Improve boxity in deferAfterPreciseExceptionwip/T18993a

As #20746 showed, the demand analyser behaved badly in a key I/O
library (`GHC.IO.Handle.Text`), by unnessarily boxing and reboxing.

This patch adjusts the subtle function deferAfterPreciseException;
it's quite easy, just a bit subtle.

See the new Note [deferAfterPreciseException]

And this MR deals only with Problem 2 in #20746.
Problem 1 is still open.

1 files changed, 14 insertions, 0 deletions

diff --git a/testsuite/tests/stranal/should_compile/T20746b.hs b/testsuite/tests/stranal/should_compile/T20746b.hs
new file mode 100644
index 0000000000..6804fb4449
--- /dev/null
+++ b/testsuite/tests/stranal/should_compile/T20746b.hs
@@ -0,0 +1,14 @@
+module T20746b where
+
+mightThrow :: Int -> IO Int
+{-# NOINLINE mightThrow #-}
+mightThrow n = return n
+
+-- Should not unbox trp
+-- Recursive because if it's too small
+-- we don't do worker/wrapper at all
+f :: Bool -> (Int, Int, Int) -> IO (Int, Int, Int)
+f False trp = f True trp
+f True trp@(a,b,c) = do
+  _ <- mightThrow a -- this potentially throwing IO action should not force unboxing of trp
+  return trp