Author martin.panter
Recipients martin.panter
Date 2016-01-28.01:24:34
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1453944276.21.0.50318794169.issue26173@psf.upfronthosting.co.za>
In-reply-to
Content
There are more mistakes in the history of test_nonexisting_cert(). In revision 8a281bfc058d (Python 2.6), the method was added as testWrongCert(), with an existing but non-matching certificate file. But when this was ported to Python 3 in r66311, the wrongcert.pem file was not added, so Python 3 was actually testing the behaviour when the specified certificate file was missing. Then in r80596, the test method was renamed and a comment added assuming the Python 3 version with the missing file. However we already test the behaviour of missing files in test_errors().

I do not understand the ECONNRESET failure on Windows. Perhaps there is a race to do with the server closing the connection when the client should be reporting a certificate mismatch. It seems like a bug, and I suspect r80534 is not the correct fix. But Iā€™m not in a position to investigate so I will leave that code as it is.

For Python 2 I propose wrong-cert-py2.patch:

* Rename WRONGCERT ā†’ NONEXISTINGCERT, not to be confused with wrongcert.pem
* Repurpose test_nonexisting_cert() ā†’ test_wrong_cert()
* Remove ENOENT exception handling from bad_cert_test()
History
Date User Action Args
2016-01-28 01:24:36martin.pantersetrecipients: + martin.panter
2016-01-28 01:24:36martin.pantersetmessageid: <1453944276.21.0.50318794169.issue26173@psf.upfronthosting.co.za>
2016-01-28 01:24:36martin.panterlinkissue26173 messages
2016-01-28 01:24:34martin.pantercreate