qcc toolset

Hi Jim,

> Having spent some time trying to get your version of
qcc.jam working I
> have taken the executive decision to revert to my
original plan and
> produce qcc.jam as a sibling of gcc.jam and, not as a
child if gcc.jam.
>
> Although qcc is based on gcc the command sets are *so*
different and I
> found myself overloading almost every statement. In
addition I would
> have needed to modify some of the platform conditional
logic in gcc.jam
> which I did not want to do. I trust you will not be
offended at this
> decision.

No problem. If you end up overriding everything single
option from gcc.jam, 
then there's no point in inheriting from gcc.

> The file is attached and it runs the hello &
libraries test but there
> are errors with test_all.py.

If you send me the output of test_all.py, I'll take a look.

> I have also started a run of the regression tests but I
stopped when I
> saw that there is a problem with qcc.link. I need to
add 'libm' to each
> of the linking rules because it has been omitted from
the default search
> list on this platform.  I have tried adding this as
FINDLIBS and
> FINDLIBS-SA but it does not always show up in the right
places.

When I run "bjam -n qcc" on the hello example, I
get this:

    qcc.link bin/qcc/debug/main

      QCC  -o "bin/qcc/debug/main"
"bin/qcc/debug/main.o"
"bin/qcc/debug/a.o"  
      -lm -gstabs+

That is, the "-lm" is there. Is that right? Do
you get "-lm" when building 
"hello"? And "libraries"?

Or "-lm" is incorrect and something else should
be placed? Or you get problem 
only when running regressions?

In fact, if -lm should always be added, you can hardcode it
in the qcc.link 
and qcc.link.dll actions.

- Volodya
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build

Vladimir Prus wrote:

Just to complete the picture my user-config.jam contains:

using qcc : 3.3.5_cpp : :
<compileflags>-V3.3.5,gcc_ntox86_cpp 
<linkflags>-V3.3.5,gcc_ntox86_cpp ;

using qcc : 3.3.5_gpp : :
<compileflags>-V3.3.5,gcc_ntox86_gpp 
<linkflags>-V3.3.5,gcc_ntox86_gpp ;

> 
> If you send me the output of test_all.py, I'll take a
look.
> 
Attached as test_all_*.txt.bz2. I did a run with each
toolset in case 
there are differences.

>>I have also started a run of the regression tests
but I stopped when I
>>saw that there is a problem with qcc.link. I need to
add 'libm' to each
>>of the linking rules because it has been omitted
from the default search
>>list on this platform.  I have tried adding this as
FINDLIBS and
>>FINDLIBS-SA but it does not always show up in the
right places.
> 
> When I run "bjam -n qcc" on the hello
example, I get this:
> 
>     qcc.link bin/qcc/debug/main
> 
>       QCC  -o "bin/qcc/debug/main"
"bin/qcc/debug/main.o"
"bin/qcc/debug/a.o"  
>       -lm -gstabs+
> 
> That is, the "-lm" is there. Is that right?
Do you get "-lm" when building 
> "hello"? And "libraries"?

In this context it is correct.

When I run the regression test from regression.py I get the
following 
failure of "math_info" with the commands:

qcc.link 
/hd1/boost-HEAD/results/boost/bin.v2/status/math_info.test/q
cc-3.3.5_cpp/debug/math_info

     "QCC"    -o 
"/hd1/boost-HEAD/results/boost/bin.v2/status/math_info
.test/qcc-3.3.5_cpp/debug/math_info" 
"/hd1/boost-HEAD/results/boost/bin.v2/status/math_info
.test/qcc-3.3.5_cpp/debug/math_info.o" 
    -gstabs+ -V3.3.5,gcc_ntox86_cpp


But then if I cd to lib/config/test and run "bjam --v2
qcc-3.3.5_cpp 
math_info" I get a pass with these commands:

qcc.link 
../../../bin.v2/libs/config/test/math_info.test/qcc-3.3.5_cp
p/debug/threading-multi/math_info

     "QCC"    -o 
"../../../bin.v2/libs/config/test/math_info.test/qcc-3
.3.5_cpp/debug/threading-multi/math_info" 
"../../../bin.v2/libs/config/test/math_info.test/qcc-3
.3.5_cpp/debug/threading-multi/math_info.o" 
   -lm -gstabs+ -V3.3.5,gcc_ntox86_cpp


Ah! I can see now that it has something to do with
threading, but then 
all qcc stuff should default to multithreading and there is
no switch as 
with gcc. AFAIK qcc always inserts the
"-pthread" option automatically 
for gcc.

What's going on here?

Jim
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build