test_gettext.py fails on 64-bit architectures
(Tru64 Unix) - test_gettext fails with the message: IOError: [Errno 0] Bad magic number: './xx/LC_MESSAGES/gettext.mo' This is because the magic number is read in by the code in Lib/gettext.py as FFFFFFFF950412DE (hex) (using unpack('<i', buf[:4])[0]), and checked against LE_MAGIC (defined as 950412DE) and BE_MAGIC (calculated as FFFFFFFFDE120495 using struct.unpack('>i',struct.pack('<i', LE_MAGIC))[0]) These format strings work for machines where a Python integer is the same size as a C int, but not for machines where a Python integer is larger than a C int. The problem arises because the LE_MAGIC number is negative if a 32-bit int, but positive if Python integers are 64-bit. Replacing the "i" in the code that generates BE_MAGIC and reads in "magic" by "I" makes the test work for me, but there's other uses of "i" and "ii" when the rest of the .mo file is processed that I'm unsure about with different inputs. Mark -- Email - m.favas@per.dem.csiro.au Postal - Mark C Favas Phone - +61 8 9333 6268, 041 892 6074 CSIRO Exploration & Mining Fax - +61 8 9333 6121 Private Bag No 5 Wembley, Western Australia 6913
"MF" == Mark Favas <m.favas@per.dem.csiro.au> writes:
MF> This is because the magic number is read in by the code in MF> Lib/gettext.py as FFFFFFFF950412DE (hex) (using unpack('<i', MF> buf[:4])[0]), and checked against LE_MAGIC (defined as MF> 950412DE) and BE_MAGIC (calculated as FFFFFFFFDE120495 using MF> struct.unpack('>i',struct.pack('<i', LE_MAGIC))[0]) I was trying to be too clever. Just replace the BE_MAGIC value with 0xde120495, as in the included patch. MF> Replacing the "i" in the code that generates BE_MAGIC and MF> reads in "magic" by "I" makes the test work for me, but MF> there's other uses of "i" and "ii" when the rest of the .mo MF> file is processed that I'm unsure about with different inputs. Should be fine, I think. With < and > leading characters, those format strings should select `standard' sizes: Standard size and alignment are as follows: no alignment is required for any type (so you have to use pad bytes); short is 2 bytes; int and long are 4 bytes. float and double are 32-bit and 64-bit IEEE floating point numbers, respectively. Please run the test again with this patch and let me know. -Barry Index: gettext.py =================================================================== RCS file: /cvsroot/python/python/dist/src/Lib/gettext.py,v retrieving revision 1.4 diff -u -r1.4 gettext.py --- gettext.py 2000/08/30 03:29:58 1.4 +++ gettext.py 2000/08/31 10:40:41 @@ -125,7 +125,7 @@ class GNUTranslations(NullTranslations): # Magic number of .mo files LE_MAGIC = 0x950412de - BE_MAGIC = struct.unpack('>i', struct.pack('<i', LE_MAGIC))[0] + BE_MAGIC = 0xde120495 def _parse(self, fp): """Override this method to support alternative .mo formats."""
participants (2)
-
bwarsaw＠beopen.com -
Mark Favas