| msg77718 - (view) |
Author: Johnny Lee (typo.pl) |
Date: 2008-12-13 07:56 |
| I ran my typo.pl perl script that locates possible C/C++ typos. I found four that looked valid. Two of the typos were in the Python directory {pythonrun.c, dynload_win.c}, two were in PC/bdist_wininst {install.c, extract.c}. Python/dynload_win.c: Win32 API FormatMessageW() expects the 6th parameter to be the count of characters, NOT the count of bytes. Python/pythonrun.c: The source code contains "if (ferr != NULL | |
ferr != Py_None)". This does not work as expected - if ferr == NULL, then the second part of the if expression will succeed. Look at the code handling fout about 8 lines up to see the correct code. PC/bdist_wininst/extract.c: Win32 API CreateFileMapping returns NULL on error, not INVALID_HANDLE_VALUE. PC/bdist_wininst/install.c: Win32 API CreateFileMapping returns NULL on error, not INVALID_HANDLE_VALUE. |
|
| msg77764 - (view) |
Author: Antoine Pitrou (pitrou) *  |
Date: 2008-12-13 21:09 |
| I don't know Windows APIs but the pythonrun.c bug is genuine. |
|
|
| msg77765 - (view) |
Author: Johnny Lee (typo.pl) |
Date: 2008-12-13 21:26 |
| Here are the URLs to the MSDN documentation for CreateFileMapping and FormatMessage[A|W]: <http://msdn.microsoft.com/en-us/library/aa366537.aspx> <http://msdn.microsoft.com/en-us/library/ms679351.aspx> For CreateFileMapping(), from the Return Value section: "If the function fails, the return value is NULL. To get extended error information, call GetLastError." For FormatMessage[A |
W]: nSize [in] If the FORMAT_MESSAGE_ALLOCATE_BUFFER flag is not set, this parameter specifies the size of the output buffer, in TCHARs. |
|
| msg77767 - (view) |
Author: Johnny Lee (typo.pl) |
Date: 2008-12-13 23:06 |
| For the dynload_win.c typo, it's technically a possible buffer overflow, but you'd need to find an error that had an error message that's longer than 259 chars. In pythonrun.c, the if statements for fout and ferr and almost identical. Probably a good idea to convert the if-statement to a separate function and call that function with fout and then ferr. |
|
|
| msg77916 - (view) |
Author: Mark Dickinson (mark.dickinson) * |
Date: 2008-12-16 17:01 |
| It looks to me as though all these are valid, and the patch should be applied to py3k and the 3.0 maintenance branch. Minor nit: the "sizeof(theInfo)" line exceeds 79 characters, in violation of PEP 7. Is there any core developer who's in a position to test this patch on Windows? |
|
|
| msg77948 - (view) |
Author: Amaury Forgeot d'Arc (amaury.forgeotdarc) * |
Date: 2008-12-17 00:24 |
| It's difficult to really test such errors. Moreover, on the top of the bdist_wininst files, a comment says loudly: IF CHANGES TO THIS FILE ARE CHECKED INTO PYTHON CVS, THE RECOMPILED BINARIES MUST BE CHECKED IN AS WELL! |
|
|
| msg77956 - (view) |
Author: Johnny Lee (typo.pl) |
Date: 2008-12-17 05:52 |
| attached modified diff patch so line length <=79 chars |
|
|
| msg78054 - (view) |
Author: Johnny Lee (typo.pl) |
Date: 2008-12-19 06:34 |
| > It's difficult to really test such errors. When I can't control the called function, I usually step through the code in a debugger and change the result variable in question to the appropriate value to see if the code handles failed function calls correctly. In the three Win32 API cases, it's clear that the failure path was not tested. And I can't see how the ferr if statement was tested with a failure value either. |
|
|
| msg105675 - (view) |
Author: STINNER Victor (vstinner) * |
Date: 2010-05-14 01:00 |
| I fixed the typo in pythonrun.c: r81156. I don't use Windows, so I cannot review the other fixes. |
|
|
| msg106636 - (view) |
Author: Alexander Belopolsky (belopolsky) * |
Date: 2010-05-28 04:46 |
| I don't run Windows either but I believe sizeof(char) == 1 is guaranteed by the C standard, so the first report (Python/dynload_win.c) is invalid. The last two appear to be valid [1], and a common mistake. [2] Unit tests are needed for these. I also notice that r81156 does not add a unit test either. ISTM, a test case can be crafted by setting sys.stderr to None. [1] http://msdn.microsoft.com/en-us/library/aa366537(VS.85).aspx [2] http://blogs.msdn.com/b/oldnewthing/archive/2004/03/02/82639.aspx |
|
|
| msg106637 - (view) |
Author: Alexander Belopolsky (belopolsky) * |
Date: 2010-05-28 05:03 |
| Strike my comment about Python/dynload_win.c - I was looking at the trunk version instead of py3k. In py3k theInfo is wchar_t and patch seems to be valid. |
|
|
| msg138220 - (view) |
Author: Éric Araujo (eric.araujo) * |
Date: 2011-06-12 20:36 |
| Martin would be the expert for bdist_wininst. |
|
|
| msg185104 - (view) |
Author: Roundup Robot (python-dev)  |
Date: 2013-03-23 23:06 |
| New changeset 4db1b0bb3683 by Gregory P. Smith in branch '3.3': Fixes - Correctly specify the buffer size to FormatMessageW and http://hg.python.org/cpython/rev/4db1b0bb3683 New changeset ace52be8da89 by Gregory P. Smith in branch 'default': Fixes - Correctly specify the buffer size to FormatMessageW and http://hg.python.org/cpython/rev/ace52be8da89 |
|
|
| msg185105 - (view) |
Author: Gregory P. Smith (gregory.p.smith) * |
Date: 2013-03-23 23:08 |
| the pythonrun issue had already been fixed. the others were still there (valid bugs, but were unlikely to be causing problems given the codepaths). fixed. I didn't write a Misc/NEWS entry on the commit because I didn't know how to usefully describe these minor fixes. someone else is welcome to contribute one if they feel the need. |
|
|