so levels of "thisContext senders senders etc" for post crash analysis to help isolate the loop.

cheers -ben

P.S. While I was browsing senders-of-senders-of-senders I quickly got lost and wondered

if a tool that graphically showed a senders-tree might be a useful demo for Diagrammer.

Sean P. DeNigris

Re: [Vm-dev] [Pharo-dev] Sign of Impending Iceberg Doom on Pharo 6.1

Administrator

Ben Coman wrote
> Do you mean seconds or minutes or more away from a crash?

It seems typically several minutes, but not to happen all the time and not
be easy to interrupt.

-----
Cheers,
Sean
--
Sent from: http://forum.world.st/Squeak-Dev-f45488.html

Cheers,
Sean