I agree on not re-using accidentally available structures. But it got me thinking that maybe the BufSpan could be something else.
I’m somewhat confused by the fact that we agreed to do it post-merge, but all further work has been in some other direction.
My work since the merge has been on consolidating the actual use-case for the exact print annotations, which means making sure that the (still external) ghc-exactprint library is still able to support the existing use-cases. My current driver for that is porting retrie[1] to use it, which has brought up the changes that I have been doing to GHC recently.
Once that is done, I plan to update the in-tree exact printing to match, and then contemplate actioning plan B.
Alan