On 2016-12-17 Sat 06:58 GMT-0800, Eli Zaretskii <eliz@gnu.org> wrote:

>> From: Hong Xu <hong@topbug.net>
>> Date: Thu, 01 Dec 2016 20:04:00 -0800
>>
>> +;;;###autoload
>>  (defun parse-iso8601-time-string (date-string)
>> +  "Parse an ISO 8601 time string, such as 2016-12-01T23:35:06-05:00.
>> +If DATE-STRING cannot be parsed, it falls back to
>> +`parse-time-string'."
>
> Why did you want to autoload this?  The log message doesn't mention
> this change at all.

The attached new version now fixes this issue in the log message. It
autoloads because parse-time-string autoloads -- If parse-time-string is
useful for autoloading, there is no reason that
parse-iso8601-time-string is not useful for that.

>
>> +                 '(42 35 19 22 2 2016 1 t -25200)))
>> +  (should (equal (parse-iso8601-time-string "1998-09-12T12:21:54-0200")
>> +                 '(13818 33666)))
>> +  (should (equal (parse-iso8601-time-string "1998-09-12T12:21:54-02:00")
>> +                 '(13818 33666)))
>> +  (should (equal (parse-iso8601-time-string "1998-09-12T12:21:54-02")
>> +                 '(13818 33666)))
>> +  (should (equal (parse-iso8601-time-string "1998-09-12T12:21:54+02")
>> +                 '(13818 19266))))
>
> Does this also test this part of your changes:
>
>> @@ -204,13 +204,17 @@ parse-time-iso8601-regexp
>>  	 (date-time (concat full-date "T" full-time)))
>>      (list (concat "^" full-date)
>>  	  (concat "T" partial-time)
>> -	  (concat "Z" time-numoffset)))
>> +	  (concat "Z?" time-numoffset)))
>
> If not, could you please add tests for that?
>

I've added the tests now. And indeed, it exposes a bug in the original patch.

Again, I decomposed the patch to two different parts, one for the stable
branch and the other (with only test files) for the master branch.