List Info

Thread: Should parrellel builds work?
Should parrellel builds work?
country flaguser name
Canada		 2007-10-22 02:15:21 
Folks,

I seem to remember that someone (Dave A.?) fixed bjam/bbv2
so that parallel 
builds of the regression runs should work OK?

I'm asking because many of the IBM xlc result failures have
no error message 
associated with them.  Looking at the error log files posted
to 
ftp://fx.meta-comm.com/boost-regression/trunk/logs I see
things like:

"../libs/serialization/test/test_non_intrusive.cpp"
;, line 68.36: 1540-0130 
(S) "std::numeric_limits" is not declared.
"../libs/serialization/test/test_non_intrusive.cpp"
;, line 83.34: 1540-0130 
(S) "std::numeric_limits" is not declared.
MkDir1 
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_default_ctor_text_archive.test/vacp
p

    mkdir 
"/home/ccambly/boost_regression/results/boost/bin.v2/li
bs/serialization/test/test_non_default_ctor_text_archive.tes
t/vacpp"

MkDir1 
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_intrusive_xml_warchive_dll.test/vac
pp/debug

    mkdir 
"/home/ccambly/boost_regression/results/boost/bin.v2/li
bs/serialization/test/test_non_intrusive_xml_warchive_dll.te
st/vacpp/debug"

vacpp.compile.c++ 
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_intrusive_text_archive_dll.test/vac
pp/debug/test_non_intrusive.o

    xlC -c   -DBOOST_ALL_DYN_LINK=1 -DBOOST_ALL_NO_LIB=1
-DBOOST_ARCHIVE_TEST=text_archive.hpp 
 -DBOOST_LIB_DIAGNOSTIC=1 -qcpluscmt -qNOOPTimize -qnoinline
-g -qfullpath -qfuncsect 
 -qeh -qrtti -I".."  -o 
"/home/ccambly/boost_regression/results/boost/bin.v2/li
bs/serialization/test/test_non_intrusive_text_archive_dll.te
st/vacpp/debug/test_non_intrusive.o" 
"../libs/serialization/test/test_non_intrusive.cpp"
;

MkDir1 
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_intrusive_xml_warchive.test/vacpp/d
ebug

    mkdir 
"/home/ccambly/boost_regression/results/boost/bin.v2/li
bs/serialization/test/test_non_intrusive_xml_warchive.test/v
acpp/debug"

MkDir1 
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_default_ctor_text_archive_dll.test

    mkdir 
"/home/ccambly/boost_regression/results/boost/bin.v2/li
bs/serialization/test/test_non_default_ctor_text_archive_dll
.test"

...failed vacpp.compile.c++ 
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_intrusive_binary_archive.test/vacpp
/debug/test_non_intrusive.o...

So it looks like the error messages are getting disconnected
from the 
failure message?

The tests are run with -j16 I believe, so is this to be
expected, or should 
this work now?  Any special requirements on building bjam or
on the C std 
lib for this to work correctly?

Many thanks, John. 

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build
Re: Should parrellel builds work?
country flaguser name
United States		 2007-10-22 10:18:07 
John Maddock wrote:
> I seem to remember that someone (Dave A.?) fixed
bjam/bbv2 so that parallel 
> builds of the regression runs should work OK?

Don't know if Dave said it... But I certainly have.

> I'm asking because many of the IBM xlc result failures
have no error message 
> associated with them.  Looking at the error log files
posted to 
> ftp://fx.meta-comm.com/boost-regression/trunk/logs I
see things like:
> 
[...]
> vacpp.compile.c++ 
>
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_intrusive_text_archive_dll.test/vac
pp/debug/test_non_intrusive.o
> 
>     xlC -c   -DBOOST_ALL_DYN_LINK=1
-DBOOST_ALL_NO_LIB=1 -DBOOST_ARCHIVE_TEST=text_archive.hpp 
>  -DBOOST_LIB_DIAGNOSTIC=1 -qcpluscmt -qNOOPTimize
-qnoinline -g -qfullpath -qfuncsect 
>  -qeh -qrtti -I".."  -o 
>
"/home/ccambly/boost_regression/results/boost/bin.v2/li
bs/serialization/test/test_non_intrusive_text_archive_dll.te
st/vacpp/debug/test_non_intrusive.o" 
>
"../libs/serialization/test/test_non_intrusive.cpp"
;
> 
> MkDir1 
>
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_intrusive_xml_warchive.test/vacpp/d
ebug
> 
>     mkdir 
>
"/home/ccambly/boost_regression/results/boost/bin.v2/li
bs/serialization/test/test_non_intrusive_xml_warchive.test/v
acpp/debug"
> 
> MkDir1 
>
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_default_ctor_text_archive_dll.test
> 
>     mkdir 
>
"/home/ccambly/boost_regression/results/boost/bin.v2/li
bs/serialization/test/test_non_default_ctor_text_archive_dll
.test"
> 
> ...failed vacpp.compile.c++ 
>
/home/ccambly/boost_regression/results/boost/bin.v2/libs/ser
ialization/test/test_non_intrusive_binary_archive.test/vacpp
/debug/test_non_intrusive.o...
> 
> So it looks like the error messages are getting
disconnected from the 
> failure message?
> 
> The tests are run with -j16 I believe, so is this to be
expected, or should 
> this work now?

It should work... But it does depend on which version of
bjam is used.

> Any special requirements on building bjam or on the C
std 
> lib for this to work correctly?

AFAIK, only POSIX stuff.


-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software
.com
-- rrivera/acm.org - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim - grafikrobot/yahoo
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build
Re: Should parrellel builds work?
country flaguser name
United States		 2007-10-22 20:38:46 
John Maddock wrote:
> Folks,
> 
> I seem to remember that someone (Dave A.?) fixed
bjam/bbv2 so that parallel 
> builds of the regression runs should work OK?

I have the same recollection, but I tried -j2 on an Intel
dual core 
machine a month or two ago with similar problems.  Bjam
itself ran OK 
but the log file output was all garbled so it was useless
for regression 
testing:-(

--Beman
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build
Re: Should parrellel builds work?
country flaguser name
United States		 2007-10-22 20:50:04 
Beman Dawes wrote:
> John Maddock wrote:
>> Folks,
>>
>> I seem to remember that someone (Dave A.?) fixed
bjam/bbv2 so that parallel 
>> builds of the regression runs should work OK?
> 
> I have the same recollection, but I tried -j2 on an
Intel dual core 
> machine a month or two ago with similar problems.  Bjam
itself ran OK 
> but the log file output was all garbled so it was
useless for regression 
> testing:-(

Well it's been implemented on Windows long ago now. But if
you are 
unsure there's a test for this feature:

   cd tools/jam/src
   build.bat (or build.sh)
   cd ../test
   test.bat (or test.sh)

And I use it with -j2 on my regular test run, RSI Droid,
without problems.


-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software
.com
-- rrivera/acm.org - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim - grafikrobot/yahoo
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build
[1-4]

about | contact  Other archives ( Real Estate discussion Medical topics )