Explicitly passing -package-env causes "Loaded package environment from .ghc.environment" message to be printed twice

changed weight to 5

Actually, the window of time in which this regression was introduced can be narrowed down even further, since this does not occur in GHC 8.6.2:

$ /opt/ghc/8.6.2/bin/ghc -package-env .ghc.environment.x86_64-linux-8.6.2 -e "putStrLn \"Hello\""
Loaded package environment from .ghc.environment.x86_64-linux-8.6.2
Hello

I'm reasonably confident that commit e67bebbf (Load plugins in interactive session) is the culprit:

-ghciUI :: [(FilePath, Maybe Phase)] -> Maybe [String] -> Ghc ()
+ghciUI :: HscEnv -> DynFlags -> [(FilePath, Maybe Phase)] -> Maybe [String]
+       -> Ghc ()
 #if !defined(GHCI)
-ghciUI _ _ = throwGhcException (CmdLineError "not built for interactive use")
+ghciUI _ _ _ _ =
+  throwGhcException (CmdLineError "not built for interactive use")
 #else
-ghciUI     = interactiveUI defaultGhciSettings
+ghciUI hsc_env dflags0 srcs maybe_expr = do
+  dflags1 <- liftIO (initializePlugins hsc_env dflags0)
+  _ <- GHC.setSessionDynFlags dflags1
+  interactiveUI defaultGhciSettings srcs maybe_expr
 #endif

This inserts a call to setSessionDynFlags, which transitively invokes initPackages, which is the function that logs the "Loaded package environment from ..." message. But if you look at the source code for interactiveUI, you'll see that it invokes setProgramDynFlags, which //itself// transitively invokes initPackages!

It seems like we might want to decouple the action of logging package environments from initPackages. Thoughts, hvr?

added package system label

added Pnormal label

This breaks hlint and hlint editor integrations. Do you think such a refactor is needed @RyanGlScott or can we just pass a flag or function that controls the output? I would hope to fix this in the next point release if there is agreement on the strategy.

If your goal is to just suppress the Loading package environment message, then I think that's tracked separately in #16879 (closed). Of course, that doesn't fix this issue, since (AFAICT) the message will still be printed twice if -v0 isn't used. I'm not terribly familiar with the part of the codebase that generates this message, however, so I'm afraid I don't have a concrete recommendation to make vis-à-vis how it should be changed. Thoughts, @hvr?

mentioned in issue #18125 (closed)

This seems to be fixed in 8.10. @RyanGlScott is it ok to close?

I can still reproduce the bug in this issue, although you now need to pass -v1 or greater due to !1308 (closed):

$ /opt/ghc/8.10.1/bin/ghc -package-env .ghc.environment.x86_64-linux-8.10.1 -e "putStrLn \"Hello\"" -v1
Loaded package environment from .ghc.environment.x86_64-linux-8.10.1
Loaded package environment from .ghc.environment.x86_64-linux-8.10.1
Hello
Leaving GHCi.

mentioned in merge request !3203 (closed)

assigned to @ulysses4ever

mentioned in commit c76d6346

mentioned in commit b373a2d5

mentioned in commit 9878e607

mentioned in commit 8f9179c3

mentioned in commit d15847db

mentioned in commit 927b3a5e

mentioned in commit d880d6b2

Closed by d880d6b2.

closed

mentioned in issue #18422 (closed)

Trac field	Value
Version	8.6.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	GHCi
Test case
Differential revisions
BlockedBy
Related
Blocking
CC	hvr
Operating system
Architecture

Explicitly passing -package-env causes "Loaded package environment from .ghc.environment" message to be printed twice

Child items 0

Activity