Add fallback to ISO 8601 date format in ContentDispositionParser.

[reitermarkus]

Some websites incorrectly use ISO 8601 for the dates, causing an exception to be raised and rendering all other correctly specified fields unusable.

[reitermarkus]

Ping?

[flavorjones]

Hi, thanks for proposing this change, and apologies you haven't gotten any feedback until now.

There are tests covering the behavior of Mechanize::HTTP::ContentDispositionParser in test/test_mechanize_http_content_disposition_parser.rb, but notably they do not address the parser's behavior when a header is encountered that does not follow RFC 2183 to use RFC 822 dates.

I'd be open to merging this if I saw, as part of this PR:

• a test describing current behavior when given an unparseable date string
• test describing the new behavior when using ISO8601 date strings
• and the first test should continue to pass before and after the changes to ContentDispositionParser

Thanks for being open to this request.

[reitermarkus]

a test describing current behavior when given an unparseable date string

The current behavior is that it raises an ArgumentError, however, it should return nil given that all other wrong fields also don't result in an error but nil being returned.

[reitermarkus]

@flavorjones, I added some tests. The other test failures seem unrelated.

[flavorjones]

Hi @reitermarkus, thanks for adding the test coverage. The change of behavior (from raising an exception to returning nil) makes sense to me (though I'm not sure why the current implementation returns nil rather than simply omitting the key or value from the returned hash -- that's for another day).

The existing test failures are ... well, that's not great. Mechanize has not received much attention from maintainers recently. I'll try to set aside some time over the holidays to try to get the tests in shape again.

[flavorjones]

I just rebased off master, let's see what CI thinks now.

[flavorjones]

Looks good, merging. Will be in the next release, v2.8.0.