I'm trying to compile CMake using a non-default GCC installed in /usr/local/gcc530, on Solaris 2.11.
I have LD_LIBRARY_PATH=/usr/local/gcc530/lib/sparcv9
Bootstrap proceeds fine, bootstrapped cmake successfully compiles various object files, but when it tries to link the real cmake (and other executables), I get pages of "undefined reference" errors to various standard library functions, because, as running the link command manually with -Wl,-verbose shows, the linker links with /usr/lib/64/libstdc++.so of the system default, much older GCC.
This is because apparently CMake tries to find curses/ncurses libraries (even if I tell it BUILD_CursesDialog:BOOL=OFF), finds them in /usr/lib/64, and adds -L/usr/lib/64 to build/Source/CMakeFiles/cmake.dir/link.txt, which causes the linker to use libstdc++.so from there, and not my actual GCC's own.
I found a workaround: I can get the path to proper libraries from $CC -m64 -print-file-name=libstdc++.so then put it with -L into LDFLAGS when running ./configure, and all works well then.
Is there a less hacky way? It's really weird that I can't tell GCC to prioritize its own libraries.
Also, is there some way to have CMake explain where different parts of a resulting command line came from?
Related
I'm a student doing research involving extending the TM capabilities of gcc. My goal is to make changes to gcc source, build gcc from the modified source, and, use the new executable the same way I'd use my distro's vanilla gcc.
I built and installed gcc in a different location (not /usr/bin/gcc), specifically because the modified gcc will be unstable, and because our project goal is to compare transactional programs compiled with the two different versions.
Our changes to gcc source impact both /gcc and /libitm. This means we are making a change to libitm.so, one of the shared libraries that get built.
My expectation:
when compiling myprogram.cpp with /usr/bin/g++, the version of libitm.so that will get linked should be the one that came with my distro;
when compiling it with ~/project/install-dir/bin/g++, the version of libitm.so that will get linked should be the one that just got built when I built my modified gcc.
But in reality it seems both native gcc and mine are using the same libitm, /usr/lib/x86_64-linux-gnu/libitm.so.1.
I only have a rough grasp of gcc internals as they apply to our project, but this is my understanding:
Our changes tell one compiler pass to conditionally insert our own "function builtin" instead of one it would normally use, and this is / becomes a "symbol" which needs to link to libitm.
When I use the new gcc to compile my program, that pass detects those conditions and successfully inserts the symbol, but then at runtime my program gives a "relocation error" indicating the symbol is not defined in the file it is searching in: ./test: relocation error: ./test: symbol _ITM_S1RU4, version LIBITM_1.0 not defined in file libitm.so.1 with link time reference
readelf shows me that /usr/lib/x86_64-linux-gnu/libitm.so.1 does not contain our new symbols while ~/project/install-dir/lib64/libitm.so.1 does; if I re-run my program after simply copying the latter libitm over the former (backing it up first, of course), it does not produce the relocation error anymore. But naturally this is not a permanent solution.
So I want the gcc I built to use the shared libs that were built along with it when linking. And I don't want to have to tell it where they are every time - my feeling is that it should know where to look for them since I deliberately built it somewhere else to behave differently.
This sounds like the kind of problem any amateur gcc developer would have when trying to make a dev environment and still be able to use both versions of gcc, but I had difficulty finding similar questions. I am thinking this is a matter of lacking certain config options when I configure gcc before building it. What is the right configuration to do this?
My small understanding of the instructions for building and installing gcc led me to do the following:
cd ~/project/
mkdir objdir
cd objdir
../source-dir/configure --enable-languages=c,c++ --prefix=/home/myusername/project/install-dir
make -j2
make install
I only have those config options because they seemed like the ones closest related to "only building the parts I need" and "not overwriting native gcc", but I could be wrong. After the initial config step I just re-run make -j2 and make install every time I change the code. All these steps do complete without errors, and they produce the ~/project/install-dir/bin/ folder, containing the gcc and g++ which behave as described.
I use ~/project/install-dir/bin/g++ -fgnu-tm -o myprogram myprogram.cpp to compile a transactional program, possibly with other options for programs with threads.
