Per

What's happening here?

To resolve this, I find myself using:

#if MIN_VERSION_base(4,16,0)

But why would MIN_VERSION_GLASGOW_HASKELL not work?

CodePudding user response：

GHC versions consist of a major version, minor version, first patchlevel, and optional second patchlevel. So GHC 9.2.1 has major 9, minor 2, first patchlevel 1, and no second patch level.

There's a bug in GHC, possibly in existence since the macro MIN_VERSION_GLASGOW_HASKELL was introduced, where that macro depends on a macro variable __GLASGOW_HASKELL_PATCHLEVEL2__ which is not defined when, as is usual for release versions, the GHC version has only one patch level (like 9.2.1).

This usually doesn't matter, except if the -Wcpp-undef flag is in force AND the test is performed on a version that matches on the major, minor, and first patch level. Then, the check on the second patch level will generate a compiler warning message. (Despite the fact that it looks like an "error" rather than a warning, the compilation appears to complete.)

You can work around this by:

• ignoring it -- despite appearances, it looks like it's only a warning, not an error, and compilation should complete

• shutting off -Wcpp-undef by either removing the flag or, if feasible, adding a -Wno-cpp-undef flag;

• adding some code to define the missing patchlevel before calling the macro:

   #ifndef __GLASGOW_HASKELL_PATCHLEVEL2__
   #define __GLASGOW_HASKELL_PATCHLEVEL2__ 0
   #endif