[latex3-commits] [l3svn] 02/04: "deprecated" here causes an error ... [ci skip]

[Bruno Le Floch]

On 05/13/2017 01:47 PM, Bruno Le Floch wrote:
> On 05/13/2017 12:44 PM, Joseph Wright wrote:
>> On 13/05/2017 16:13, Bruno Le Floch wrote:
>>>> I wanted to ship and the logic wasn't clear, so I've 'forced' the issue.
>>>> To me, putting in the date _when something is deprecated_ seems logical:
>>>> the date for removal is a policy decision that should appear at most
>>>> once in any code (in l3doc).
>>>>
>>>> Joseph
>>>>
>>>
>>> While I agree "deprecated = <date-where-it-will-be-removed>" is
>>> illogical, I don't understand in what sense anything is duplicated by
>>> doing that.  What markup would you like?  I guess one option would be to do
>>>
>>>    \begin{macro}[deprecated = 2017-04-05]{...}
>>>
>>> and make l3doc complain if the current year is > (year+1).
>>
>> What's duplicated is the decision about the time gap between 'marked for
>> removal' and 'actually should be removed'. If you put in the date of the
>> 'marked for removal' case, any change in the time between that and the
>> second case is only in l3doc. On the other hand, if you mark with 'date
>> actually to be removed', any change of policy means tracking each one
>> down and adjusting by hand.
>>
>>> By the way, I'm not complaining about what you did removing the date,
>>> just trying to figure out how to set things up better.
>>
>> Understood.
>>
>> Joseph
> 
> Ok, I'll change to date deprecated and make l3doc complain if current
> year > year+1.  So something deprecated now is to be deleted at the end
> of 2018.  Expect commits soon.
> 
> Bruno
> 

Actually it's trickier than expected (and my time's up): several places
need to know deprecation or removal dates.

- l3doc needs to know when something will be removed,
- deprecated functions need to know their deprecation date to put up a
warning,
- l3deprecation needs to know removed functions' removal date, and
deprecated functions' future removal date (so that I can use the same
auxiliary for both removed and deprecated functions when
\deprecation_error: is used).

Right now I don't see how to avoid writing down explicitly the future
removal date in l3deprecation.  There is probably some obvious solution
I'm missing.

Bruno