According to the docs, CreateProcess can be passed an executable name as first argument, or a command line as a second argument (from which the executable name will be extracted).
If you pass an executable name, the docs say PATH won't be searched.
if you pass a command line instead, the first token is extracted to be used
as the executable name and PATH is supposed to be searched.
In my case, though, my call to CreateProcess ---with a command line only and
with a modified environment--- doesn't find the sought executable. It only
succeeds if I precede the command line with cmd.exe /c (I understand why
it works this way).
For completeness, I'm not actually using the Windows API directly, but
subprocess.Popen in Python, although I think I've narrowed down the problem
to the above circumstances. With shell = True, the right environment is
picked up; with shell = False (my desired way of creating the subprocess),
the call fails to locate my executable. The executable is a standalone exe, not an intrinsic command of cmd.exe.
Can someone please tell my what I'm doing wrong here or where's my misunderstanding?
Example code:
from subprocess import Popen
import os, sys
exe = "wc.exe" # No other wc.exe on the PATH
env = os.environ.copy()
new_path = os.path.expandvars(r"%HOMEDRIVE%%HOMEPATH%\SmallApps\GnuWin32\bin;%PATH%")
env["PATH"] = os.path.expandvars(new_path).encode(sys.getfilesystemencoding())
Popen(
args=[exe, "*.*"],
env=env,
# shell=True # Works if you uncomment this line.
)