(I am using Xubuntu 16.04.3 (64 bit), within VirtualBox on Windows. The installed /usr/bin/gcc is version 5.4.0. Our source at ~/project/source-dir/ is a modified version of 5.3.0.)
You’re running into build- versus run-time linking differences. When you build with -fgnu-tm, the compiler knows where the library it needs is found, and it tells the linker where to find it; you can see this by adding -v to your g++ command. However when you run the resulting program, the dynamic linker doesn’t know it should look somewhere special for the ITM library, so it uses the default library in /usr/lib/x86_64-linux-gnu.
Things get even more confusing with ITM on Ubuntu because the library is installed system-wide, but the link script is installed in a GCC-private directory. This doesn’t happen with the default GCC build, so your own GCC build doesn’t do this, and you’ll see libitm.so in ~/project/install-dir/lib64.
To fix this at run-time, you need to tell the dynamic linker where to find the right library. You can do this either by setting LD_LIBRARY_PATH (to /home/.../project/install-dir/lib64), or by storing the path in the binary using -Wl,-rpath=/home/.../project/install-dir/lib64 when you build it.
[Shamelessly cross-posted from the CMake help list]
I'm trying to create binaries as statically as possible. The fortran code I've got has got X11 and quadmath as dependencies, and I've come across a number of issues (maybe each of these issues should be in a different question?):
My variables are currently
set(CMAKE_LIBRARY_PATH /usr/X11/lib /usr/X11/include/X11 ${CMAKE_LIBRARY_PATH})
find_package(X11 REQUIRED)
find_library(X11 NAMES X11.a PATHS /usr/X11/include/X11/ /usr/X11/lib)
find_library(X11_Xaw_LIB NAMES Xaw Xaw /usr/X11/include/X11/ /usr/X11/lib ${X11_LIB_SEARCH_PATH})
find_library(Xaw Xaw7 PATHS ${X11_LIB_SEARCH_PATH})
set(CMAKE_LIBRARY_PATH /usr/lib/gcc/x86_64-linux-gnu/4.7 /usr/lib/gcc/x86_64-linux-gnu/4.7/x32 /usr/lib/gcc/x86_64-linux-gnu/4.7/32 ${CMAKE_LIBRARY_PATH})
find_library(quadmath NAMES quadmath.a)
set(BUILD_SHARED_LIBS ON)
set(CMAKE_FIND_LIBRARY_SUFFIXES .a ${CMAKE_FIND_LIBRARY_SUFFIXES})
set(LINK_SEARCH_START_STATIC TRUE)
set(LINK_SEARCH_END_STATIC TRUE)
set(SHARED_LIBS OFF)
set(STATIC_LIBS ON)
set(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE)
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static")
Using these, CMake attempts to build every program statically (as expected) - however, it fails because I don't have Xaw.a - I can't find out whether this actually should exist. I have installed the latest libxaw7-dev which I was expecting to fix it. One option would be to compile the X11 libraries myself, but I don't really want to do that...
if I comment out only set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static"), then CMake compiles everything, but uses shared libraries for every program, even though I specify the location of .a X11 libraries in my find_library() calls. I was expecting CMake to use the .a files where it could and then only use shared libraries - is there a way to force this behaviour?
does anyone know yet of a fix for the bug described here: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46539; whereby gfortran seemingly can't statically link libquadmath? I tried the fix using gcc but I can't get CMake to recognise the libgfortran flag:
cmake -DCMAKE_Fortran_COMPILER=gcc -DCMAKE_Fortran_FLAGS=-gfortran
results in
-- The Fortran compiler identification is unknown
-- Check for working Fortran compiler: /usr/bin/gcc
-- Check for working Fortran compiler: /usr/bin/gcc -- broken
CMake Error at /usr/share/cmake-2.8/Modules/CMakeTestFortranCompiler.cmake:54 (message):
The Fortran compiler "/usr/bin/gcc" is not able to compile a simple test program.
However, as you might have noticed, I set the location of the libquadmath.a; when I build a program which doesn't use X11 but does use quadmath when I use
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static")
then the program does compile successfully (running ldd reports 'not a dynamic executable') - does this mean that the bug has been fixed, or does it only work because I set the location in CMake?
