Under Python 2.6.4 64-bit on Windows 7 64-bit, I found that when launching a script under the debugger, if backslashes were in the script pathname, they were not interpreted correctly by the interpreter.

For example, create a simple test script, "t-helloworld.py" with the canonical "hello-world" content. The script name must start with a backslash escape character such as 't' or 'n'. Then, from the command prompt:

> python -m pdb .\t-helloworld.py
IOError: (2, 'No such file or directory', '.\t-helloworld.py')
> <string>(1)<module>()
(Pdb)

However, using forward slashes works just fine.

> python -m pdb ./t-helloworld.py
> c:\debug\t-helloworld.py(1)<module>()
-> print "hello world"
(Pdb)

Note that launching the script from the python directly does not exhibit the error - it seems to be only when pdb is used.

Expected behavior: pdb should interpret the command-line parameters the same way Python does.

You can see the same thing by doing execfile(".\test.py"), which is what pdb makes under the hood, then bdb exec's it.

I'm changing the title back to the original title. I don't believe the problem is reproducible using execfile. That is, if a properly-escaped path is passed to execfile, it works fine.

So I believe the problem lies between when pdb takes control and where execfile is called - the command-line is not properly parsed.

I suspect this patch may fix the problem. I haven't yet had time to test it.

Index: Lib/pdb.py
===================================================================
--- Lib/pdb.py  (revision 77683)
+++ Lib/pdb.py  (working copy)
@@ -1200,7 +1200,7 @@
         self._wait_for_mainpyfile = 1
         self.mainpyfile = self.canonic(filename)
         self._user_requested_quit = 0
-        statement = 'execfile( "%s")' % filename
+        statement = 'execfile(%r)' % filename
         self.run(statement)

 # Simplified interface

Works for me, repr will escape the slash.

The attached test-case identifies the failure mode (and verifies that the proposed fix corrects the issue).

Can you add your test to Lib/test/test_pdb.py?

If the test fails the "hello world" script won't be removed, it won't make it to os.remove(). You'll need to make sure the file gets deleted in any case. Could use a temporary file.

Can you condense the cmd stuff into less lines? Super minor thing, but it's all short lines anyways.

Rather than testing using failIf, assertFalse is a better alternative. In 2.7 there is assertIn which would be ideal for this test, but this will go into 2.6 as well.

Here's a new patch against the trunk that addresses Brian's concerns.

For completeness, I've back-ported the patch to the release26-maint branch.

this bug exist also in python 2.5 (of course)

and the same patch as above fixes it (of course).

New changeset f7dd5178f36a by Jason R. Coombs in branch '2.7':
PDB now will properly escape backslashes in the names of modules it executes. Fixes #7750
http://hg.python.org/cpython/rev/f7dd5178f36a

This bug doesn't appear to be present in 3.1, so the 2.7 patch is all that applies at this time.