Re: [random] Merging to master for 1.75 release

From: Edward Diener (eldiener_at_[hidden])
Date: 2020-11-25 00:19:50

On 11/24/2020 5:52 PM, Marshall Clow via Boost wrote:
> 
> 
>> On Nov 24, 2020, at 12:47 PM, Edward Diener via Boost <boost_at_[hidden]> wrote:
>>
>> On 11/24/2020 2:10 PM, Marshall Clow via Boost wrote:
>>> On Nov 24, 2020, at 10:55 AM, Steven Watanabe via Boost <boost_at_[hidden]> wrote:
>>>>
>>>> AMDG
>>>>
>>>> On 11/18/20 11:18 AM, Edward Diener via Boost wrote:
>>>>> On 11/18/2020 10:18 AM, Glen Fernandes via Boost wrote:
>>>>>> On Wed, Nov 18, 2020 at 5:55 AM Edward Diener  wrote:
>>>>>>>
>>>>>>> Is there any possibility of merging 'develop' to 'master' for the random
>>>>>>> library for the 1.75 release ? The 'master' branch has not been updated
>>>>>>> for 2 years, or since the 1.70 release. While I am not claiming all
>>>>>>> invaluable fixes I know my fix for Embarcadero C++ clang-based support
>>>>>>> and another fix for correctly auto-linking random_device is among the
>>>>>>> 'develop' branch fixes which have never made it to the 'master' branch.
>>>>>>
>>>>>> I'm assuming this is for Steven (who is the active maintainer of
>>>>>> Random). i.e. Unlike the other repositories which we've helped out
>>>>>> with, that are either under CMT or no active maintainer.
>>>>>
>>>>> A maintainer of a library has the right to do what he wants. But it does
>>>>> not seem that Steven is aware of the situation. Also others may have
>>>>> write access to 'random' for all I know, and be able to review the merge
>>>>> and make the change. In the 'develop' regression tests for 'random'
>>>>> there are no issues which I can see in the results and the last change
>>>>> was over 4 months ago.
>>>>>
>>>>
>>>> I can merge it now if it's okay with the release managers.
>>> I looked at the commits since the last merge, and only this one looks problematic for a post-beta merge.
>>> (We try not to add new features after the beta)
>>> https://github.com/boostorg/random/commit/b8d5df31168a07e04a175ace751f0defa7be5e28 <https://github.com/boostorg/random/commit/b8d5df31168a07e04a175ace751f0defa7be5e28><https://github.com/boostorg/random/commit/b8d5df31168a07e04a175ace751f0defa7be5e28 <https://github.com/boostorg/random/commit/b8d5df31168a07e04a175ace751f0defa7be5e28>>
>>> All the rest are fine
>>
>> The commit you cite above does not change anything for the end-user. All the fix does is use type_traits rather than mpl in a number of cases, which removes the mpl dependency for the 'random' library, and all those cases are  private implementation details which does not affect the end-user's use of the library in any way. This commit is hardly adding a new feature to the 'random' library.
> 
> Then I copy/pasted the wrong URL; sorry.
> 
> https://github.com/boostorg/random/commit/1182a4071e6bcdaa94e4c49e4a220885b0dc2748 <https://github.com/boostorg/random/commit/1182a4071e6bcdaa94e4c49e4a220885b0dc2748>
> 
> This looks like a new feature to me.
> (I am willing to be convinced otherwise, though)

It is a new feature, a new generator called 'mixmax'. It appears to be 
tested pretty thoroughly and does not affect any other generator AFAICS. 
It is your call to Steven, but I would very much like to see 'random' 
develop merged to 'master' so the Embarcadero clang C++ support goes in. 
That would complete the basic support for the Embarcadero C++ 
clang-based compilers for all the Boost libraries. I would like to 
report that back to my contact at Embarcadero and then maybe they would 
take a more pro-active stance in supporting current Boost with their 
clang-based compilers.