We have forkIOWithUnmask[1] and forkOnWithUnmask[2], but we don't have forkOSWithUnmask. This proposal would add it:

-- | Like 'forkIOWithUnmask', but the child thread is a bound thread,

-- as with 'forkOS'.

forkOSWithUnmask :: ((forall a . IO a -> IO a) -> IO ()) -> IO ThreadId

forkOSWithUnmask io = forkOS (io unsafeUnmask)

For GHC < 7.8, forkOSWithUnmask can be achieved using forkOS and block[3], but block is removed in GHC 7.8. unsafeUnmask can still be imported from GHC.IO, however.

Impact: virtually no compatibility break (just adds a new definition), but we will probably want to update the async package[4] at some point, adding asyncBoundWithUnmask and withAsyncBoundWithUnmask. Such an addition would not require base >= 4.7 in async, as we could use the `block` workaround described above.

The biggest objection I can imagine would be that we now have six basic ways to fork a thread: forkIO, forkOn, forkOS, and their fork*WithUnmask variants. Libraries like async end up having to replicate all these. Perhaps in the future, we should consolidate forkIO, forkOn, and forkOS:

data ThreadMode

= ForkIO

| ForkOn Int

| ForkBound

forkMode :: ThreadMode -> IO () -> IO ()

forkModeWithUnmask :: ThreadMode

-> ((forall a. IO a -> IO a) -> IO ())

-> IO ThreadId

But for now, adding forkOSWithUnmask mirrors forkIOWithUnmask and forkOnWithUnmask, which are already in place.

Ticket: http://hackage.haskell.org/trac/ghc/ticket/8010 . The title "Add forkOSUnmasked" is a mistake (should be "Add forkOSWithUnmask"), but the patch should be right.

[1]: http://hackage.haskell.org/packages/archive/base/latest/doc/html/Control-Concurrent.html#v:forkIOWithUnmask

[2]: http://hackage.haskell.org/packages/archive/base/latest/doc/html/Control-Concurrent.html#v:forkOnWithUnmask

[3]: http://hackage.haskell.org/packages/archive/base/4.6.0.1/doc/html/Control-Exception.html#v:block

[4]: http://hackage.haskell.org/packages/archive/async/latest/doc/html/Control-Concurrent-Async.html