AMDG On 02/15/2016 12:50 AM, Vladimir Prus wrote:

On 13-Feb-16 3:20 PM, Abel Sinkovics wrote:

...

Both the compile and link steps succeed and there is nothing about the execution of the compiled code. Does someone know why these are marked as failures?

Hi Abel,

I'll try to figure why there is failure without no output, but don't yet know how long it will take me.

In the meantime, is there a chance you'd run these tests locally and see what's wrong?

I know what the problem is, but fixing it is non-trivial. process_jam_log randomly picks one compilation, for targets that consist of more than one translation unit. This is a side-effect of handling incremental builds, which discards an existing compile log when it finds a new one. i.e. the last one that it sees will be the only one shown. In Christ, Steven Watanabe

Hi Vladimir, On 2016-02-15 23:04, Abel Sinkovics wrote:

Hi Vladimir,

On 2016-02-15 08:50, Vladimir Prus wrote:

...

I'll try to figure why there is failure without no output, but don't yet know how long it will take me.

In the meantime, is there a chance you'd run these tests locally and see what's wrong? Metaparse has a TravisCI job checking it against gcc 4.6, which builds and runs the tests in the Boost source tree (using b2). Tests build and run fine there (see https://travis-ci.org/boostorg/metaparse/jobs/109227535 for the last result). I can look at setting up a Windows-based gcc 4.6 environment, as the failing regression tests seem to run on Windows. I've built and run the tests on Windows with MinGW gcc 4.6 (and -std=c++0x) and there were no errors.

Note that only the tests that have to be run (not only compiled) fail. The library has only a few of them, as it is a template metaprogramming library. Regards, Ábel