sigh, soname troubles plus this nssov thing - this clearly is one of the more ugly merged.
/me hugs Andreas
Re-reviewing the MP as it is right now ...
- Ack on not removing nssov (for now)
- Ack on keeping the bad symbols until we can soname bump
(both as discussed)
+1 on the new set of kept/dropped changes.
One thing if you want to experiment a bit more with it since we can't get rid of the extra features/symbols we have that came to my mind last weekend was deprecating them.
Would it be a reasonable delta to throw in some "deprecated" attributes via [1]?
That way - once we some day remove it - everyone linking against them would have had quite some time being told that they are deprecated.
One could think of a similar strategy for nssov to now yell/warn/message about that it will be dropped later in all places you can - to reduce the impact when you do it some time down the road. IIRC you already have the code to detect nssov and while I agree messing with the config is error-prone, warning that it should not be used would be fine IMHO.
sigh, soname troubles plus this nssov thing - this clearly is one of the more ugly merged.
/me hugs Andreas
Re-reviewing the MP as it is right now ...
- Ack on not removing nssov (for now)
- Ack on keeping the bad symbols until we can soname bump
(both as discussed)
+1 on the new set of kept/dropped changes.
One thing if you want to experiment a bit more with it since we can't get rid of the extra features/symbols we have that came to my mind last weekend was deprecating them.
Would it be a reasonable delta to throw in some "deprecated" attributes via [1]?
That way - once we some day remove it - everyone linking against them would have had quite some time being told that they are deprecated.
One could think of a similar strategy for nssov to now yell/warn/message about that it will be dropped later in all places you can - to reduce the impact when you do it some time down the road. IIRC you already have the code to detect nssov and while I agree messing with the config is error-prone, warning that it should not be used would be fine IMHO.
[1]: https:/ /gcc.gnu. org/onlinedocs/ gcc-4.7. 1/gcc/Type- Attributes. html#Type- Attributes