Re: [boost] "peer reviewed" - Rights and responsibilities of maintainers

Subject: Re: [boost] "peer reviewed" - Rights and responsibilities of maintainers
From: Alexander Grund (alexander.grund_at_[hidden])
Date: 2018-10-17 07:27:49

I really wonder how someone can show that much ignorance if all facts 
are against him.

"#111 doesn't show anything", " I have no problem if the author want's 
to improve his test.  It's just not happening. "
EVERYONE sees how it does show a crash. I explained that it is a MWE 
reproducer of a real application crash. So while the test may SEEM to 
not be doing much useful it DOES use Boost.Serialization and it DOES 
crash as described. There were NO arguments on WHY this test is invalid 
C++ code.

Based on his comments (after months) I slightly changed the test and now 
I even created #128 which shall show a small, but valid and regular use 
of Boost.Serialization and it shows the exact same behaviour.

Isn't it important to add tests for bugs found? Isn't it important to 
have these tests be as minimal as possible, so side effects are avoided? 
How can he still claim the test "doesn't show anything" if it clearly 
shows a construction/destruction problem IN the library?

He also still fails to acknowledge the broken `is_destroyed` function. A 
good C++ programmer can see this from the code (just ask: When is 
is_destroyed set to false? When can this fail to happen?) and there is a 
test case where CI says "|Assertion failed: 
boost::serialization::singleton<plainSingleton>::is_destroyed()|" (see 
also comment 
https://github.com/boostorg/serialization/pull/110#issuecomment-429561420)

Note how this was independently noticed by others too: 
https://github.com/boostorg/serialization/pull/110#issuecomment-417601919

And again and again he lacks reasoning:
> It's based on a mis-understanding of how the serialization uses the 
> singleton and of how the singleton works.

If it was, it would be easy to point out what mis-understanding this is. 
And it all boils down THAT the singleton is used. So even if the values 
in the singleton are bogus, all that is tested is the construction and 
destruction of the singleton. And the library DOES do that too. #127 
clearly shows this.

> >As I as the author of the patch might
>> understand the patch better, it is quite possible, that I'm right, 
>> isn't it?
>
> It's possible - but after having considered that possibility, I've 
> concluded that it's False

Again: No reasons, just claims. My claims are all backed by references 
to actual executions on CI and findings of other people coming to the 
same conclusion as me.

Maybe it is time to put Boost.Serialization under CMT...