homepage

$ LANG=ISO-8859-15 ./python -m test.regrtest test_doctest
[1/1] test_doctest
**********************************************************************
File "/home/antoine/py3k/__svn__/Lib/test/test_doctest.py", line 1676, in test.test_doctest.test_debug
Failed example:
 try: doctest.debug_src(s)
 finally: sys.stdin = real_stdin
Expected:
 > <string>(1)<module>()
 (Pdb) next
 12
 --Return--
 > <string>(1)<module>()->None
 (Pdb) print(x)
 12
 (Pdb) continue
Got:
 > /home/antoine/py3k/__svn__/Lib/encodings/iso8859_15.py(15)decode()
 -> return codecs.charmap_decode(input,errors,decoding_table)
 (Pdb) next
 --Return--
 > /home/antoine/py3k/__svn__/Lib/encodings/iso8859_15.py(15)decode()->('<string>', 8)
 -> return codecs.charmap_decode(input,errors,decoding_table)
 (Pdb) print(x)
 *** NameError: name 'x' is not defined
 (Pdb) continue
 12
**********************************************************************
1 items had failures:
 1 of 4 in test.test_doctest.test_debug
***Test Failed*** 1 failures.
test test_doctest failed -- 1 of 429 doctests failed
1 test failed:
 test_doctest
Also visible on the following buildbot:
http://www.python.org/dev/buildbot/all/builders/x86%20debian%20parallel%203.x/builds/934/steps/test/logs/stdio 

tests are OK for me
running on ubuntu 10.04 64bits and py3k from svn repository.

You can reproduce the bug with:
$ LANG=fr_FR.iso885915@euro ./python -c 'import pdb; pdb.Pdb(nosigint=True).run("exec(%r)" % "x=12")'
> /home/haypo/prog/SVN/py3k/Lib/encodings/iso8859_15.py(15)decode()
-> return codecs.charmap_decode(input,errors,decoding_table)
(Pdb) quit
(it should print "x=12" in the backtrace, not ...iso8859_15.py...)
Simplified C backtrace: builtin_exec() -> PyRun_StringFlags() -> PyAST_CompileEx() -> makecode() -> PyUnicode_DecodeFSDefault().
ISO-8859-15 codec is implemented in Python whereas ASCII, ISO-8859-1 and UTF-8 are implemented in C. Pdb stops at the first Python instruction. The user expects that the first instruction is "x=12", but no, the real first Python instruction is calling ISO-8859-15 to decode the byte string "<string>" (script filename).
I see two solutions:
 - set the trace function later. Eg. replace exec(cmd, ...) by code=compile(cmd, ...) + exec(code) and set the trace function after the call to compile. I don't know if both codes are equivalent.
 - reimplement ISO-8859-15 in Python: it doesn't solve the issue, there are other encodings implemented in Python

See a more complex solution: #3080 (don't decode the filename in the parser, keep unicode strings).

> set the trace function later. Eg. replace exec(cmd, ...)
> by code=compile(cmd, ...) + exec(code) and set the trace function 
> after the call to compile.
Implemented in the attached patch, bdb.patch: trace the execution of the code, not the compilation of the code.
> I don't know if both codes are equivalent.
I still don't know :-p

bdb.patch doesn't work if cmd is not a string (if cmd is a code object).
r87780 fixes this issue: bdb.Bdb.run() only traces the execution of the code, not the compilation (if the input is a string).
With this fix, the whole test suite pass on Linux with ISO-8859-1, ISO-8859-15 and UTF-8 locale encodings (I only tested in an ASCII path).

"x86 debian parallel 3.x" buildbot is green again! :-)