Closed 23Skidoo closed 6 years ago
So basically base-4.10
deprecated GCStats
in favour of RTSStats
, which was introduced in base-4.9
.
Yep, this can be a bit of a bear to migrate to. There are examples of this migration in accelerate
and criterion
.
Is anyone taking this issue? By itself changing the RTS metrics seems OK-ish. The hard bit will be to arrange the code to work with old GHCs as well (there already are a few #ifdef
around this code).
How much backward support do we want for ekg-core
(I'd ask for ekg
in general)?
I'm not aware of anyone working on this right now, so you're welcome to.
How much backward support do we want for
ekg-core
(I'd ask forekg
in general)?
Supporting the GHC versions released during the last three years is a good rule of thumb. This means that we can drop 7.6 and 7.8.
The
GHC.Stats
module seems to have been refactored in GHC 8.2.1, so we now get a bunch of warnings of the following type when compilingSystem.Metrics
:See this gist for the full log. Looks like we'll have to write a compatibility shim to support future GHC versions.