Boost logo

Boost :

Subject: Re: [boost] Boost Mingw w64 compilation issues and fixes
From: Andrew Ho (helloworld922_at_[hidden])
Date: 2013-07-15 05:43:18

> I trust you have read the Boost.Math section
> "If and How to Build a Boost.Math Library, and its Examples and Tests"
> It's most unusual to want/need to build a library - Boost.Math is most
efficiently used as
> header-only.

I'm trying to run an "as basic" build as possible, which includes building
boost math as a library. There are a few other libraries which similarly
have "optional" build libraries, I don't know how many of them by default
are built (I believe MPI is the main library which isn't built by default).

> However the tests should run, and the test suite at
> and specifically the test using 4.8
> MinGW-w64-4.8
> Boost Regression Test, Windows 7 x64, MinGW-w64, gcc 4.8
> are showing mostly green.

mostly != all :P I haven't tried running the tests without the -j flag yet
(stop on first error), so i don't know how many tests failed, and it's
possible that multiple compile errors/crashes could be present in a single
test. I don't have some "automatic terminate compiler on crash" script, so
every compile error means I have to click the mouse (which is more than I
want to).

The test log suggests there's some script for running tests with an error
timeout, I'll try to see if I can get that working so I can get the full

> However it is showing a few unexpected failures (compared to previous gcc
versions), for example:
> Test output: MinGW-w64-4.8 - math - test_remez / gcc-mingw-
> Rev 84971 / Sun, 07 Jul 2013 14:27:48 +0000
> Compile [2013-07-08 00:15:42 UTC]: fail
> "g++" -ftemplate-depth-128 -O0 -fno-inline -Wall -g -Winvalid-pch -
> -I"F:\boost\GCC\trunk\results\boost\bin.v2\libs\math\test\gcc-mingw-
4.8.0\debug\link-static" -I".."
> -I"..\libs\math\test" -c -o
> tatic\test_remez.o" "..\libs\math\test\test_remez.cpp"
> which is another similarly unhelpful message (as are the others).
> So, to me, this looks like a compiler problem.
> (And your StackOverflow link above also suggests the same).
> Please report if you make any sense of/fix this.

I concur, it does appear to be a compiler problem.

Boost list run by bdawes at, gregod at, cpdaniel at, john at