Re: [certi-dev] Solaris8/GCC-3.4.2; Was: Solaris10/SunStudio

[Eric Noulard]

2009/5/27 Martin Spott <address@hidden>:
> Eric Noulard wrote:
>
>> Is it really impossible to find a 64 bits long integer type on 32bits
>> machine using gcc 3.4.x ??
>
> If I bypass the CMake test and simply add '-DHAVE_INT64' to the
> compiler call, then the compiler doesn't complain and creates a nice
> object file - I'm even getting a nice, shared 'libHLA'.

> BUT, simply
> adding '-DHAVE_INT64' to every call (by adding it to 'CMakeLists.txt')
> doesn't lead much further:

How did you add -DHAVE_INT64 in the first place if it was not within
CMakeLists.txt ?

I think the trouble may comes from the fact CERTI is compiled with
-pedantic and -ansi which may forbids long-long usage unless one specify
-Wno-long-long.

However some version of gcc did not overide -ansi rules with -Wno-long-long
would you try to remove -ansi -pedantic from the main CMakeLists.txt?

> Would you like me to send you a copy of the involved int* and sys/int*
> header files ?

Yes why not,
I may find explanation with conditionnal definition of such types
in those headers.

I think I may have found a similar issue there:
http://gcc.gnu.org/onlinedocs/libstdc++/faq.html#faq.solaris_long_long

If this is the case this is a gcc+solaris specific issue regarding long long....

-- 
Erk