On 2/7/2012 3:15 PM, Davidson, Josh wrote:
I've found that the build configuration for the last couple of releases (1.48 and 1.49 beta 1) have been a bit flaky with MinGW. ....
In 1.48, there were issues performing complete builds using mingw-w64 (Duplicate name of actual target).
I just tried building 1.48 under MinGW-32/gcc-4.4 a couple days ago and got the same ugly "duplicate name" stack-dump, and had to fall back to the minimal build (which did work). The behavior was the same regardless of bjam's origin, too. (I.e., MinGW build or msvc-8.0). I run a batch of MinGW regression tests http://www.boost.org/development/tests/release/developer/summary.html, but those probably don't attempt to do a complete build, do they? I just searched those regressions' .log files and didn't find that error message. I haven't looked into the test-result matrix for an indication of a problem, but I suspect it's not there.