Issue14505
This issue tracker has been migrated to GitHub,
and is currently read-only.
For more information,
see the GitHub FAQs in the Python's Developer Guide.
Created on 2012-04-05 09:01 by doko, last changed 2022-04-11 14:57 by admin. This issue is now closed.
Messages (8) | |||
---|---|---|---|
msg157555 - (view) | Author: Matthias Klose (doko) * | Date: 2012-04-05 09:01 | |
[forwarded from http://bugs.debian.org/664529] seen with 2.7.3 rc2 File descriptors opened by PyFile_FromString don't get closed when the reference count is decreased. Here's my test program, pythony.c: #include <Python.h> int main() { int i = 0; PyObject *obj; Py_Initialize(); while (i++ < 5) { obj = PyFile_FromString("hello.py", "r"); assert(obj); Py_DECREF(obj); } Py_Finalize(); } hello.py is 'print("hello world")'. I'm compiling it with both Python 2.6 and 2.7. $ gcc pythony.c -lpython2.6 -L/usr/lib/python2.6/config -I/usr/include/python2.6/ -o pythony-2.6 $ gcc pythony.c -lpython2.7 -L/usr/lib/python2.7/config -I/usr/include/python2.7/ -o pythony-2.7 $ strace ./pythony-2.6 2>&1 | tail -n 20 ioctl(0, SNDCTL_TMR_TIMEBASE or TCGETS, {B38400 opost isig icanon echo ...}) = 0 ioctl(1, SNDCTL_TMR_TIMEBASE or TCGETS, 0x7fffb1d097b0) = -1 EINVAL (Invalid argument) ioctl(2, SNDCTL_TMR_TIMEBASE or TCGETS, 0x7fffb1d097b0) = -1 EINVAL (Invalid argument) open("hello.py", O_RDONLY) = 3 fstat(3, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 close(3) = 0 open("hello.py", O_RDONLY) = 3 fstat(3, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 close(3) = 0 open("hello.py", O_RDONLY) = 3 fstat(3, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 close(3) = 0 open("hello.py", O_RDONLY) = 3 fstat(3, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 close(3) = 0 open("hello.py", O_RDONLY) = 3 fstat(3, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 close(3) = 0 rt_sigaction(SIGINT, {SIG_DFL, [], SA_RESTORER, 0x7f1e1a0224f0}, {0x7f1e1a49a160, [], SA_RESTORER, 0x7f1e1a0224f0}, 8) = 0 exit_group(0) = ? $ strace ./pythony-2.7 2>&1 | tail -n 20 fstat(4, {st_mode=S_IFREG|0644, st_size=1950, ...}) = 0 read(4, "", 4096) = 0 close(4) = 0 munmap(0x7fa41f10f000, 4096) = 0 close(3) = 0 ioctl(0, SNDCTL_TMR_TIMEBASE or TCGETS, {B38400 opost isig icanon echo ...}) = 0 ioctl(1, SNDCTL_TMR_TIMEBASE or TCGETS, 0x7ffff7bd33f0) = -1 EINVAL (Invalid argument) ioctl(2, SNDCTL_TMR_TIMEBASE or TCGETS, 0x7ffff7bd33f0) = -1 EINVAL (Invalid argument) open("hello.py", O_RDONLY) = 3 fstat(3, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 open("hello.py", O_RDONLY) = 4 fstat(4, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 open("hello.py", O_RDONLY) = 5 fstat(5, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 open("hello.py", O_RDONLY) = 6 fstat(6, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 open("hello.py", O_RDONLY) = 7 fstat(7, {st_mode=S_IFREG|0644, st_size=21, ...}) = 0 rt_sigaction(SIGINT, {SIG_DFL, [], SA_RESTORER, 0x7fa4206e24f0}, {0x7fa420b8dd50, [], SA_RESTORER, 0x7fa4206e24f0}, 8) = 0 exit_group(0) = ? The Python 2.7 version never calls close, not even at Py_Finalize(). On #d-d, jwilk suspected that this change might be the cause: http://hg.python.org/cpython/rev/0f5b64630fda/#l4.46 |
|||
msg157559 - (view) | Author: STINNER Victor (vstinner) * | Date: 2012-04-05 10:30 | |
