Re: [Bug] setup regression

[Christian Franke <Christian.Franke@t-online.de>]

Jon Turney wrote:
> On 27/09/2022 14:51, Christian Franke wrote:
>> Christian Franke wrote:
>>> ...
>>> I made the false assumption that default_version=empty in 
>>> set_action() always implies that the default version is not 
>>> accessible. This is not the case for packages selected for 
>>> installation before chooser is visible.
>>>
>>> I'm working on a new fix for the "Ctrl+I pressed but current version 
>>> is not accessible" case. ...
>>
>> See attached patch. It also fixes the same problem for the "Category" 
>> view. Testing shows that the problem only affects the display of the 
>> version number as the solver later silently removes such install 
>> requests.
>>
>>
>>> The correct logic is already in toggle_action(): Install the most 
>>> recent accessible non-test ('naively_preferred') version.
>>>
>>
>> I dropped this idea and aligned Ctrl+I behavior with "Install" from 
>> "Category" view instead. Toggle_action() behaves different in such 
>> corner cases as it always installs something.
>
> Thanks, I applied both patches.
>
> Do you have a simple way of causing a non-accessible package for 
> testing with?

Unfortunately there is only a simple one if the current installation and 
cache directory matches some conditions. The "Ignore install requests if 
version is not accessible" case appears if:

- "Install from Local Directory" is selected.
- A previous version of a no longer installed package is still in the 
cache directory. Otherwise the package would not appear in the chooser.
- The current version of this package is not in the cache directory. The 
"Size" column shows "?" for such packages.

Without commit 63a2c90, Ctrl+I or "Install" from "Category" view shows 
an empty version in "New" column. The "Install" from "Category" case 
could also be reproduced with setup 2.919.

With this commit, the Install request does nothing.