Problems with python 1.0.3 on Solaris - but in general (fwd)

Brian de Alwis (bsdealwi@math.uwaterloo.ca)
Wed, 20 Jul 1994 13:16:14 -0400

This is a mail message I've sent to Guido. But, since he's on vacation,
I'm appealing to the python gurus on the list.

Forwarded message:
> Subject: Problems with python 1.0.3 on Solaris - but in general
> To: guido@cwi.nl
> Date: Wed, 20 Jul 1994 13:11:42 -0400 (EDT)
>
> Hello Guido,
>
> I'm trying to get python 1.0.3 up and running on an DEC Alpha box
> (OSF/1) but got a core dump on `make test'. I decided to try it on
> a Solaris 2 box as well, since it has Purify.
>
> I'm getting a whole mess of warnings like:
>
> ----------------
> **** Purify'd ./python (pid 19028) ****
> Purify (abr): array bounds read:
> * This is occurring while in:
> com_arglist [line 2180, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x86d20]
> compile_funcdef [line 2233, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x87250]
> compile_node [line 2300, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x8787c]
> compile [line 2462, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x88400]
> com_funcdef [line 1910, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x85874]
> com_node [line 1985, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x86104]
> * Reading 4 bytes from 0x18c59c in the heap (3 bytes at 0x18c59d illegal).
> * This is 60 bytes into a malloc'd block at 0x18c560 of 61 bytes.
> * This block was allocated from:
> malloc [bbsdealwi/src/python-1.0.3/./python, pc=0x1bbe0]
> realloc [bbsdealwi/src/python-1.0.3/./python, pc=0x1bec0]
> addchild [bbsdealwi/src/python-1.0.3/./python, pc=0xc44a8]
> shift [line 152, bbsdealwi/src/python-1.0.3/Parser/parser.c, pc=0xc7f58]
> addtoken [line 267, bbsdealwi/src/python-1.0.3/Parser/parser.c, pc=0xc8544]
> parsetok [line 147, bbsdealwi/src/python-1.0.3/Parser/parsetok.c, pc=0xc4c28]
> ----------------
>
> and
>
> ----------------
> **** Purify'd ./python (pid 19028) ****
> Purify (umr): uninitialized memory read:
> * This is occurring while in:
> com_arglist [line 2182, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x86d40]
> compile_funcdef [line 2233, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x87250]
> compile_node [line 2300, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x8787c]
> compile [line 2462, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x88400]
> com_funcdef [line 1910, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x85874]
> com_node [line 1985, bbsdealwi/src/python-1.0.3/Python/compile.c, pc=0x86104]
> * Reading 4 bytes from 0xefffd7d4 on the stack.
> * This is local variable "t" in function com_arglist.
> ----------------
>
> These occur quite often.
>
> It seems to indicate a problem with addchild - if I change the XXXROUNDUP
> macro, the array bounds problems disappear, but the the uninitialized memory
> read errors are still present.
>
> >From the purify docs:
>
> ABR, Array Bounds Read Warning
>
> An ABR report indicates that your program is about to read a
> value from before or after an allocated block.
>
> and umr is self-explanatory.
>
> Incidentally, the core dump with the Alpha box were due (I think) to
> using gcc-2.5.8 -- compilation with 2.6.0 doesn't dump. Instead, the
> test comes out like:
>
> ----------------
> PYTHONPATH=./Lib:./Lib/test ./python -c 'import autotest'
> All tests OK.
> PYTHONPATH=./Lib:./Lib/test ./python -c 'import autotest'
> Traceback (innermost last):
> File "<string>", line 0
> File "./Lib/test/autotest.py", line 67
> main()
> File "./Lib/test/autotest.py", line 62
> import testall
> File "./Lib/test/testall.py", line 22
> import test_builtin
> File "./Lib/test/test_builtin.py", line 9
> import test_b1
> ImportError: Bad code object in .pyc file
> ----------------
>
> If you need more details, more tests, or a full copy of the Purify
> report, please let me know.
>

-- 
Brian de Alwis - University of Waterloo: bsdealwis@math.uwaterloo.ca
`How extraordinarily lucky Minta is! She is marrying a man who has a gold
 watch in a wash-leather bag!' - _To_the_Lighthouse_, Virginia Woolf.