inescapable modal dialog in trunk

Well, it appears the problem is present in the production 5.2 image
with stock Preference settings too!   :(

Here are exact steps to reproduce using the production 5.2 image:

1) launch Squeak5.2.

2) Click "Skip" to remove the "Welcome To Squeak" banner.  Close the
"Welcome To Squeak" window, too.

3) Click in the search bar in the upper right, type "asFloat" (without
quotes), and press [Return].  The window showing implementors is shown
with the first method selected.

4) Make it dirty.  Click in the text pane and insert a space.

5) Click on the second method.  Squeak displays, "Changes have not been
saved, is it okay to cancel those changes?"

6) Click "No".

7) Now try to click on the desktop.  The dialog keeps coming back.
On Wed, Nov 7, 2018 at 12:22 PM tim Rowledge <[hidden email]> wrote:
>
>
>
> > On 2018-11-07, at 1:45 AM, K K Subbu <[hidden email]> wrote:
> >
> > On 07/11/18 9:38 AM, Chris Muller wrote:
> >> 1.  Make a method in a browser dirty.
> >> 2.  Click on another method.  Squeak displays, "Changes have not been
> >> saved, is it okay to cancel those changes?"
> >> 3.  Click "No".
> >> 4.  Now try click anywhere else or do anything else in other windows
> >> or the desktop.  Every click anywhere causes the dialog to reappear
> >> until you say "Yes" or cancel the method changes.
> > Chris,
> >
> > I am unable to reproduce this problem. In Step 4, I am able to open the global menu or a workspace and continue working on other tasks without having the modal dialog pop up.
>
> Same thing on 5.2 32bit Pi - no problem with the dialogue at all.
>
> Now an obvious thing is that my set of preferences is almost certainly very different to Chris' and some of those settings may have a relevant effect. I don't use smart splitters for example and maybe something related to that is interacting with the mouse presses; no idea.
>
>
> tim
> --
> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
> Useful random insult:- If she was any dumber, she'd be a green plant.
>
>
>

Hi, there.

At the time of writing, we have:

A) dialogs that close on an "outside click" (e.g., class serach in system browser)

B) dialogs that insist on being treated on a globally exclusive level (e.g., save before quit) -- wiggle wiggle :-)

C) dialogs/windows that are modally attached to their spawning window (e.g., font chooser in text fields)

All these variations have their pros and cons. Yet, we cannot reduce all scenarios to only one version. Programmers should refrain from using B too often. Only for system-wide interruptions. A and C are what seems to be less distracting for the user.

Maybe we need another variation of dialogs:

D) dialogs that are scoped to their spawning window AND restrict input to that spawning window (effectively a variation of C)

Best,

Marcel

Am 19.12.2018 05:10:44 schrieb Eliot Miranda <[hidden email]>:

> On Dec 18, 2018, at 2:20 PM, Chris Muller wrote:
>
> Hi Tim,
>
>> Possibly related functionally but certainly by subject area -
>> the modal dialogues I build for the file chooser etc are 'violated' by a yellowbuttonmenu IF you haven't turned off yellowbutton menus. Not noticed before because my normal preferences do just that; this time I tried some stuff in a very vanilla image and spotted the problem.
>>
>> Open a file chooser. cClick about a bit to make sure it is live. Click yellow button; you can now wander off and click in other windows - you've killed the modal nature of the dialog. You can 'return' by simply clicking in the dialog again.
>
> I'm glad you put 'violated' in quotes like that, since there can be
> other scopes of modality than global. For example, it could extend
> from only the window which spawned it (the ideal), OR (going in the
> other direction to the extreme), what if opening up a FileDialog in
> your Squeak image not only prevented you from interacting with other
> Squeak windows but all other OS windows, too? (Impossible, I know,
> but a question for conceptual illustration..). IMO, any scope beyond
> the minimum scope seems arbitarily restricting.
>
> I think we should consider making all modal dialogs scoped only to the
> spawner.

+10

> As long as the dialog selection works and continues to
> process upon the selection, this sounds like a feature. Because even
> after the user has opened the dialog and navigated to the directory
> they want, if they find different file-names than they expected and
> need to go look elsewhere in the image to confirm, -- it seems
> unnecessary to make them cancel out and go through all of that again.
>
> Best,
> Chris
>
>
>> (I just noticed that using the halo does the same.)
>> Turn off 'generalizedYellowButtonMenu' in the preferences and you don't get the (non-halo) problem.
>>
>>
>> tim
>> --
>> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
>> Spell checkers at maximum! Fire!
>>
>>
>>
>

Hi Eliot,

are you referring to this?

Best,

Marcel

Am 20.12.2018 01:31:24 schrieb Eliot Miranda <[hidden email]>:

On Wed, Dec 19, 2018 at 1:46 PM karl ramberg <[hidden email]> wrote:

Also the two different dialogs that pop up for unknown variable when saving a method cause some dissonance.

Two totally different dialogs for doing basically the same thing. 

See picture examples.

Plus IIRC that if there's a subsequent syntax error the variable declarations are lost.  We really should do one edit and then automatically launch a recompile (or compile as much as we can and collect multiple undeclared together in one dialogue?)

Cheers,

Karl

On Wed, Dec 19, 2018 at 8:33 AM Marcel Taeumel <[hidden email]> wrote:

Hi, there.

