Let subprocess decode stdout based on localebschubert/no-subprocess-decode

Subprocesses can return decoded strings if we give them a
"universal_newlines=True" argument. We can therefore offload that to
them, and not explicitly decode output ourselves.

This also fixes multiple bugs where we would not be respecting the
locale used by the user, and in some cases force it to "ascii".

2 files changed, 2 insertions, 7 deletions

diff --git a/buildstream/_platform/linux.py b/buildstream/_platform/linux.py
index 702059a5d..85e810c26 100644
--- a/buildstream/_platform/linux.py
+++ b/buildstream/_platform/linux.py
@@ -143,8 +143,7 @@ class Linux(Platform):
                 '--unshare-user',
                 '--uid', '0', '--gid', '0',
                 whoami,
-            ])
-            output = output.decode('UTF-8').strip()
+            ], universal_newlines=True).strip()
         except subprocess.CalledProcessError:
             output = ''
 
diff --git a/buildstream/utils.py b/buildstream/utils.py
index a4e161ed4..2960348e9 100644
--- a/buildstream/utils.py
+++ b/buildstream/utils.py
@@ -1122,14 +1122,10 @@ def _call(*popenargs, terminate=False, **kwargs):
 
     with _signals.suspendable(suspend_proc, resume_proc), _signals.terminator(kill_proc):
         process = subprocess.Popen(  # pylint: disable=subprocess-popen-preexec-fn
-            *popenargs, preexec_fn=preexec_fn, **kwargs)
+            *popenargs, preexec_fn=preexec_fn, universal_newlines=True, **kwargs)
         output, _ = process.communicate()
         exit_code = process.poll()
 
-    # Program output is returned as bytes, we want utf8 strings
-    if output is not None:
-        output = output.decode('UTF-8')
-
     return (exit_code, output)