Change memory handling in attrs_from_List() #106
Conversation
08b0ae9 to
c392093
Compare
Codecov Report
@@ Coverage Diff @@
## master #106 +/- ##
==========================================
+ Coverage 68% 68.08% +0.08%
==========================================
Files 48 48
Lines 4675 4684 +9
Branches 784 787 +3
==========================================
+ Hits 3179 3189 +10
+ Misses 1145 1143 -2
- Partials 351 352 +1
Continue to review full report at Codecov.
|
| @@ -289,16 +294,27 @@ attrs_from_List( PyObject *attrlist, char***attrsp, PyObject** seq) { | |||
| LDAPerror_TypeError("expected bytes in list", item); | |||
There was a problem hiding this comment.
As I was reading this, I also changed the comment “Encoded by Python to UTF-8” to “Encoded in Python to UTF-8”, since it's python-ldap's Python code doing the encoding.
Modules/LDAPObject.c
Outdated
| const char *str; | ||
| #else | ||
| char *str; | ||
| #endif |
There was a problem hiding this comment.
It's should be fine to use const char everywhere.
210a6ce to
06b8ac1
Compare
|
Rebased with the two changes above folded in. |
attrs_from_List() no longer uses internal buffers of bytes/str PyObject* to maintain memory. Instead it creates its own copies, which are then passed to ldap_search_ext() and then cleaned up. The modification is required for Python 3.7. PyUnicode_AsUTF8() returns const char* but ldap_search_ext() does not use const. Closes: python-ldap#105 Signed-off-by: Christian Heimes <cheimes@redhat.com>
06b8ac1 to
3a9c490
Compare
| if (*seq == NULL) | ||
| PyObject *item = NULL; | ||
| Py_ssize_t i, len, strlen; | ||
| #if PY_MAJOR_VERSION >= 3 |
There was a problem hiding this comment.
That's wrong, it must be #if PY_VERSION_HEX >= 0x03070000. Python < 3.7 has PyAPI_FUNC(char *) PyUnicode_AsUTF8AndSize(). Only 3.7+ has PyAPI_FUNC(const char *) PyUnicode_AsUTF8AndSize()
There was a problem hiding this comment.
I thought implicitly converting char * to const char * is perfectly fine – it just limits what can be done with the string later.
Why is const char *str = PyUnicode_AsUTF8AndSize(...) a problem on 3.6-?
There was a problem hiding this comment.
I discussed the issue with @vstinner. He assured me that const is always fine here. Let's go for your version check.
| #endif | ||
| /* Make a copy. PyBytes_AsString* / PyUnicode_AsUTF8* return | ||
| * internal values that must be treated like const char. Python | ||
| * 3.7 actually returns a const char. |
|
Your modifications LGTM |
attrs_from_List() no longer uses internal buffers of bytes/str PyObject*
to maintain memory. Instead it creates its own copies, which are then
passed to ldap_search_ext() and then cleaned up.
The modification is required for Python 3.7. PyUnicode_AsUTF8() returns
const char* but ldap_search_ext() does not use const.
Closes: #105
Signed-off-by: Christian Heimes cheimes@redhat.com