At the time of writing, we have:

A) dialogs that close on an "outside click" (e.g., class serach in system browser)

B) dialogs that insist on being treated on a globally exclusive level (e.g., save before quit) -- wiggle wiggle :-)

C) dialogs/windows that are modally attached to their spawning window (e.g., font chooser in text fields)

All these variations have their pros and cons. Yet, we cannot reduce all scenarios to only one version. Programmers should refrain from using B too often. Only for system-wide interruptions. A and C are what seems to be less distracting for the user.

Maybe we need another variation of dialogs:

D) dialogs that are scoped to their spawning window AND restrict input to that spawning window (effectively a variation of C)

Best,

Marcel

Am 19.12.2018 05:10:44 schrieb Eliot Miranda <[hidden email]>:

> On Dec 18, 2018, at 2:20 PM, Chris Muller wrote:
>
> Hi Tim,
>
>> Possibly related functionally but certainly by subject area -
>> the modal dialogues I build for the file chooser etc are 'violated' by a yellowbuttonmenu IF you haven't turned off yellowbutton menus. Not noticed before because my normal preferences do just that; this time I tried some stuff in a very vanilla image and spotted the problem.
>>
>> Open a file chooser. cClick about a bit to make sure it is live. Click yellow button; you can now wander off and click in other windows - you've killed the modal nature of the dialog. You can 'return' by simply clicking in the dialog again.
>
> I'm glad you put 'violated' in quotes like that, since there can be
> other scopes of modality than global. For example, it could extend
> from only the window which spawned it (the ideal), OR (going in the
> other direction to the extreme), what if opening up a FileDialog in
> your Squeak image not only prevented you from interacting with other
> Squeak windows but all other OS windows, too? (Impossible, I know,
> but a question for conceptual illustration..). IMO, any scope beyond
> the minimum scope seems arbitarily restricting.
>
> I think we should consider making all modal dialogs scoped only to the
> spawner.

+10

> As long as the dialog selection works and continues to
> process upon the selection, this sounds like a feature. Because even
> after the user has opened the dialog and navigated to the directory
> they want, if they find different file-names than they expected and
> need to go look elsewhere in the image to confirm, -- it seems
> unnecessary to make them cancel out and go through all of that again.
>
> Best,
> Chris
>
>
>> (I just noticed that using the halo does the same.)
>> Turn off 'generalizedYellowButtonMenu' in the preferences and you don't get the (non-halo) problem.
>>
>>
>> tim
>> --
>> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
>> Spell checkers at maximum! Fire!
>>
>>
>>
>

--

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

best, Eliot

Hi Karl,

yeah, some projects (or code) use "list chooser" for "yes/no" stuff. That's why we had to come up with this mixed versions of the dialog for list choosing. So: backwards compatibility.

Or maybe it is still a good idea to present buttons for smaller lists of stuff. Like we do now.

Best,

Marcel

Am 19.12.2018 22:46:56 schrieb karl ramberg <[hidden email]>:

Also the two different dialogs that pop up for unknown variable when saving a method cause some dissonance.

Two totally different dialogs for doing basically the same thing. 

See picture examples.

Cheers,

Karl

On Wed, Dec 19, 2018 at 8:33 AM Marcel Taeumel <[hidden email]> wrote:

Hi, there.

At the time of writing, we have:

A) dialogs that close on an "outside click" (e.g., class serach in system browser)

B) dialogs that insist on being treated on a globally exclusive level (e.g., save before quit) -- wiggle wiggle :-)

C) dialogs/windows that are modally attached to their spawning window (e.g., font chooser in text fields)

All these variations have their pros and cons. Yet, we cannot reduce all scenarios to only one version. Programmers should refrain from using B too often. Only for system-wide interruptions. A and C are what seems to be less distracting for the user.

Maybe we need another variation of dialogs:

D) dialogs that are scoped to their spawning window AND restrict input to that spawning window (effectively a variation of C)

Best,

Marcel

Am 19.12.2018 05:10:44 schrieb Eliot Miranda <[hidden email]>:

> On Dec 18, 2018, at 2:20 PM, Chris Muller wrote:
>
> Hi Tim,
>
>> Possibly related functionally but certainly by subject area -
>> the modal dialogues I build for the file chooser etc are 'violated' by a yellowbuttonmenu IF you haven't turned off yellowbutton menus. Not noticed before because my normal preferences do just that; this time I tried some stuff in a very vanilla image and spotted the problem.
>>
>> Open a file chooser. cClick about a bit to make sure it is live. Click yellow button; you can now wander off and click in other windows - you've killed the modal nature of the dialog. You can 'return' by simply clicking in the dialog again.
>
> I'm glad you put 'violated' in quotes like that, since there can be
> other scopes of modality than global. For example, it could extend
> from only the window which spawned it (the ideal), OR (going in the
> other direction to the extreme), what if opening up a FileDialog in
> your Squeak image not only prevented you from interacting with other
> Squeak windows but all other OS windows, too? (Impossible, I know,
> but a question for conceptual illustration..). IMO, any scope beyond
> the minimum scope seems arbitarily restricting.
>
> I think we should consider making all modal dialogs scoped only to the
> spawner.

+10

> As long as the dialog selection works and continues to
> process upon the selection, this sounds like a feature. Because even
> after the user has opened the dialog and navigated to the directory
> they want, if they find different file-names than they expected and
> need to go look elsewhere in the image to confirm, -- it seems
> unnecessary to make them cancel out and go through all of that again.
>
> Best,
> Chris
>
>
>> (I just noticed that using the halo does the same.)
>> Turn off 'generalizedYellowButtonMenu' in the preferences and you don't get the (non-halo) problem.
>>
>>
>> tim
>> --
>> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
>> Spell checkers at maximum! Fire!
>>
>>
>>
>