Re: [boost] Solving Robert's singleton problem

Subject: Re: [boost] Solving Robert's singleton problem
From: Alexander Grund (alexander.grund_at_[hidden])
Date: 2018-11-01 07:55:28

> I merged your latest version of your "test" and it failed no where.

Can you point me to the merge or commit and the passing travis build? 
All I have seen is some of my PR manually put into other commits 
65f05dc2099e376f5306cc0c16c29ec0a1a1d2d4 and 
237ff81ac62cb9cf8206a050647a1b47601ce4bd and due to not using the full 
MR it fails to even link on linux: 
https://travis-ci.org/boostorg/serialization/jobs/446809035 which is 
probably why you removed it in c5549b2d6ba79a60a98ddceb854954c090c70cb1

Why don't you just merge the open PR which does work as intended and 
DOES fail on travis?

> I also managed to reproduce the remaining test failure - 
> test_dll_exported - on my own machine and have traced it to an related 
> issue which has now been fixed.  I've run all tests on clang and 
> several gcc compilers on my own system.  I'm waiting for final results 
> on the test matrix and appveyor for windows.  These were failing but I 
> need to upload a jamfile tweak.
Good to hear another issue is resolved.
> Sooo - I believe that that there are no pending issues regarding 
> singletons, visibility and export.

Unfortunately not true.

1) The crash-producing test is not or wrongly included
2) The is_destroyed test is not merged and the bug not fixed

Again: Please simply merge my PRs (at least the "simple ones" with the 
tests) or tell me, what you don't understand or like about them so I can 
change them without breaking them. We are running in circles here where 
you do not seem to understand my intentions and hence the bugs, change 
things "based on my changes" but without using the important parts. 
Others have already confirmed the validity of my tests and that my fix 
works, so it can't be bad.