Prev: Sun Grid Engine / NFS and Python shell execution question
Next: James Gosling the Creator of EMACS and JAVA - leaves ORACLE - But then reports started coming in of odd failures. Systems would crash strangely. We'd get crashes in applications. All applications. Crashes in the kernel.
From: Burton Samograd on 22 Jul 2010 12:12
Hello,

I have written an importing extension along the lines of PEP 302
(http://www.python.org/dev/peps/pep-0302/) and have run into a bit of a
problem with stack backtraces after exceptions.

When I run code with the using my importing extension, backtraces come
up looking like this:

Traceback (most recent call last):
File "<string>", line 134, in <module>
File "<string>", line 9, in <module>
File "<string>", line 7, in a
File "<string>", line 4, in b
TypeError

What I'm having a problem with is that the file names are no longer
being printed during the backtrace. The following code is from my
importer function load_module:

try:
mod = sys.modules[module_name]
already_in_sys_modules = True
except KeyError:
mod = sys.modules.setdefault(module_name,
imp.new_module(module_name))
already_in_sys_modules = False
mod.__file__ = "%s" % module_path
mod.__loader__ = self
if is_package:
mod.__path__ = [ module_path ]
mod.__dict__['__file__'] = module_path
try:
exec code in mod.__dict__
except:
import traceback
print traceback.format_exc(5)
print "ERROR: could not load module: %s" % module_path
if not already_in_sys_modules:
del sys.modules[module_name]
mod = None

As shown in PEP 302, I am setting the mod.__file__ variable and as my
own idea I set __file__ in mod.__dict__. Niether of these steps seem to
set the file properly in the backtrace.

So the question here is, where does the backtrace printing module get
it's __file__ parameters from, or what am I missing to set it properly?

Thanks in advance.

--
Burton Samograd

From: Peter Otten on 22 Jul 2010 13:00
Burton Samograd wrote:

> Hello,
>
> I have written an importing extension along the lines of PEP 302
> (http://www.python.org/dev/peps/pep-0302/) and have run into a bit of a
> problem with stack backtraces after exceptions.
>
> When I run code with the using my importing extension, backtraces come
> up looking like this:
>
> Traceback (most recent call last):
> File "<string>", line 134, in <module>
> File "<string>", line 9, in <module>
> File "<string>", line 7, in a
> File "<string>", line 4, in b
> TypeError
>
> What I'm having a problem with is that the file names are no longer
> being printed during the backtrace. The following code is from my
> importer function load_module:
>
> try:
> mod = sys.modules[module_name]
> already_in_sys_modules = True
> except KeyError:
> mod = sys.modules.setdefault(module_name,
> imp.new_module(module_name))
> already_in_sys_modules = False
> mod.__file__ = "%s" % module_path
> mod.__loader__ = self
> if is_package:
> mod.__path__ = [ module_path ]
> mod.__dict__['__file__'] = module_path
> try:
> exec code in mod.__dict__
> except:
> import traceback
> print traceback.format_exc(5)
> print "ERROR: could not load module: %s" % module_path
> if not already_in_sys_modules:
> del sys.modules[module_name]
> mod = None
>
> As shown in PEP 302, I am setting the mod.__file__ variable and as my
> own idea I set __file__ in mod.__dict__. Niether of these steps seem to
> set the file properly in the backtrace.
>
> So the question here is, where does the backtrace printing module get
> it's __file__ parameters from, or what am I missing to set it properly?
>
> Thanks in advance.
>
> --
> Burton Samograd

exec code

implicitly compiles code with "<string>" as the filename.
The filename is stored in f.func_code.co_filename and it is read-only.

>> exec "def f(x): return f(x-1) if x else 1/0"
>>> f(3)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "<string>", line 1, in f
File "<string>", line 1, in f
File "<string>", line 1, in f
File "<string>", line 1, in f
ZeroDivisionError: integer division or modulo by zero

If you make the compilation step explicit you can pass a filename:

>>> exec compile("def f(x): return f(x-1) if x else 1/0", "yadda.py",
"exec")
>>> f(3)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "yadda.py", line 1, in f
File "yadda.py", line 1, in f
File "yadda.py", line 1, in f
File "yadda.py", line 1, in f
ZeroDivisionError: integer division or modulo by zero

Of course the offending line isn't quoted because there is no file
"yadda.py" on my harddisk. But I can change that:

>>> with open("yadda.py", "w") as out: print >> out, "I was fooled!"
....
>>> f(3)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "yadda.py", line 1, in f
I was fooled!
File "yadda.py", line 1, in f
I was fooled!
File "yadda.py", line 1, in f
I was fooled!
File "yadda.py", line 1, in f
I was fooled!
ZeroDivisionError: integer division or modulo by zero

Peter


From: Burton Samograd on 22 Jul 2010 13:48
Peter Otten <__peter__(a)web.de> writes:

<snip>

> If you make the compilation step explicit you can pass a filename:
>
>>>> exec compile("def f(x): return f(x-1) if x else 1/0", "yadda.py",
> "exec")

The works great. Problem solved. Thanks.

--
Burton Samograd

 | 
Pages: 1
Prev: Sun Grid Engine / NFS and Python shell execution question
Next: James Gosling the Creator of EMACS and JAVA - leaves ORACLE - But then reports started coming in of odd failures. Systems would crash strangely. We'd get crashes in applications. All applications. Crashes in the kernel.