GitLab

Create a standalone Haskell program "/tmp/delme.hs". Then do

$ cd /tmp
{- 0 -}$ ghc --make -o delme.exe /tmp/delme.hs
[1 of 1] Compiling Main ( /tmp/delme.hs, /tmp/delme.o )
Linking delme.exe ...
$ rm delme.o
{- 1 -}$ ghc --make -o delme.exe /tmp/delme
[1 of 1] Compiling Main ( /tmp/delme.hs, /tmp/delme.o )
Linking delme.exe ...
$ rm delme.o
$ mv delme.hs delme
{- 2 -}$ ghc --make -o delme.exe /tmp/delme
target `/tmp/delme' is not a module name or a source file

We see that:[[BR]]

• in "1" GHC compiles "/tmp/delme.hs" though I say "/tmp/delme";[[BR]]
• in "2" GHC can not see file "/tmp/delme".

The problem is that a target (the last argument) may be a module name or a source file. GHC uses heuristics to determine this. Though heuristics may seem neat and clever, they draw GHC more intricate and unpredictable (see also KISS).

I encountered this intricacy when ran "ghc" on a shell script. Shell scripts in Unix commonly do not have extensions, because interpreter name is included in a shell script. And "runghc" runs them fine. "ghc" in theory also does not require a file to have an extension because it clearly contains Haskell source code — what else?

I propose to introduce a GHC flag which says explicitly which type a target is of.

Trac metadata

Trac field	Value
Version	6.12.2
Type	FeatureRequest
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture