64-bit bug(s)

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

64-bit bug(s)

Eliot Miranda-2
Hi Timothy, Hi All,

    I know some people have seen UI bugs with the 64-bit VM and image.  This is just to report that I see one very obvious bug.  In a recent messages browser on 64-bits any attempt to select a method other than the first in the list works, but immediately the selection reverts to the first element in the list.  I suspect something in the event chain.  But it's nice to have an obvious bug to chase instead of the much harder visual bugs people have been reporting up until now. Hopefully the same root cause will account for both bugs.  Fingers crossed.

_,,,^..^,,,_
best, Eliot


tty
Reply | Threaded
Open this post in threaded view
|

Re: 64-bit bug(s)

tty
Hi Eliot.

First thank you, thank you, thank you, for making my dream of developing on Squeak on a 64x64 bit system a reality.

The UI bug I had was from a custom-compiled VM. The cause was a rectangle such as 1@1 evaluating to random reals

1@1 inspect => 0.79@0.02
(exagerated for effect)

The effect was a "Modern Art" layout manager with no recognizable perspective and a refusal to abide by the norms of established practice. Picasso-ish.

When I upgraded to a pre-built VM from your site, the "fun rectangles" went away and where replaced with the "predictable rectangles" we programmer types like to depend on.

cheers,

tty

---- On Fri, 17 Jun 2016 16:49:32 -0400 Eliot Miranda<[hidden email]> wrote ----
Hi Timothy, Hi All,

    I know some people have seen UI bugs with the 64-bit VM and image.  This is just to report that I see one very obvious bug.  In a recent messages browser on 64-bits any attempt to select a method other than the first in the list works, but immediately the selection reverts to the first element in the list.  I suspect something in the event chain.  But it's nice to have an obvious bug to chase instead of the much harder visual bugs people have been reporting up until now. Hopefully the same root cause will account for both bugs.  Fingers crossed.

_,,,^..^,,,_
best, Eliot





Reply | Threaded
Open this post in threaded view
|

Re: 64-bit bug(s)

marcel.taeumel
In reply to this post by Eliot Miranda-2
Eliot Miranda-2 wrote
Hi Timothy, Hi All,

    I know some people have seen UI bugs with the 64-bit VM and image.
This is just to report that I see one very obvious bug.  In a recent
messages browser on 64-bits any attempt to select a method other than the
first in the list works, but immediately the selection reverts to the first
element in the list.  I suspect something in the event chain.  But it's
nice to have an obvious bug to chase instead of the much harder visual bugs
people have been reporting up until now. Hopefully the same root cause will
account for both bugs.  Fingers crossed.

_,,,^..^,,,_
best, Eliot
Hi Eliot,

once we have a 64-bit VM for Windows, I can help. :)

Best,
Marcel
Reply | Threaded
Open this post in threaded view
|

Re: 64-bit bug(s)

marcel.taeumel
In reply to this post by Eliot Miranda-2
Eliot Miranda-2 wrote
Hi Timothy, Hi All,

    I know some people have seen UI bugs with the 64-bit VM and image.
This is just to report that I see one very obvious bug.  In a recent
messages browser on 64-bits any attempt to select a method other than the
first in the list works, but immediately the selection reverts to the first
element in the list.  I suspect something in the event chain.  But it's
nice to have an obvious bug to chase instead of the much harder visual bugs
people have been reporting up until now. Hopefully the same root cause will
account for both bugs.  Fingers crossed.

_,,,^..^,,,_
best, Eliot
Don't we have tests for numbers and rectangles? Do they pass?

A quick look at the selection mechanism in lists revealed these calls:
PluggableListMorph >> #rowAtLocation: (with a coordinate transformation)
LazyListMorph >> #rowAtLocation: (with calls to #min: #min:max: etc.)

You could write event positions to the transcript.

Best,
Marcel
Reply | Threaded
Open this post in threaded view
|

Re: 64-bit bug(s)

Eliot Miranda-2
In reply to this post by tty
Tim,

On Jun 17, 2016, at 2:04 PM, gettimothy <[hidden email]> wrote:

Hi Eliot.

First thank you, thank you, thank you, for making my dream of developing on Squeak on a 64x64 bit system a reality.

The UI bug I had was from a custom-compiled VM. The cause was a rectangle such as 1@1 evaluating to random reals

1@1 inspect => 0.79@0.02
(exagerated for effect)

The effect was a "Modern Art" layout manager with no recognizable perspective and a refusal to abide by the norms of established practice. Picasso-ish.

When I upgraded to a pre-built VM from your site, the "fun rectangles" went away and where replaced with the "predictable rectangles" we programmer types like to depend on.

ROTLF!

and there's a moral in there somewhere too :-)

cheers,

tty

---- On Fri, 17 Jun 2016 16:49:32 -0400 Eliot Miranda<[hidden email]> wrote ----
Hi Timothy, Hi All,

    I know some people have seen UI bugs with the 64-bit VM and image.  This is just to report that I see one very obvious bug.  In a recent messages browser on 64-bits any attempt to select a method other than the first in the list works, but immediately the selection reverts to the first element in the list.  I suspect something in the event chain.  But it's nice to have an obvious bug to chase instead of the much harder visual bugs people have been reporting up until now. Hopefully the same root cause will account for both bugs.  Fingers crossed.

_,,,^..^,,,_
best, Eliot






Reply | Threaded
Open this post in threaded view
|

Re: 64-bit bug(s)

Eliot Miranda-2
In reply to this post by marcel.taeumel
Hi Marcel,


> On Jun 17, 2016, at 2:12 PM, marcel.taeumel <[hidden email]> wrote:
>
> Eliot Miranda-2 wrote
>> Hi Timothy, Hi All,
>>
>>    I know some people have seen UI bugs with the 64-bit VM and image.
>> This is just to report that I see one very obvious bug.  In a recent
>> messages browser on 64-bits any attempt to select a method other than the
>> first in the list works, but immediately the selection reverts to the
>> first
>> element in the list.  I suspect something in the event chain.  But it's
>> nice to have an obvious bug to chase instead of the much harder visual
>> bugs
>> people have been reporting up until now. Hopefully the same root cause
>> will
>> account for both bugs.  Fingers crossed.
>>
>> _,,,^..^,,,_
>> best, Eliot
>
> Don't we have tests for numbers and rectangles? Do they pass?
>
> A quick look at the selection mechanism in lists revealed these calls:
> PluggableListMorph >> #rowAtLocation: (with a coordinate transformation)
> LazyListMorph >> #rowAtLocation: (with calls to #min: #min:max: etc.)
>
> You could write event positions to the transcript.

Well the annoying thing is that the bug has disappeared and the #%^*£> recent submissions browser is working fine.  Sigh...

>
> Best,
> Marcel
>
>
>
> --
> View this message in context: http://forum.world.st/64-bit-bug-s-tp4901548p4901567.html
> Sent from the Squeak - Dev mailing list archive at Nabble.com.
>