Change ExecLinkWrapper to not buffer all tool output

build/toolchain/win/tool_wrapper.py

Index: build/toolchain/win/tool_wrapper.py
diff --git a/build/toolchain/win/tool_wrapper.py b/build/toolchain/win/tool_wrapper.py
index 281298c65c2aabe4100ad320917c2cc7aaeecbc3..4adfe92ccd54bd8b6f94b8adb4e2add10afe55b8 100644
--- a/build/toolchain/win/tool_wrapper.py
+++ b/build/toolchain/win/tool_wrapper.py
@@ -133,12 +133,17 @@ class WinTool(object):
     # non-Windows don't do that there.
     link = subprocess.Popen(args, shell=sys.platform == 'win32', env=env,
                             stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
-    out, _ = link.communicate()
-    for line in out.splitlines():
+    # Read output one line at a time as it shows up to avoid OOM failures when
+    # GBs of output is produced.
+    while True:
+      line = link.stdout.readline()
+      if len(line) == 0: # Read of zero bytes means EOF

[scottmg, 2016/12/07 22:20:56]

I'm not sure about the loop termination...

for line in link.stdout:
  if (not ...):
    print line,
link.wait()

should be the same though and a bit nicer looking?

[brucedawson, 2016/12/07 23:00:24]

Agreed. I started with a solution that read stderr and stdout and didn't remove
the complexity when I changed it to stdout only.

+        break
       if (not line.startswith('   Creating library ') and
           not line.startswith('Generating code') and
           not line.startswith('Finished generating code')):
-        print line
+        print line,
+    link.wait()
     return link.returncode
 
   def ExecLinkWithManifests(self, arch, embed_manifest, out, ldcmd, resname,