| msg121954 - (view) |
Author: Antoine Pitrou (pitrou) *  |
Date: 2010-11-21 17:24 |
| $ 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: > (1)() (Pdb) next 12 --Return-- > (1)()->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()->('', 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 |
|
|
| msg121957 - (view) |
Author: Bruno Gola (brunogola) |
Date: 2010-11-21 18:26 |
| tests are OK for me running on ubuntu 10.04 64bits and py3k from svn repository. |
|
|
| msg123771 - (view) |
Author: STINNER Victor (vstinner) * |
Date: 2010-12-11 03:38 |
| 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 "" (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 |
|
|
| msg123774 - (view) |
Author: STINNER Victor (vstinner) * |
Date: 2010-12-11 04:00 |
| See a more complex solution: #3080 (don't decode the filename in the parser, keep unicode strings). |
|
|
| msg125487 - (view) |
Author: STINNER Victor (vstinner) * |
Date: 2011-01-06 00:23 |
| > 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 |
|
|
| msg125492 - (view) |
Author: STINNER Victor (vstinner) * |
Date: 2011-01-06 00:51 |
| 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). |
|
|
| msg125496 - (view) |
Author: STINNER Victor (vstinner) * |
Date: 2011-01-06 01:11 |
| "x86 debian parallel 3.x" buildbot is green again! :-) |
|
|