| msg15547 - (view) |
Author: Jack Jansen (jackjansen) *  |
Date: 2003-04-20 22:14 |
| Adding this just so it doesn't fall by the roadside (it's been discussed on python-dev): test_getargs2 is currently failing, at least on MacOSX. |
|
|
| msg15548 - (view) |
Author: Neal Norwitz (nnorwitz) * |
Date: 2003-04-21 17:42 |
| Logged In: YES user_id=33168 This is failling on several other architectures in the snake farm: http://www.lysator.liu.se/xenofarm/python/latest.html |
|
|
| msg15549 - (view) |
Author: Neal Norwitz (nnorwitz) * |
Date: 2003-04-22 14:16 |
| Logged In: YES user_id=33168 This is fixed on all the snake-farm machines AFAIK. Jack, are you still having the problems? |
|
|
| msg15550 - (view) |
Author: Jack Jansen (jackjansen) * |
Date: 2003-04-23 10:39 |
| Logged In: YES user_id=45365 Yes, I still have the problem. The test output is attached. |
|
|
| msg15551 - (view) |
Author: Thomas Heller (theller) * |
Date: 2003-04-23 11:29 |
| Logged In: YES user_id=11105 Strange, since 196608 == 0x30000, and 50331648 == 0x3000000. Even stranger (to me) is that the test for upper case i works, and it uses nearly the same code. Does it help the test_H case to change line 485 of getargs.c from 'long ival;' to 'unsigned short ival;', but I'm only guessing? |
|
|
| msg15552 - (view) |
Author: Thomas Heller (theller) * |
Date: 2003-04-23 11:49 |
| Logged In: YES user_id=11105 Um, I think I found the reason. test_getargs2 uses getargs_ul (in _testcapimodule.c) to call PyArg_ParseTuple(), and this always passes a pointer to an unsigned long, even for H and B format codes. On little endian machines casting this pointer into a unsigned short pointer works ok, but not on big endian machines. Makes sense? |
|
|
| msg15553 - (view) |
Author: Guido van Rossum (gvanrossum) * |
Date: 2003-04-23 18:04 |
| Logged In: YES user_id=6380 Yes, that makes sense, and I remember seeing that and thinking, "hmm, that's not right". You probably need separate C functions for each C data type to test. :-( |
|
|
| msg15554 - (view) |
Author: Thomas Heller (theller) * |
Date: 2003-04-23 19:17 |
| Logged In: YES user_id=11105 Jack, if I try to fix this (by implementing the C functions needed) would you be able to test this, or should we simply comment out the failing tests and do this after the release? |
|
|
| msg15555 - (view) |
Author: Jack Jansen (jackjansen) * |
Date: 2003-04-24 13:12 |
| Logged In: YES user_id=45365 Yes, I will be able to test it. But it will have to be tomorrow. BTW: does this mean that there are *no* bigendian machines in the snakefarm? That surprises me, I seem to remember than almost all machines except intels (i.e. suns, hps, sgi's) are big-endian... |
|
|
| msg15556 - (view) |
Author: Thomas Heller (theller) * |
Date: 2003-04-24 16:20 |
| Logged In: YES user_id=11105 I've commited changes to Modules/_testcapimodule.c and Lib/test/test_getargs2.py which will hopefully fix the problems. Since this is only test code, I've checked it in without asking ;-) Re Big endian: I've got this impression simply by reading the code. But I've never programmed on such a machine. |
|
|
| msg15557 - (view) |
Author: Neal Norwitz (nnorwitz) * |
Date: 2003-04-24 19:10 |
| Logged In: YES user_id=33168 I think HPs and Alphas are not big-endian, but Sun's are. The Sun was the only one failling. With the last checkin, the Sun is now working. Hopefully, it works on OS X too. |
|
|
| msg15558 - (view) |
Author: Jack Jansen (jackjansen) * |
Date: 2003-04-25 09:03 |
| Logged In: YES user_id=45365 It works! Thanks everyone! |
|
|