Redetect machines after every change to languages

Currently, detect_machine_info() is called:

- from Environment.__init__(), before we have any compiler available

- from Interpreter.__init__() with the list of languages provided to
project() (determined via the initial parse_project())

This is not sufficent in the case where no languages are specified to
project() and are later added with add_languages().

We cannot correctly detect that the host machine should be treated as
x86 (on x86_64 hardware) until we have a compiler we are told to use.

Redetect machines after project(), and after every add_languages().

1 files changed, 3 insertions, 0 deletions

diff --git a/mesonbuild/interpreter.py b/mesonbuild/interpreter.py
index c29ed89..b00324d 100644
--- a/mesonbuild/interpreter.py
+++ b/mesonbuild/interpreter.py
@@ -2216,7 +2216,9 @@ class Interpreter(InterpreterBase):
         self.build_def_files = [os.path.join(self.subdir, environment.build_filename)]
         if not mock:
             self.parse_project()
+        self._redetect_machines()
 
+    def _redetect_machines(self):
         # Re-initialize machine descriptions. We can do a better job now because we
         # have the compilers needed to gain more knowledge, so wipe out old
         # inference and start over.
@@ -3014,6 +3016,7 @@ external dependencies (including libraries) must go to "dependencies".''')
         success &= self.add_languages_for(args, required, MachineChoice.HOST)
         if not self.coredata.is_cross_build():
             self.coredata.copy_build_options_from_regular_ones()
+        self._redetect_machines()
         return success
 
     def add_languages_for(self, args, required, for_machine: MachineChoice):