➜
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 2012-05-12 16:40 by pitrou, last changed 2022-04-11 14:57 by admin.
| Files | ||||
|---|---|---|---|---|
| File name | Uploaded | Description | Edit | |
| setup_lib64.patch | pitrou, 2012-05-12 16:40 | review | ||
| setup.diff | doko, 2012-07-08 00:24 | review | ||
| Messages (6) | |||
|---|---|---|---|
| msg160478 - (view) | Author: Antoine Pitrou (pitrou) *  |
Date: 2012-05-12 16:40 | |
setup.py adds <prefix>/lib to the list of directories where libraries are looked for, but not <prefix>/lib64. Patch attached. |
|||
| msg160483 - (view) | Author: Arfrever Frehtes Taifersar Arahesis (Arfrever) *  |
Date: 2012-05-12 17:02 | |
It would be better to fix issue #1294959 and use os.path.join(sys.prefix, sys.libdir). IIRC some mips architectures need "/usr/lib32", while x32 architecture needs "/usr/libx32". |
|||
| msg164956 - (view) | Author: Matthias Klose (doko) * |
Date: 2012-07-08 00:24 | |
using sys.maxsize isn't safe for cross builds. it should at least guarded by 'if not cross_compiling'. I still think that the preferred way to get the library dirs is to ask the compiler (patch attached), however this will add "incompatible" library dirs as well, at least for multilib'd gcc builds, and leading to linker warnings when building the extensions. At least when using gcc this should address the issue finding the lib32 and libx32 dirs. |
|||
| msg164965 - (view) | Author: Ned Deily (ned.deily) * |
Date: 2012-07-08 04:44 | |
Also OS X universal builds have many of the characteristics of a cross-build. In particular, one executable (like the python interpreter) can contain both 32-bit and 64-bit architecture files. So tests for 32-bit vs 64-bit at build time using the running compiler's arch are problematic. Another issue is the support for building with an OS X SDK. Basically, all /usr include and lib file paths used for searching in setup.py need to be prefixed by the SDK path if configured. There are some holes in this right now. I plan to fix them soon; it would better to have this issue resolved and applied first. |
|||
| msg380458 - (view) | Author: Irit Katriel (iritkatriel) * |
Date: 2020-11-06 17:05 | |
Was this resolved by the fix for https://bugs.python.org/issue32059? |
|||
| msg380492 - (view) | Author: Arfrever Frehtes Taifersar Arahesis (Arfrever) * |
Date: 2020-11-07 04:19 | |
No. setup.py still contains several hardcoded paths with "lib" (e.g. "/usr/local/lib", "/usr/lib/termcap"). There are also some places in setup.py which check both "lib64" and "lib", but such checks can possibly return incorrect results on multilib systems (when a library happens to be available for another ABI (32-bit vs 64-bit) but not ABI for which Python is being built). |
|||
| History | |||
|---|---|---|---|
| Date | User | Action | Args |
| 2022-04-11 14:57:30 | admin | set | github: 58996 |
| 2020-11-07 10:02:02 | iritkatriel | set | versions: + Python 3.8, Python 3.9, Python 3.10, - Python 3.2, Python 3.3 |
| 2020-11-07 04:19:06 | Arfrever | set | messages: + msg380492 |
| 2020-11-06 17:05:13 | iritkatriel | set | nosy:
+ iritkatriel messages: + msg380458 |
| 2012-07-08 04:44:29 | ned.deily | set | nosy:
+ ned.deily messages: + msg164965 |
| 2012-07-08 00:24:12 | doko | set | files:
+ setup.diff messages: + msg164956 |
| 2012-07-07 17:47:47 | pitrou | set | nosy:
+ doko |
| 2012-05-14 10:47:38 | jcea | set | nosy:
+ jcea |
| 2012-05-12 17:02:47 | Arfrever | set | nosy:
+ Arfrever messages: + msg160483 |
| 2012-05-12 16:40:42 | pitrou | create | |