Smalltalk › Instantiations

floating point exception in DateAndTime now class

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

6 messages Options

Denis Ermolin

floating point exception in DateAndTime now class

I am encountering a problem with date and time class on some machines (not everyone is experiencing the problem).

This is from the dev image:

if I try to convert the large integer to float in my workspace, it works fine, but during execution the conversion to float fails on some machines.

Anyone help is appreciated!

--
You received this message because you are subscribed to the Google Groups "VA Smalltalk" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [hidden email].
To post to this group, send email to [hidden email].
Visit this group at https://groups.google.com/group/va-smalltalk.
For more options, visit https://groups.google.com/d/optout.

Seth Berman

Re: floating point exception in DateAndTime now class

Hello Denis,

What version/bitness of the product? Judging by the font in the debugger, I'm going to guess something < 8.6.1?

-- Seth

On Tuesday, December 19, 2017 at 10:03:45 AM UTC-5, Denis na wrote:

I am encountering a problem with date and time class on some machines (not everyone is experiencing the problem).

This is from the dev image:

if I try to convert the large integer to float in my workspace, it works fine, but during execution the conversion to float fails on some machines.

Anyone help is appreciated!

Denis Ermolin

Re: floating point exception in DateAndTime now class

8.5.2 , I think its 32 bit

On Tuesday, December 19, 2017 at 12:33:08 PM UTC-5, Seth Berman wrote:

Hello Denis,

What version/bitness of the product? Judging by the font in the debugger, I'm going to guess something < 8.6.1?

-- Seth

On Tuesday, December 19, 2017 at 10:03:45 AM UTC-5, Denis na wrote:
I am encountering a problem with date and time class on some machines (not everyone is experiencing the problem).

This is from the dev image:

if I try to convert the large integer to float in my workspace, it works fine, but during execution the conversion to float fails on some machines.

Anyone help is appreciated!

Seth Berman

Re: floating point exception in DateAndTime now class

Greetings,

It sounds like a red-herring of sorts and that the actual floating point exception happened earlier.

On x86, the FPE might not be triggered until a subsequent floating point (native) instruction. So it may just be showing up here...but may not actually originate here.

The symptom you are reporting would seem to indicate this.

I'm assuming the argument large integer (which I can't fully see in the screenshot) can execute #asFloat just fine all day long?

So either the the large integer that got #asFloat sent to it, in all appearances, seems to be fine...but may internally have some trash bits set

An FPE occurred earlier...and your just seeing the hardware exception being triggered here.

-- Seth

On Tuesday, December 19, 2017 at 1:52:35 PM UTC-5, Denis na wrote:

8.5.2 , I think its 32 bit

On Tuesday, December 19, 2017 at 12:33:08 PM UTC-5, Seth Berman wrote:
Hello Denis,

What version/bitness of the product? Judging by the font in the debugger, I'm going to guess something < 8.6.1?

-- Seth

On Tuesday, December 19, 2017 at 10:03:45 AM UTC-5, Denis na wrote:
I am encountering a problem with date and time class on some machines (not everyone is experiencing the problem).

This is from the dev image:

if I try to convert the large integer to float in my workspace, it works fine, but during execution the conversion to float fails on some machines.

Anyone help is appreciated!

Denis Ermolin

Re: floating point exception in DateAndTime now class

Thank you for your answer. Here is more info , looks like windows call is failing calling user32.dll

On Tuesday, December 19, 2017 at 2:23:32 PM UTC-5, Seth Berman wrote:

Greetings,

It sounds like a red-herring of sorts and that the actual floating point exception happened earlier.
On x86, the FPE might not be triggered until a subsequent floating point (native) instruction. So it may just be showing up here...but may not actually originate here.
The symptom you are reporting would seem to indicate this.

I'm assuming the argument large integer (which I can't fully see in the screenshot) can execute #asFloat just fine all day long?

So either the the large integer that got #asFloat sent to it, in all appearances, seems to be fine...but may internally have some trash bits set
or
An FPE occurred earlier...and your just seeing the hardware exception being triggered here.

-- Seth

On Tuesday, December 19, 2017 at 1:52:35 PM UTC-5, Denis na wrote:
8.5.2 , I think its 32 bit

On Tuesday, December 19, 2017 at 12:33:08 PM UTC-5, Seth Berman wrote:
Hello Denis,

What version/bitness of the product? Judging by the font in the debugger, I'm going to guess something < 8.6.1?

-- Seth

On Tuesday, December 19, 2017 at 10:03:45 AM UTC-5, Denis na wrote:
I am encountering a problem with date and time class on some machines (not everyone is experiencing the problem).

This is from the dev image:

if I try to convert the large integer to float in my workspace, it works fine, but during execution the conversion to float fails on some machines.

Anyone help is appreciated!

12-19-2017 4-54-29 PM.png (420K) Download Attachment

Seth Berman

Re: floating point exception in DateAndTime now class

Hello Denis,

I recognize that and have been down this rabbit hole...that is another red-herring. There is some float instruction in the windows CallWindowProc...but that is not the source.

Check if your doing any call-outs to other C libraries (3rd party functionality?)...the likely culprit is somewhere in those.

- Seth

On Tuesday, December 19, 2017 at 4:56:16 PM UTC-5, Denis na wrote:

Thank you for your answer. Here is more info , looks like windows call is failing calling user32.dll

On Tuesday, December 19, 2017 at 2:23:32 PM UTC-5, Seth Berman wrote:
Greetings,

It sounds like a red-herring of sorts and that the actual floating point exception happened earlier.
On x86, the FPE might not be triggered until a subsequent floating point (native) instruction. So it may just be showing up here...but may not actually originate here.
The symptom you are reporting would seem to indicate this.

I'm assuming the argument large integer (which I can't fully see in the screenshot) can execute #asFloat just fine all day long?

So either the the large integer that got #asFloat sent to it, in all appearances, seems to be fine...but may internally have some trash bits set
or
An FPE occurred earlier...and your just seeing the hardware exception being triggered here.

-- Seth

On Tuesday, December 19, 2017 at 1:52:35 PM UTC-5, Denis na wrote:
8.5.2 , I think its 32 bit

On Tuesday, December 19, 2017 at 12:33:08 PM UTC-5, Seth Berman wrote:
Hello Denis,

What version/bitness of the product? Judging by the font in the debugger, I'm going to guess something < 8.6.1?

-- Seth

On Tuesday, December 19, 2017 at 10:03:45 AM UTC-5, Denis na wrote:
I am encountering a problem with date and time class on some machines (not everyone is experiencing the problem).

This is from the dev image:

if I try to convert the large integer to float in my workspace, it works fine, but during execution the conversion to float fails on some machines.

Anyone help is appreciated!