| msg40306 - (view) |
Author: Gustavo Niemeyer (niemeyer) *  |
Date: 2002-06-14 18:53 |
| As discussed in python-dev, st_rdev provides an integer, which usually needs to be translated in a (major, minor) pair before it's useful. There are no implementations of major() and minor() in the current library, and maybe there shouldn't be, since their functionality is not portable. This patch provides a portable way to access this information, by providing a new attribute to the os.stat() return tuple named st_rdev_pair. |
|
|
| msg40307 - (view) |
Author: Neal Norwitz (nnorwitz) * |
Date: 2002-06-14 21:36 |
| Logged In: YES user_id=33168 Are major/minor guaranteed to exist? Should a check be added to configure? There is a missing DECREF. In the last line of the patch to posixmodule.c, there is if (PyErr_Occurred()) { In that block, w needs to be DECREFed. You can initialize w = NULL in the declaration, then do an XDECREF in the error block. That way you don't need to handle the error creating w, if w != NULL, set the items in the tuple. |
|
|
| msg40308 - (view) |
Author: Gustavo Niemeyer (niemeyer) * |
Date: 2002-06-14 21:47 |
| Logged In: YES user_id=7887 I don't think a test is needed. If st_rdev is available, this information must be decodable by the system, and this is done by the major/minor pair, AFAIK. About the DECREF, I don't think it's needed. w's reference is stolen by v when PyStructSequence_SET_ITEM is run, and will be destroyed with v once it's DECREF'd. If that's not true, the whole logic is wrong in that function. |
|
|
| msg40309 - (view) |
Author: Neal Norwitz (nnorwitz) * |
Date: 2002-06-14 22:02 |
| Logged In: YES user_id=33168 You are correct about the DECREF, it is not necessary. I'm not sure if any standards define major/minor (that's why I asked). What platform(s) did you test on? |
|
|
| msg40310 - (view) |
Author: Gustavo Niemeyer (niemeyer) * |
Date: 2002-06-14 22:23 |
| Logged In: YES user_id=7887 I'll redo the patch with a test. Thanks! |
|
|
| msg40311 - (view) |
Author: Gustavo Niemeyer (niemeyer) * |
Date: 2002-06-15 19:04 |
| Logged In: YES user_id=7887 I'm not sure if that's really the right way to implement the needed features for tarfile. Maybe including major, minor, and makedev could be the way to go. I'm sending the rationale to python-dev for discussion, and attaching a new patch with their implementations, instead of st_rdev_pair. |
|
|
| msg40312 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2002-10-07 21:48 |
| Logged In: YES user_id=21627 Is anything stopping this patch? If not, I'll apply it. |
|
|
| msg40313 - (view) |
Author: Neal Norwitz (nnorwitz) * |
Date: 2002-10-08 19:17 |
| Logged In: YES user_id=33168 I'm fine with checking in. |
|
|
| msg40314 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2002-10-10 14:28 |
| Logged In: YES user_id=21627 Thanks for the patch. Committed as configure 1.345; configure.in 1.356; pyconfig.h.in 1.53; libos.tex 1.98; posixmodule.c 2.265; |
|
|