I was having a similar problem. Turns out that cmake was implicitly linking against libgfortran and libquadmath. To fix this I put the following in my top level CMakeLists.txt:
unset(CMAKE_Fortran_IMPLICIT_LINK_LIBRARIES)
I could then explicitly link again the libraries using:
SET_TARGET_PROPERTIES(main_f PROPERTIES LINKER_LANGUAGE "C"
LINK_FLAGS
"/usr/local/Cellar/gcc/7.1.0/lib/gcc/7/libgfortran.a
/usr/local/Cellar/gcc/7.1.0/lib/gcc/7/libquadmath.a -lm -lgcc"
)
The static version of libgfortran is necessary because the shared library also depends on libquadmath. The added "-lm" and "-lgcc" bring in the system dynamic versions of these libraries. On a mac system, you would want to use the full path to your libm.a as well.
I guess your questions are not that much related, I don't know the answer for all of them.
For your static linking problems, since you're using GCC, you can pass multiple -static and -dynamic flags to it:
set(CMAKE_EXE_LINKER_FLAGS "-static ${STATIC_LIBS} -dynamic ${EVERYTHING ELSE} -static ${MORE_STATIC_LIBS}")
I don't know why Xaw.a isn't available on your system, probably because the package maintainer of your Linux distribution didn't really make them available.
Also, compiling everything static might make things not compatible between all distros out there and you cripple the ability for others to use improved, up-to-date libraries with your program, it might not be what you want.
If you intend to make a self-contained package of your program, it might be better just to include the shared libraries you used together, like Dropbox and many other proprietary applications do (Humble Bundle games are other example).
I have Lion 10.7.3 with the Command-line tool installed. I wanted to experiment with C++11, so I used homebrew to install GCC 4.7 as documented here.
How can I now upgrade the /usr/bin/g++ to be the one installed by Homebrew? Is it as simple as symlinking it? I just want to double check and make sure. Thanks!
First, are you sure you need g++ 4.7? As you can see from the C++11 implementation status page, recent versions of clang support most of C++11 too. Of course there are still things that g++ handles and clang doesn't, but there are also still things that clang supports and g++ doesn't. And, more importantly, you already have a recent version of clang, from Apple, configured and ready to go, as your default compiler. Plus, g++ after 4.2 doesn't support Mac extensions like, say, -arch, which means you can't use it to build a whole lot of third-party software (because most configure scripts assume that if you're on a Mac, your compiler supports Mac extensions).
But if you want g++ 4.7, you can do it. Just not by trying to replace /usr/bin/g++ with a different version. Never replace anything in /usr/bin (or /System) with non-Apple stuff except in a few very rare cases (when you have a strong reassurance from someone who knows what they're talking about).
A better thing to do is to just install another compiler in parallel. Just let Homebrew install its favorite way (so it installs into some prefix like /usr/local/Cellar/gcc/4.7, then symlinks all the appropriate stuff into /usr/local/bin, etc.), and use it that way.
When compiling your code, instead of writing g++, write /usr/local/bin/g++, or g++-4.7.
If you get tired of doing that, put /usr/local/bin higher on your PATH that /usr/bin, or create a shell alias, or stick it in the environment variable CXX and write $CXX instead of g++.
If you're using a GUI IDE, you should be able to configure it to use your compiler by setting the path to it somewhere. (Unless you're using Xcode, which you can only configure to work with Apple-tested compilers.)
This is all you need for experimenting with your own code. If you want to compile third-party applications with this compiler, that may be a bit more complicated. You don't often actually compile each source file and link the result together; you just do configure && make and let them do the heavy lifting for you.
Fortunately, most packages will respect the standard environment variables, especially CXX for specifying a default C++ compiler and CC for a default C compiler. (That's why I suggested the name CXX above.)
Just remember that, again, g++ 4.7 doesn't support Mac extensions, so if you're not prepared to debug a bunch of autoconf-based configure scripts complaining that your compiler can't generate code because it assumed it could throw -arch x86_64 at any compiler on a Mac, etc., don't do this.
