| msg75779 - (view) |
Author: Matthias Klose (doko) *  |
Date: 2008-11-12 09:00 |
| ctypes fails to build on arm-linux-gnu, see http://python.org/dev/buildbot/2.5/ARM Linux 2.5/builds/3/step-compile/0 gcc -pthread -fPIC -fno-strict-aliasing -g -Wall -Wstrict-prototypes -I. -I/home/pybot/buildarea-arm/2.5.klose-linux-arm/build/./Include -Ibuild/temp.linux-armv5tel-2.5/libffi/include -Ibuild/temp.linux-armv5tel-2.5/libffi -I/home/pybot/buildarea-arm/2.5.klose-linux-arm/build/Modules/_ctypes/libffi/src -I. -IInclude -I./Include -I/usr/local/include -I/home/pybot/buildarea-arm/2.5.klose-linux-arm/build/Include -I/home/pybot/buildarea-arm/2.5.klose-linux-arm/build -c /home/pybot/buildarea-arm/2.5.klose-linux-arm/build/Modules/_ctypes/_ctypes.c -o build/temp.linux-armv5tel-2.5/home/pybot/buildarea-arm/2.5.klose-linux-arm/build/Modules/_ctypes/_ctypes.o In file included from /home/pybot/buildarea-arm/2.5.klose-linux-arm/build/Modules/_ctypes/_ctypes.c:127: /home/pybot/buildarea-arm/2.5.klose-linux-arm/build/Modules/_ctypes/ctypes.h:72: error: expected specifier-qualifier-list before 'ffi_closure' /home/pybot/buildarea-arm/2.5.klose-linux-arm/build/Modules/_ctypes/_ctypes.c: In function 'CFuncPtr_new': /home/pybot/buildarea-arm/2.5.klose-linux-arm/build/Modules/_ctypes/_ctypes.c:2955: error: 'CThunkObject' has no member named 'pcl' this is a regression compared to 2.5.2. Not seen on 2.6 or 3.0. Please consider a fix before 2.5.3 is released. |
|
|
| msg76185 - (view) |
Author: Barry A. Warsaw (barry) * |
Date: 2008-11-21 15:18 |
| Deferring until 3.0 final is out. |
|
|
| msg77491 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2008-12-10 08:26 |
| Thomas, do you have an idea what might be causing this problem, or a proposal how to resolve it? |
|
|
| msg77522 - (view) |
Author: Thomas Heller (theller) * |
Date: 2008-12-10 09:20 |
| Is this really a regression? I do not think so; the libffi library included with the Python 2.5 ctypes does not support closures on the arm. See file Modules/_ctypes/libffi/src/arm/ffitarget.h, it contains '#define FFI_CLOSURES 0'. |
|
|
| msg77524 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2008-12-10 09:38 |
| The question is whether it used to build in 2.5.2, but doesn't build anymore now. If that's what happening, it is a regression. It seems that ffi_info moved into ctypes.h: could that have caused the breakage? |
|
|
| msg77531 - (view) |
Author: Thomas Heller (theller) * |
Date: 2008-12-10 10:21 |
| > It seems that ffi_info moved into ctypes.h: could that have caused the > breakage? Not sure what you mean exactly (there is no ffi_info in ctypes.h), but I assume you are referring to svn rev. 62504? |
|
|
| msg77557 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2008-12-10 17:07 |
| That's the revision I meant, yes. It seems that ffi_info just got renamed, so I misread the change. In any case, it's (now) unclear how this did compile before, as ffi_closure was always referenced in ctypes.h... |
|
|
| msg77558 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2008-12-10 17:08 |
| Unfortunately, the ARM buildbot is down right now, so no tests can be performed. |
|
|
| msg77733 - (view) |
Author: Martin v. Löwis (loewis) * |
Date: 2008-12-13 14:15 |
| As the ARM buildbot is still down, we have no way of analysing or fixing this problem, so I'm closing this as "won't fix". |
|
|