[issue5665] Add more pickling tests
Collin Winter coll...@gmail.com added the comment: Committed as r71408 (trunk) and r71638 (py3k). -- resolution: - fixed status: open - closed ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Benjamin Peterson benja...@python.org added the comment: Ok. Thank you! -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Benjamin Peterson benja...@python.org added the comment: Collin, can you port these new tests to Py3k? -- nosy: +benjamin.peterson ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Collin Winter coll...@gmail.com added the comment: Yes, I'm porting them. I got started, but got distracted by other things since there were a lot of conflicts in the merge -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Collin Winter coll...@gmail.com added the comment: If no-one has any objections to the xpickle resource included in the latest version of the patch, I'd like to commit this soon so that we can be more confident in the other changes I have queued up. If I no-one objects, I'll commit this sometime early tomorrow morning PDT. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Antoine Pitrou pit...@free.fr added the comment: If no-one has any objections to the xpickle resource included in the latest version of the patch, I'd like to commit this soon so that we can be more confident in the other changes I have queued up. If I no-one objects, I'll commit this sometime early tomorrow morning PDT. Ok if it doesn't take too long to run the tests (which may imply implementing something like Raymond's suggestion of randomizing test order, if you haven't already done so). Thanks for your work, cheers Antoine. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Collin Winter coll...@gmail.com added the comment: Ok if it doesn't take too long to run the tests (which may imply implementing something like Raymond's suggestion of randomizing test order, if you haven't already done so). I did something similar: if you don't pass the -uxpickle flag to regrtest, only a small subset of the tests will be run, which takes less than half a second on my machine; if you do pass -uxpickle, the full battery of tests will run, which takes ~5 minutes. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Collin Winter coll...@gmail.com added the comment: I've made test_xpickle support Python 2.4 because it uses Python 2.4, 2.5 and 2.6 to check that we haven't broken backwards compatibility with those versions. I made AbstractCompatTests use cPickle because that's what I've been changing :) I can add tests for the pure-Python pickle module if you want. At that point, though, test_xpickle will be very slow and we'll want to add a regrtest resource to control how much time test_xpickle spends running. Does xpickle work for the resource name (ie, regrtest.py -u xpickle)? I agree with your other points. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Antoine Pitrou pit...@free.fr added the comment: How much is very slow? If it's less than 10-15 seconds I think it's ok. -- nosy: +pitrou ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Collin Winter coll...@gmail.com added the comment: test_xpickle currently takes over three minutes on my MacBook Pro to test compatibility with Python 2.4, 2.5 and 2.6, and adding tests for the pickle module will at least double that, if not push it well above 10 minutes. That's why I recommend using the resource flag if we want to add the same tests for the pickle module. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Raymond Hettinger rhettin...@users.sourceforge.net added the comment: We took a similar issue in test_decimal and came up with a better approach to using the resource flag. If the resource is not enabled, the tester runs a random subset of the tests. Over time, that means that all the tests will get run even if testers are routinely not enabling the resource. -- nosy: +rhettinger ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Antoine Pitrou pit...@free.fr added the comment: test_xpickle currently takes over three minutes on my MacBook Pro to test compatibility with Python 2.4, 2.5 and 2.6, and adding tests for the pickle module will at least double that, if not push it well above 10 minutes. Ouch! A couple of minutes is already too much for a single test IMHO. I fully agree with the desire to test as deeply as possible, but if all modules were to become as well tested, nobody would bother running the regression suite anymore :-) -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Antoine Pitrou pit...@free.fr added the comment: Raymond, how would you reconcile your random approach with the fact that buildbots only display errors after a second verification run? -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Raymond Hettinger rhettin...@users.sourceforge.net added the comment: The buildbots should be running with the resource enabled (run the full set of tests, every time). If we need the reproducible results with the resource disabled, then a random seed (different for each successful run) can be saved in a logfile so that a particular pattern of tests can be re-run. The idea to run some random tests when the resource flag is disabled provides more coverage than not running any tests at all. This was very helpful during the early development of the decimal module. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Collin Winter coll...@gmail.com added the comment: Updated the patch to include a regrtest xpickle resource, which restricts if the backwards compat tests will be run. For normal development, the existing tests should be fine; you only need these tests if you're mucking with the pickle output. Added backwards compat tests for the pure-Python pickle module. Total run time with -uxpickle maxes out a little above five minutes on my machine, with a lot of variation between runs. Also addressed the comment about [0, 1, 2] - protocols. -- Added file: http://bugs.python.org/file13599/pickle_tests.patch ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Changes by Antoine Pitrou pit...@free.fr: -- nosy: +alexandre.vassalotti ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
Alexandre Vassalotti alexan...@peadrop.com added the comment: Overall, the patch looks good. I haven't reviewed the patch thoroughly yet, but there is a few things I am not sure about. First, why do you bother supporting 2.4 (i.e. with a copy of run_with_locale) in a patch aimed at 2.7? In test_many_puts_and_gets(), I believe this: for proto in [0, 1, 2]: ... should be changed to: for proto in protocols: ... Also, I think the tests in AbstractPicklerUnpicklerObjectTests could be stronger if they used the pickletools module to decompile to check that only PUT opcodes have changed or not. However, the extra complexity may not worth it. And one last thing, why AbstractCompatTests is hard-coded to use cPickle? Shouldn't the standard pickle module be tested too? -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue5665] Add more pickling tests
New submission from Collin Winter coll...@gmail.com: The attached patch adds more tests for pickling: - Add tests for the module-level load() and dump() functions. - Add tests for cPickle's internal data structures, stressing workloads with many gets/puts. - Add tests for the Pickler and Unpickler classes, in particular the memo attribute. - test_xpickle is extended to test backwards compatibility with Python 2.4, 2.5 and 2.6 by round-tripping pickled objects through a worker process. 2.3 was too difficult to support. I'll port to py3k after reviewed for trunk. -- components: Tests files: pickle_tests.patch keywords: needs review, patch messages: 85162 nosy: collinwinter severity: normal status: open title: Add more pickling tests type: behavior versions: Python 2.7, Python 3.1 Added file: http://bugs.python.org/file13571/pickle_tests.patch ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue5665 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
