Failing tests on gcc of serialization library

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Failing tests on gcc of serialization library

Robert Ramey
I notice a lot of the serialization library tests are failing.

In a number of cases the test is failing to compile.

A cause might be the command line switch "g++-5  -ftemplate-depth-128"
The Jamfiles for the test specify

<toolset>gcc:<cxxflags>"-ftemplate-depth-255"

But this information doesn't seem to be getting used by the test itself.
  I should note that exactly the same compiler passes on my own machine.

Why are the settings used by the tester different from what I expect?

Robert Ramey


_______________________________________________
Boost-Testing mailing list
[hidden email]
http://lists.boost.org/mailman/listinfo.cgi/boost-testing
Reply | Threaded
Open this post in threaded view
|

Re: Failing tests on gcc of serialization library

Robert Ramey
On 10/20/15 7:31 AM, Robert Ramey wrote:

> I notice a lot of the serialization library tests are failing.
>
> In a number of cases the test is failing to compile.
>
> A cause might be the command line switch "g++-5  -ftemplate-depth-128"
> The Jamfiles for the test specify
>
> <toolset>gcc:<cxxflags>"-ftemplate-depth-255"
>
> But this information doesn't seem to be getting used by the test itself.
>   I should note that exactly the same compiler passes on my own machine.
>
> Why are the settings used by the tester different from what I expect?
>
> Robert Ramey

I quadriple checked my jamfiles and it seems that I had the stack depth
extension only in my test builds but not in the library builds.  I've
now fixed this and checked it in.  We'll see if that fixes this problem.

Robert Ramey


_______________________________________________
Boost-Testing mailing list
[hidden email]
http://lists.boost.org/mailman/listinfo.cgi/boost-testing
Reply | Threaded
Open this post in threaded view
|

Re: Failing tests on gcc of serialization library

Niklas Angare
"Robert Ramey" <[hidden email]> wrote:
> On 10/20/15 7:31 AM, Robert Ramey wrote:
>> The Jamfiles for the test specify
>>
>> <toolset>gcc:<cxxflags>"-ftemplate-depth-255"

Boost.Build has the feature c++-template-depth for specifying that compiler
option so it may be more appropriate to use that. I think the value needs to
be divisible by 10 or 64 so the closest would be <c++-template-depth>256.

It seems strange that Boost.Serialization tests would need a template depth
of 255 when libraries with heavy use of template meta-programming can get by
with 128. Are you sure there's not a bug somewhere?

Regards,

Niklas Angare
 


_______________________________________________
Boost-Testing mailing list
[hidden email]
http://lists.boost.org/mailman/listinfo.cgi/boost-testing