pushpdb: Don't loop indefinitely if symbol store is broken

If the store is not accessible the script should not get stuck
in an inifinite loop trying to create the lock file. This commit
adds another check which aborts the pushpdb attempt in such a
case.

Also fixing 'timeout' invocation to make sure it runs the Windows
tool and not the GNU one when running under Cygwin.

Signed-off-by: Ladi Prosek <lprosek@redhat.com>

Tools/pushpdb.cmd

@@ -203,6 +203,9 @@ if exist "%~n1.inf" (
     for /f "tokens=2 delims=," %%t in ('findstr /b "DriverVer" "%~n1.inf"') do (
         for /f "tokens=1 delims= " %%v in ('echo %%~t') do (
             call :lock_file
+            if  ERRORLEVEL 1 (
+                exit /b 1
+            )
             echo Pushing %~1 version %%~v to the symbol store.
             if "%DSTORE_BUILD_TAG%"=="" (
                 call "%path_to_systore_exe%" add /s "%store_address%" /f "%~1" /t "%~n1" /v "%%~v" /c "eng" > NUL
@@ -225,6 +228,8 @@ exit /b 0
 :: Creates a lock file with its unique stamp, tries to copy it to the stores
 :: location, then compares the lock at the store with its own, if equal then
 :: we locked and we continue, if not then we wait for 2 seconds then try again.
+:: If we're unable to create a uniquely named file in the store directory we
+:: abort the loop and return an error.
 :lock_file
 :loop
 (
@@ -241,8 +246,14 @@ fc %GUID%.lock "%store_address%\lock" > NUL 2>&1
 if not ERRORLEVEL 1 (
     exit /b 0
 ) else (
+    copy /y %GUID%.lock "%store_address%\%GUID%.lock" > NUL
+    if ERRORLEVEL 1 (
+        echo store appears to be broken, aborting
+        exit /b 1
+    )
+    del "%store_address%\%GUID%.lock" > NUL
     echo someone else is pushing, waiting for 2 secs
-    timeout /t 2 /nobreak > NUL
+    %windir%\system32\timeout /t 2 /nobreak > NUL
     goto :loop
 )