I've installed GCC 4.6.3 into a non-system path on a Mac system and it works fine. However, GCC wants to use code from libgcc for all the binaries I compile, and running otool -L shows that these compiled programs look for libgcc_s.1.dylib in GCC's install path. I can override this by passing -static-libgcc, which just compiles the stuff needed into the binary and that's fine. The problem is this only seems to work with executables, not shared libraries. If I use GCC to compile some third-party lib I want to use in one of my programs as a .dylib, these libraries still look for libgcc_s.1.dylib in the local GCC install path even if I specify -static-libgcc! Needless to say, this is a problem as there's no guarantee that those libraries will find libgcc when run on some other system.
I tried this with ffmpeg. If I look at config.log, the -static-libgcc is most certainly being used. GCC is just not linking libgcc statically with the resulting dylibs. I even tried the -nostdlib, -nostartfiles and -nodefaultlibs options but they were ignored. Again, I checked config.log and they're definitely there!
I believe this is to do with throwing exceptions across the shared library boundary. This page says:
There are several situations in which an application should use the
shared libgcc instead of the static version. The most common of these
is when the application wishes to throw and catch exceptions across
different shared libraries. In that case, each of the libraries as
well as the application itself should use the shared libgcc.
Therefore, the G++ and GCJ drivers automatically add -shared-libgcc
whenever you build a shared library or a main executable, because C++
and Java programs typically use exceptions, so this is the right thing
to do.
The rest of that sections gives a possible workaround (it appears) and that is to use the GCC driver to link your shared library, however if the statically-linked library throws exceptions you'll probably get a Segmentation Violation.
I'm extremely new to using Makefiles and autoconf. I'm using the Camellia image library and trying to statically link my code against their libraries. When I run "make" on the Camellia image library, I get libCamellia.a, .so, .la, and .so.0.0.0 files inside my /usr/local/lib directory. This is the command I use to compile my code with their libraries:
gcc -L/usr/local/lib -lCamellia -o myprogram myprogram.c
This works fine, but when I try to statically link, this is what I get:
gcc -static -L/usr/local/lib -lCamellia -o myprogram myprogram.c
/tmp/cck0pw70.o: In function `main':
myprogram.c:(.text+0x23): undefined reference to `camLoadPGM'
myprogram.c:(.text+0x55): undefined reference to `camAllocateImage'
myprogram.c:(.text+0x97): undefined reference to `camZoom2x'
myprogram.c:(.text+0x104): undefined reference to `camSavePGM'
collect2: ld returned 1 exit status
I want to statically link because I'm trying to modify the Camellia source code and I want to compare my version against theirs. So after some googling, I tried adding AM_DISABLE_SHARED into the configure.in file. But after running ./configure, I still get the exact same Makefile. After I "make install", I still get the same results above.
What is an easy way to get two versions of my code, one with the original Camellia source code compiled and one with my modified version? I think static libraries should work. There is an easy way to get static libraries working or are there other simple solutions to my problem? I just don't want to re-"make" and re-"make install" everytime I want to compare my version against the original.
Did you re-run autoconf after adding AM_DISABLE_SHARED and before configure, make, make install? You also can just use configure --disable-dynamic to stop it building the shared libraries. Make sure you delete any previously installed ones - make uninstall should do that. I can't see anything else obviously wrong. Try being explicit:
gcc -static -o myprogram myprogram.c /usr/local/lib/libCamellia.a
or break it down into two steps and check the symbols in myprogram.o are what you expect with nm myprogram.o.
I am not skillful with autoconf and I don't know why your attempt to link statically fails, but if linking dynamically works I think using shared libraries would actually solve your problem a little better.
Just make two shared libraries, one with the original Camellia code and one with your modified version. Put them in two different directories, and when you run myprogram you can choose between them either by switching LD_LIBRARY_PATH (or whatever you're using to find libraries) or by keeping a symbolic link in /usr/local/lib and switching it between libraries. The advantage of this over static libraries (apart from the fact that this works) is that you can tinker with your modified code, rebuild the shared library and run without having to rebuild myprogram (as long as you don't modify the signatures).
P.S. An experiment: try removing the shared libraries from /usr/local/lib and rebuilding without the -static flag, just as if you were using the shared libraries. In theory this should cause gcc to use the static libraries instead. The results may give a clue to why the static link is failing.