Open mpilgrem opened 4 months ago
CI where it first happened: https://github.com/haskell/ghcup-hs/actions/runs/9996272096/job/27630268254?pr=1108
The second run did not suffer from this issue: https://github.com/haskell/ghcup-hs/actions/runs/9996272096/job/27695488770?pr=1108
This suggests it's a spurious failure.
While building
happy-1.20.1.1
with GHC 9.4.8 on Windows in CI, @hausfell experienced:This was during the
Setup.hs copy
step, which sees Cabal (the library), by default, runstrip
after having moved the executable file to a new location. For example, duringstack --verbose build --cabal-verbose --ghc-options=-v
:I think it must be
strip.exe
seeking to overwritehappy.exe
when it is not yet free to do so, for some reason.Cabal's
Distribution.Simple.GHC.installExe
has (extract):So, it looks to me that things are properly sequenced. Could it be a problem within the MSYS2-supplied
strip.exe
itself?