bpo-17852: Maintain a list of BufferedWriter objects. Flush them on exit. (#1908)

* Maintain a list of BufferedWriter objects.  Flush them on exit.

In Python 3, the buffer and the underlying file object are separate
and so the order in which objects are finalized matters.  This is
unlike Python 2 where the file and buffer were a single object and
finalization was done for both at the same time.  In Python 3, if
the file is finalized and closed before the buffer then the data in
the buffer is lost.

This change adds a doubly linked list of open file buffers.  An atexit
hook ensures they are flushed before proceeding with interpreter
shutdown.  This is addition does not remove the need to properly close
files as there are other reasons why buffered data could get lost during
finalization.

Initial patch by Armin Rigo.

* Use weakref.WeakSet instead of WeakKeyDictionary.

* Simplify buffered double-linked list types.

* In _flush_all_writers(), suppress errors from flush().

* Remove NEWS entry, use blurb.
This commit is contained in:
Neil Schemenauer 2017-09-04 20:18:38 -07:00 committed by GitHub
parent 64263dfd18
commit e38d12ed34
5 changed files with 75 additions and 1 deletions

View File

@ -1185,6 +1185,7 @@ class BufferedWriter(_BufferedIOMixin):
self.buffer_size = buffer_size
self._write_buf = bytearray()
self._write_lock = Lock()
_register_writer(self)
def writable(self):
return self.raw.writable()
@ -2574,3 +2575,26 @@ class StringIO(TextIOWrapper):
def detach(self):
# This doesn't make sense on StringIO.
self._unsupported("detach")
# ____________________________________________________________
import atexit, weakref
_all_writers = weakref.WeakSet()
def _register_writer(w):
# keep weak-ref to buffered writer
_all_writers.add(w)
def _flush_all_writers():
# Ensure all buffered writers are flushed before proceeding with
# normal shutdown. Otherwise, if the underlying file objects get
# finalized before the buffered writer wrapping it then any buffered
# data will be lost.
for w in _all_writers:
try:
w.flush()
except:
pass
atexit.register(_flush_all_writers)

View File

@ -0,0 +1,2 @@
Maintain a list of open buffered files, flush them before exiting the
interpreter. Based on a patch from Armin Rigo.

View File

@ -766,6 +766,8 @@ PyInit__io(void)
!(_PyIO_empty_bytes = PyBytes_FromStringAndSize(NULL, 0)))
goto fail;
_Py_PyAtExit(_PyIO_atexit_flush);
state->initialized = 1;
return m;

View File

@ -183,3 +183,5 @@ extern PyObject *_PyIO_empty_str;
extern PyObject *_PyIO_empty_bytes;
extern PyTypeObject _PyBytesIOBuffer_Type;
extern void _PyIO_atexit_flush(void);

View File

@ -196,7 +196,7 @@ bufferediobase_write(PyObject *self, PyObject *args)
}
typedef struct {
typedef struct _buffered {
PyObject_HEAD
PyObject *raw;
@ -240,8 +240,18 @@ typedef struct {
PyObject *dict;
PyObject *weakreflist;
/* a doubly-linked chained list of "buffered" objects that need to
be flushed when the process exits */
struct _buffered *next, *prev;
} buffered;
/* the actual list of buffered objects */
static buffered buffer_list_end = {
.next = &buffer_list_end,
.prev = &buffer_list_end
};
/*
Implementation notes:
@ -386,6 +396,15 @@ _enter_buffered_busy(buffered *self)
(self->buffer_size * (size / self->buffer_size)))
static void
remove_from_linked_list(buffered *self)
{
self->next->prev = self->prev;
self->prev->next = self->next;
self->prev = NULL;
self->next = NULL;
}
static void
buffered_dealloc(buffered *self)
{
@ -394,6 +413,8 @@ buffered_dealloc(buffered *self)
return;
_PyObject_GC_UNTRACK(self);
self->ok = 0;
if (self->next != NULL)
remove_from_linked_list(self);
if (self->weakreflist != NULL)
PyObject_ClearWeakRefs((PyObject *)self);
Py_CLEAR(self->raw);
@ -1817,10 +1838,33 @@ _io_BufferedWriter___init___impl(buffered *self, PyObject *raw,
self->fast_closed_checks = (Py_TYPE(self) == &PyBufferedWriter_Type &&
Py_TYPE(raw) == &PyFileIO_Type);
if (self->next == NULL) {
self->prev = &buffer_list_end;
self->next = buffer_list_end.next;
buffer_list_end.next->prev = self;
buffer_list_end.next = self;
}
self->ok = 1;
return 0;
}
/*
* Ensure all buffered writers are flushed before proceeding with
* normal shutdown. Otherwise, if the underlying file objects get
* finalized before the buffered writer wrapping it then any buffered
* data will be lost.
*/
void _PyIO_atexit_flush(void)
{
while (buffer_list_end.next != &buffer_list_end) {
buffered *buf = buffer_list_end.next;
remove_from_linked_list(buf);
buffered_flush(buf, NULL);
PyErr_Clear();
}
}
static Py_ssize_t
_bufferedwriter_raw_write(buffered *self, char *start, Py_ssize_t len)
{