Re: svn commit: r956106 - /subversion/trunk/build/run_tests.py

[Greg Stein <gs...@gmail.com>]

On Fri, Jun 18, 2010 at 15:34,  <hw...@apache.org> wrote:
> Author: hwright
> Date: Fri Jun 18 19:34:39 2010
> New Revision: 956106
>
> URL: http://svn.apache.org/viewvc?rev=956106&view=rev
> Log:
> * build/run_tests.py
>  (_run_py_test): Don't write to the placeholder stdout in the progress func,
>    write to the real one, or the real bogus one.  (This fixes a bug which
>    could occur if old_stdout isn't valid.)
>
> Modified:
>    subversion/trunk/build/run_tests.py
>
> Modified: subversion/trunk/build/run_tests.py
> URL: http://svn.apache.org/viewvc/subversion/trunk/build/run_tests.py?rev=956106&r1=956105&r2=956106&view=diff
> ==============================================================================
> --- subversion/trunk/build/run_tests.py (original)
> +++ subversion/trunk/build/run_tests.py Fri Jun 18 19:34:39 2010
> @@ -311,8 +311,6 @@ class TestHarness:
>       old_stderr = os.dup(2)
>       os.dup2(self.log.fileno(), 1)
>       os.dup2(self.log.fileno(), 2)
> -    else:
> -      old_stdout = sys.stdout
>
>     # This has to be class-scoped for use in the progress_func()
>     self.dots_written = 0
> @@ -320,8 +318,13 @@ class TestHarness:
>       dots = (completed * dot_count) / total
>
>       dots_to_write = dots - self.dots_written
> -      os.write(old_stdout, '.' * dots_to_write)
> -      os.fsync(old_stdout)
> +      if self.log:
> +        os.write(old_stdout, '.' * dots_to_write)
> +        os.fsync(old_stdout)
> +      else:
> +        sys.stdout.write(old_stdout, '.' * dots_to_write)
> +        sys.stdout.flush()

Doesn't the write() method take a single argument? The contents to
write? Seems like a copy/paste problem.

And from the os.write() call, it seems that old_stdout is a file
descriptor, correct? Why is Python code dealing with file descriptors
in the first place?

Cheers,
-g