> File descriptors opened by PyFile_FromString don't get closed > when the reference count is decreased. Correct, PyFile_FromString() doesn't close the file descriptor, even if you call its close() method. You have to call manually fclose(file->f_fp). Or you can use PyFile_FromFile: fp = fopen(name, mode); if (fp == NULL) ... obj = PyFile_FromFile(fp, name, mode, fclose); if (obj == NULL) { /* no need to call fclose(fp) here, it's done by PyFile_FromFile() */ ... } ... Py_DECREF(obj); Would you like to write a patch for the documentation? |
|||
msg157560 - (view) | Author: Antoine Pitrou (pitrou) * | Date: 2012-04-05 10:41 | |
Victor, that sounds like a strange behaviour to me. PyFile_FromString is a public API and maybe it shouldn't have changed between 2.6 and 2.7. |
|||
msg157561 - (view) | Author: STINNER Victor (vstinner) * | Date: 2012-04-05 11:20 | |
> Victor, that sounds like a strange behaviour to me. We may change PyFile_FromString() to call fclose() when the file is closed explicitly (call its close() method), but it may break backward compatibility. I prefer to document the behaviour instead. > PyFile_FromString is a public API and maybe it shouldn't > have changed between 2.6 and 2.7. PyFile_FromString() didn't change in Python 2.7. I changed PyFile_FromFile() in Python 2.7 to fix the issue #7732. changeset: 72456:0f5b64630fda branch: 2.7 parent: 72450:c02e790c4535 user: Victor Stinner <victor.stinner@haypocalc.com> date: Fri Sep 23 19:37:03 2011 +0200 files: Doc/c-api/file.rst Lib/test/test_import.py Misc/NEWS Objects/fileobject.c Python/import.c description: Issue #7732: Fix a crash on importing a module if a directory has the same name than a Python module (e.g. "__init__.py"): don't close the file twice. PyFile_FromFile() does also close the file if PyString_FromString() failed. It did already close the file on fill_file_fields() error (e.g. if the file is a directory). |
|||
msg157563 - (view) | Author: Antoine Pitrou (pitrou) * | Date: 2012-04-05 11:27 | |
Victor, what exactly are you talking about? http://hg.python.org/cpython/rev/0f5b64630fda/ *does* change PyFile_FromString. And Matthias mentioned that the problem didn't occur with 2.6. |
|||
msg157574 - (view) | Author: Roundup Robot (python-dev) | Date: 2012-04-05 12:13 | |
New changeset 8258e5fa4a19 by Antoine Pitrou in branch '2.7': Issue #14505: Fix file descriptor leak when deallocating file objects created with PyFile_FromString(). http://hg.python.org/cpython/rev/8258e5fa4a19 |
|||
msg157575 - (view) | Author: Antoine Pitrou (pitrou) * | Date: 2012-04-05 12:16 | |
Matthias, this should be fixed now. |
|||
msg157607 - (view) | Author: STINNER Victor (vstinner) * | Date: 2012-04-05 17:25 | |
> Victor, what exactly are you talking about? http://hg.python.org/cpython/rev/0f5b64630fda/ *does* change PyFile_FromString. Oh. I don't know (remember) why I did this change!? I suppose that I changed it to test something and then forget to remove the temporary hack :-/ |
History | |||
---|---|---|---|
Date | User | Action | Args |
2022-04-11 14:57:28 | admin | set | github: 58710 |
2012-04-05 17:25:05 | vstinner | set | messages: + msg157607 |
2012-04-05 12:16:26 | pitrou | set | status: open -> closed type: resource usage messages: + msg157575 resolution: fixed stage: resolved |
2012-04-05 12:13:21 | python-dev | set | nosy:
+ python-dev messages: + msg157574 |
2012-04-05 11:27:12 | pitrou | set | messages: + msg157563 |
2012-04-05 11:20:07 | vstinner | set | messages: + msg157561 |
2012-04-05 10:41:49 | pitrou | set | messages: + msg157560 |
2012-04-05 10:30:12 | vstinner | set | nosy:
+ pitrou messages: + msg157559 |
2012-04-05 09:01:54 | doko | set | nosy:
+ vstinner |
2012-04-05 09:01:22 | doko | create |