[PATCH 02 of 10 V2] revlog: fix pure python slicing code when chain contains nullrev

[Boris FELD]

On 30/12/2018 09:06, Yuya Nishihara wrote:
> On Fri, 28 Dec 2018 19:07:04 +0100, Boris FELD wrote:
>> On 28/12/2018 06:23, Yuya Nishihara wrote:
>>> On Fri, 28 Dec 2018 00:24:48 +0100, Boris FELD wrote:
>>>>> Can you turn these tests into a unittest to cover both pure and C
>>>>> implementations?
>>>> Turning this specific test is a bit more complicated than we would like
>>>> as we are taking advantage of _testrevlog flexibility to produce a
>>>> simple but clear test.
>>>>
>>>> However, we've already added some unittest using real revlog in patches
>>>> you have already taken. Is that enough for you?
>>> The unittest you wrote is nice, thanks, but it's even nicer if we had
>>> a single test set that covers both pure and cext. It'll work as a source
>>> of truth. At this point, we have to carefully compare both implementations
>>> to be sure they are identical.
>> I'm not sure to understand. The unittest will use the pure code when run
>> with --pure so both the C and pure code is tested. Am I missing something?
> If the unittest covers all scenarios in the doctest, perhaps the doctest can
> be removed. I just guessed it wouldn't because you've sent this patch.
As you pointed out, the error was actually in _testrevlog not in the
implementation.
> _______________________________________________
> Mercurial-devel mailing list
> Mercurial-devel at mercurial-scm.org
> https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel