Vadim Belman
Vadim Belman
The downside of the current implementation is broken `zef` and, likely, other modules having their own distribution classes. This happens because of a new required method `modified` they have to...
**UPD** Replacing `method modified {...}` with `method modified { fail ... }` does make `zef` compile and work. Since, as I said, `Zef::Distribution` wouldn't make into CUR internals, nothing else...
@ugexe But in order for the task to be solved the modification time is necessary. Yet, unless I add the method to the `Distribution` and implement it for other internal...
I consider the distribution in more generic way where its source is not only a file system. Say, it could be a centralized network resource. META6.json may not even exists...
BTW, thinking a bit more about it, the default `modification` method can simply return epoch and it should be totally safe.
CURFS doesn't work for a local development directory (i.e. `raku -I.`). It already includes ``, but its modification doesn't result in re-compilation. `Distribution::Path` can have the `modified` method on its...
> It also wouldn't surprise me if the solution can be implemented inside of I don't even know where exactly it is used. Will consider. Speaking about keeping `Distribution` simple,...
Ok, I have reduced it to the absolute minimum now. :)
@niner @ugexe ping for re-review of checksum-based re-implementation.
> An API that only provides a yes/no answer on if its an identical distribution would be a safer public API to expose because the implementation details of that are...