You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@santuario.apache.org by Berin Lautenbach <be...@wingsofhermes.org> on 2006/10/15 12:46:43 UTC
C++ 1.3.1-RC1
Hiya,
For those interested in the C++ library there is a release candidate for
1.3.1 in http://people.apache.org/~blautenb
Includes
- NIX build now based on Automake (aimed at making it more robust)
- Initial support for the Xerces 3.0 codebase (not official, as 3.0 is
not yet released)
- Bug fixes
Please try it out and let me know how it builds and runs.
Cheers,
Berin
Re: C++ 1.3.1-RC1
Posted by Berin Lautenbach <be...@wingsofhermes.org>.
Hi all,
I will be doing a 1.3.1 release this weekend - so let me know if you
find any problems with the rc-1 build!
Cheers,
Berin
Berin Lautenbach wrote:
> Hiya,
>
> For those interested in the C++ library there is a release candidate for
> 1.3.1 in http://people.apache.org/~blautenb
>
> Includes
>
> - NIX build now based on Automake (aimed at making it more robust)
> - Initial support for the Xerces 3.0 codebase (not official, as 3.0 is
> not yet released)
> - Bug fixes
>
> Please try it out and let me know how it builds and runs.
>
> Cheers,
> Berin
>
>
>
RE: C++ 1.3.1-RC1
Posted by Scott Cantor <ca...@osu.edu>.
> In fact the problem with the -mt flag was the acx_pthreads macro (used
> as of 1.3.1 and taken from the autoconf-macros project).
I'm using a variant of it myself.
> It's a really
> small bug with ordering of tests of flags on the Solaris platform.
> Because the Sun workshop compiler simply passes flags it doesn't
> understand to ld, -pthread gets accepted as the correct flag, even
> though it isn't (generates warnings). Simply reordering and putting the
> -mt test first works on my systems, but it would be good if you could
> check on yours!
I'll check with rc2 and report back, hopefully by tomorrow.
-- Scott
Re: C++ 1.3.1-RC1
Posted by Berin Lautenbach <be...@wingsofhermes.org>.
Scott,
I've done what you suggested with PLATFORM_OPTIONS, but I've removed PIC
entirely (libtool takes care of that).
In fact the problem with the -mt flag was the acx_pthreads macro (used
as of 1.3.1 and taken from the autoconf-macros project). It's a really
small bug with ordering of tests of flags on the Solaris platform.
Because the Sun workshop compiler simply passes flags it doesn't
understand to ld, -pthread gets accepted as the correct flag, even
though it isn't (generates warnings). Simply reordering and putting the
-mt test first works on my systems, but it would be good if you could
check on yours!
Cheers,
Berin
Scott Cantor wrote:
> The gist of the Solaris problem I'm finding is the use of this approach to
> setting build flags:
>
> if test "x${ac_cv_cxx_compiler_gnu}" != "xyes"; then
> # Assume SunCC
> PIC=["-KPIC"]
> PLATFORM_OPTIONS=["-mt"]
> # Flags for warnings as errors
> cxxflags_warnerror="-xwe -errtags"
> # Flags for Optimise
> cxxflags_optimise="-xO2"
> # Flags for debug
> cxxflags_debug="-g"
> fi
>
> Later on, this:
>
> CC1=["${CXX} ${CXXFLAGS} ${PLATFORM_OPTIONS}"]
>
> I'm not clear on what CC1 is, but it's not used by the make process later at
> all. It's in Makefile, but is never referenced. When libtool runs the C++
> compiler, it just runs it with whatever is in CPPFLAGS and CXXFLAGS, but not
> PLATFORM_OPTIONS. So the SunCC flags are omitted from the build, in this
> case no -mt for thread safety.
>
> My suggestion is that you should append the PLATFORM_OPTIONS bits to CFLAGS
> and CXXFLAGS directly. I think that will result in the right behavior. If
> there's a more "correct" way, that's fine too, but I've never seen one
> discussed.
>
> -- Scott
>
>
>
Re: C++ 1.3.1-RC1
Posted by Berin Lautenbach <be...@wingsofhermes.org>.
Scott,
Thanks for this - I will have a look-see on the weekend. I had actually
tagged 1.3.1 and was all ready to go, so this just came in time! :>.
Cheers,
Berin
Scott Cantor wrote:
> The gist of the Solaris problem I'm finding is the use of this approach to
> setting build flags:
>
> if test "x${ac_cv_cxx_compiler_gnu}" != "xyes"; then
> # Assume SunCC
> PIC=["-KPIC"]
> PLATFORM_OPTIONS=["-mt"]
> # Flags for warnings as errors
> cxxflags_warnerror="-xwe -errtags"
> # Flags for Optimise
> cxxflags_optimise="-xO2"
> # Flags for debug
> cxxflags_debug="-g"
> fi
>
> Later on, this:
>
> CC1=["${CXX} ${CXXFLAGS} ${PLATFORM_OPTIONS}"]
>
> I'm not clear on what CC1 is, but it's not used by the make process later at
> all. It's in Makefile, but is never referenced. When libtool runs the C++
> compiler, it just runs it with whatever is in CPPFLAGS and CXXFLAGS, but not
> PLATFORM_OPTIONS. So the SunCC flags are omitted from the build, in this
> case no -mt for thread safety.
>
> My suggestion is that you should append the PLATFORM_OPTIONS bits to CFLAGS
> and CXXFLAGS directly. I think that will result in the right behavior. If
> there's a more "correct" way, that's fine too, but I've never seen one
> discussed.
>
> -- Scott
>
>
>
RE: C++ 1.3.1-RC1
Posted by Scott Cantor <ca...@osu.edu>.
The gist of the Solaris problem I'm finding is the use of this approach to
setting build flags:
if test "x${ac_cv_cxx_compiler_gnu}" != "xyes"; then
# Assume SunCC
PIC=["-KPIC"]
PLATFORM_OPTIONS=["-mt"]
# Flags for warnings as errors
cxxflags_warnerror="-xwe -errtags"
# Flags for Optimise
cxxflags_optimise="-xO2"
# Flags for debug
cxxflags_debug="-g"
fi
Later on, this:
CC1=["${CXX} ${CXXFLAGS} ${PLATFORM_OPTIONS}"]
I'm not clear on what CC1 is, but it's not used by the make process later at
all. It's in Makefile, but is never referenced. When libtool runs the C++
compiler, it just runs it with whatever is in CPPFLAGS and CXXFLAGS, but not
PLATFORM_OPTIONS. So the SunCC flags are omitted from the build, in this
case no -mt for thread safety.
My suggestion is that you should append the PLATFORM_OPTIONS bits to CFLAGS
and CXXFLAGS directly. I think that will result in the right behavior. If
there's a more "correct" way, that's fine too, but I've never seen one
discussed.
-- Scott
RE: C++ 1.3.1-RC1
Posted by Scott Cantor <ca...@osu.edu>.
Finally doing some testing of this, and I see at least a couple of compiler
flag issues on Solaris when not using gcc. I'll look further and submit a
full description and/or fix, and test on the Mac.
-- Scott