On Mon, Apr 16, 2012 at 9:37 AM, Manuel Klimek <[email protected]> wrote:
> I like the factory methods much better! Thanks :)
>
> ERROR_OK is still in there though?
>
> +        if isinstance(unsaved_files, dict):
> +            for k, v in unsaved_files.iteritems():
> +                unsaved_normalized.append((k, v))
>
> unsaved_normalized = unsaved_files.items()
>
> While I see how this is convenient, why do we not require a list and
> let people outside call .items() if they have a dict? I kind of
> dislike all this type specific code, but that's more a gut feeling
> than being able to point my finger at problems.
>
> Also, what happens if we pass parameters of incorrect type to
> TranslationUnit_parse? All that type checking in python code seems
> somewhat strange...

I agree with what you said. Looking at the code now, I don't know what
I was thinking :)

Issues addressed with attached patch. Sorry it took so long to respond.

Since the time I created this patch, the behavior of
clang_saveTranslationUnit has changed. Previously, if you created a TU
with critical errors (like bad syntax), that API would return an error
code. Now, it seems to return success and write the file. I marked the
test that exercises the exception raising bits as skipped as a result.
I'm trying to think of the best way to test this now. Perhaps try
writing to a file without write permissions?

Attachment: tu-refactor.patch
Description: Binary data

_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to