| 1 |
>>>>> On Sun, 12 Feb 2017, Alexis Ballier wrote: |
| 2 |
|
| 3 |
>> Or do you have any concrete evidence that has_m64 is still used? |
| 4 |
|
| 5 |
> Nope. It is just that it is part of an API that we export and, since |
| 6 |
> we have easy means to drop it properly, why not doing so ? Esp. since |
| 7 |
> dropping it "improperly" doesn't seem to bring any advantage. |
| 8 |
|
| 9 |
But we don't have such means. Moving it into an EAPI conditional means |
| 10 |
that we don't drop it but keep the code forever (and even add more |
| 11 |
complexity). |
| 12 |
|
| 13 |
> The 5 years deprecation is irrelevant here: With C libraries, you |
| 14 |
> deprecate a symbol/function for a few years then bump soname when |
| 15 |
> dropping it. The equivalent here is removing it in a new EAPI after |
| 16 |
> a deprecation period. |
| 17 |
|
| 18 |
That's not equivalent. For C libraries there are releases and the code |
| 19 |
gets actually removed. |
| 20 |
|
| 21 |
Anyway, I don't care enough to waste my time on anything more complex |
| 22 |
than the patch in my original posting. Let the eclass keep its stale |
| 23 |
code then. |
| 24 |
|
| 25 |
Ulrich |
Archives