Re: [boost] MinGW test failures

Subject: Re: [boost] MinGW test failures
From: Niall Douglas (ndouglas_at_[hidden])
Date: 2013-08-07 14:30:33

> I can see that almost all MinGW tests (jc-bell.com runner) for trunk are 
> failing, at
> least for Boost.Log [1] and Boost.Atomic [2]. However, test logs don't show 
> any
> errors, only compiler warnings. Does anyone know what is the problem?
>
> Also, this is not the first time when I see no errors in logs while the test 
> is
> marked as failed. This is really unhelpful. Are some logs being lost by the 
> test
> running suite?

Proposed Boost.AFIO is also seeing segfaults on Mingw only 
(https://ci.nedprod.com/job/Boost.AFIO%20Test%20WinXP%20Mingw32/14/console).

I had assumed it's the fault of the Mingw with GCC 4.7, because the Mingw with 
GCC 4.6 works fine.

[I suspect Mingw with GCC 4.7 has been compiled with the frontend set to i686 
but the backend set to i386. This causes substantial problems with any of the 
i486 ops e.g. anything atomic or locking. Mingw does point out the GCC 4.7 
build is beta.].

There is an argument in favour of deprecating Mingw completely in favour of 
Mingw-w64, which is a ground up clean reimplementation.

Niall

---
Opinions expressed here are my own and do not necessarily represent those of 
BlackBerry Inc.