David Binder

>so that we give the impression that they may not be stable (which is true). My impression was that for the modules exposed in this PR, there is the least...

My original idea was that for every instance of `Pretty T` that is currently defined in the library, the corresponding instance `PrettyAnn T ann` should also be provided. This means,...

Result of git bisect: ``` david@anaxes ~/GitRepos/hakyll (master)$ git bisect start david@anaxes ~/GitRepos/hakyll (master)$ git bisect bad david@anaxes ~/GitRepos/hakyll (master)$ git bisect good v4.15.1.1 Bisecting: 30 revisions left to test...

The bug is independent on whether the threaded runtime is used or not. If I remove the `ghc-options: -threaded -rtsopts -with-rtsopts=-N` line in `message-index.cabal`, then the problem persists.

Thanks, I have https://github.com/haskellfoundation/error-message-index/pull/461 up. So the error is on our side, and we just could not observe the erroneous result before since we relied on a specific execution order...

GHC now uses a python script to generate the header files: https://gitlab.haskell.org/ghc/ghc/-/blob/master/rts/gen_event_types.py Maybe this python script could be generalized to also replace ghc-event's use of its own EventLogFormat.h file.

> I agree that the new output is correct. Why do you think the new output is better? That is, I can't come up with a reason that applying the...

There is currently no test in CI which tests whether the dependencies of eventlog2html can be satisfied by any of the stackage snapshots. So there is no guarantee that eventlog2html...

There is probably a confusion about what eventlog2html does, in comparison to Threadscope. The eventlog is a general purpose mechanism of the runtime system of Haskell programs to emit data...

https://gitlab.haskell.org/ghc/ghc/-/issues/23626 is related, though not the same. We now have a method to at least list all the error messages that a given GHC version can emit.