Re: [Boost-testing] Martin Wille test machine out of disk space?

From: Martin Wille (mw8329_at_[hidden])
Date: 2006-11-14 15:00:50

Juergen Hunold wrote:
> Hi Martin !
> 
> On Tuesday 14 November 2006 19:12, Martin Wille wrote:
>> Rene Rivera wrote:
>>> It might be it's filled with old stuff. What is getting put in /tmp
> 
>> Did anything change in the number of tests being executed? Or did
>> anything change in the way the path names get generated? Both would
>> explain why there's suddenly demand for additional space.
> 
> Some tests now explicitly link against the static Boost.Test library on 
> Linux. This might explain the increase in disk space.
> 
> See 
> 
> grep -l "<link>static" libs/*/test/Jamfile.v2 /dev/null
> 
> libs/assign/test/Jamfile.v2
> libs/conversion/test/Jamfile.v2
[...]

> For those libraries, you would have had the shared _and_ static linked 
> version lying around when runnig incremental tests...

This is a good idea, but:

I looked for duplicates for assign and conversion and didn't find any. 
It seems the dynamic versions were overwritten by static ones.

Looking closer, I found odd things for Boost.Conversion:

The Jamfile reads:

test-suite conversion
   : [ run implicit_cast.cpp ]
     [ compile-fail implicit_cast_fail.cpp ]
     [ run ../cast_test.cpp ]
     [ run ../numeric_cast_test.cpp ]
     [ run ../lexical_cast_test.cpp 
../../test/build//boost_unit_test_framework/<link>static ]
   ;

However:

ldd cast_test.test/gcc-4.1.0_linux/debug/cast_test shows that no dynamic 
libraries are used except for standard libraries.

ldd lexical_cast_test.test/gcc-4.1.0_linux/debug/lexical_cast_test lists
libboost_unit_test_framework-gcc41-d-1_34.so.1.34.0.

(Without knowing Boost.Build well,) it seems to me that the logic is 
inverted.

Regards,
m
Send instant messages to your online friends http://au.messenger.yahoo.com