Mikael Simberg
Mikael Simberg
Attempting to reenable SBO without any UB, but this needs more testing. We disabled SBO in the past because it was causing issues at least on LUMI, and I haven't...
We currently allow dropping senders from `async_rw_mutex`. This was not a conscious design choice, and can even be confusing to users. While the current design works, as in accesses stay...
https://github.com/pika-org/pika/pull/1379 moves the storage of the value in `async_rw_mutex` into a heap-allocated `shared_ptr`. Since we're anyway keeping track of accesses to the value and know when it's no longer needed...
Once https://github.com/breathe-doc/breathe/pull/1010 is released in a tagged version of breathe, we can lift the version restriction added in https://github.com/pika-org/pika/pull/1413.
As in https://github.com/NVIDIA/stdexec/pull/1125.
There's potential for misunderstandings with the current `drop_value` sender adaptor. It's currently implemented with the semantics of `then([](auto&&...) {})`, i.e. all value categories are accepted, but always by reference, so...
It seems like there are still cases where the `BUILD_TYPE` variable is set to `Release` even though the image used for testing has `debug` in the name, indicating that it...
https://github.com/pika-org/pika/pull/1239 disables valgrind testing on the CUDA CI configuration, since valgrind on aarch64 is either noisy for no good reason, or pika is buggy there. We have to investigate which...