I'd like to specify GCC's search directory for the startfile and
endfile: crt1.o, crti.o and crtn.o. Passing -B on the command line to
the GCC driver works, but is inconvenient. How do I modify the specs
file (lib/gcc/x86_64-unknown-linux-gnu/4.9.2/specs) to specify the search path for startfile?
I tried adding the -B option to the startfile spec and got the error:
ld: unrecognized option '-B/gsc/btl/linuxbrew/lib'
I then tried adding the -B option to the cc1 spec and got the error:
cc1: error: command line option '-B/gsc/btl/linuxbrew/lib' is valid
for the driver but not for C
If it's not possible to do this via the specs file, is there an environment variable or a configure option to GCC that accomplishes the same goal?
I've installed a recent version of glibc in my home directory. Everything's working great. I've modified the specs file to link against the new version of glibc, but it's still linking against the old system version in /usr of startfile and endfile.
Here's a the unanswered question on the gcc-help mailing list. Here's a related Linuxbrew bug, gfortran is broken with stand alone Linuxbrew, and a proposed fix, gcc, binutils: link to Cellar instead of system libs.
Thanks,
Shaun
You can use an absolute path in the *startfile: and *endfile: sections in the specs file, instead of the default relative paths. This will override GCC's choice of the default location.
As per https://wiki.debian.org/Multiarch/LibraryPathOverview, gcc will look for startup files such as crt1.o in $(sysroot)/lib, so you can specify the --sysroot option when running gcc, or compile gcc with --with-sysroot.
Related
I have an OSX 10.7 computer with a non-administrator account, and was attempting to install the pre-compiled versions of gcc and g++ found here. I've attempted to use the answers presented in these questions (three different links) to compile some code with g++, to confusing avail. I have a folder structure like this:
~/code/:
usr/:
local/:
bin/ (3.6MB)
include/ (8.6MB)
lib/ (51MB)
libexec/ (49MB)
share/ (16MB)
c++/:
source/ (contains .cpp files)
g++ -v returns this:
code USER$ usr/local/bin/g++ -v
Using built-in specs.
COLLECT_GCC=usr/local/bin/g++
COLLECT_LTO_WRAPPER=/Users/USERNAME/code/usr/local/bin/../libexec/gcc/ x86_64-apple-darwin11.4.0/4.7.1/lto-wrapper
Target: x86_64-apple-darwin11.4.0
Configured with: ../gcc-4.7.1/configure --enable-languages=fortran
Thread model: posix
gcc version 4.7.1 (GCC)
An attempt at compiling a file that "#include"s only iostream:
$ usr/local/bin/g++ c++/source/test.cpp -o ex6
In file included from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/bits/postypes.h:42:0,
from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/iosfwd:42,
from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/ios:39,
from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/ostream:40,
from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/iostream:40,
from c++/source/ex6.cpp:1:
/Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/cwchar:46:19: fatal error: wchar.h: No such file or directory
compilation terminated.
I tried compiling with some flags recommended in one of the links mentioned, like this: (with all combinations of "usr/" to "usr/local/include/" and "usr/" to "/usr/local/lib" giving the same result (which is also the same as using no flags).
$ /Users/USERNAME/code/usr/local/bin/g++ source/ex6.cpp -I/Users/USERNAME/code/usr/local/include/ -L/Users/USERNAME/code/usr/local/lib/In file included from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/bits/postypes.h:42:0,
from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/iosfwd:42,
from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/ios:39,
from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/ostream:40,
from /Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/iostream:40,
from source/ex6.cpp:1:
/Users/USERNAME/code/usr/local/bin/../lib/gcc/x86_64-apple-darwin11.4.0/4.7.1/../../../../include/c++/4.7.1/cwchar:46:19: fatal error: wchar.h: No such file or directory
compilation terminated.
In short, I'm having trouble understanding what the answers in the links provided are saying to do. I saw reference to a specs file, which I could find no specific information for, and "-Wl,-rpath,$(DEFAULT_LIB_INSTALL_PATH)", for which I couldn't figure out what I was supposed to substitute for "DEFAULT_LIB_INSTALL_PATH".
What should I do to point the downloaded g++ compiler to its own files without placing them in their default location, as I do not have administrative capabilities on this account?
I will provide any information as necessary.
It looks like you don't have required header files. You need to install Command Line Tools from Apple Developers site (free registration needed). The problem is that you don't have administrator account. I suggest that you ask the administrator to install the tools for you. If it is not possible you could try to extract the contents of downloaded package (DevSDK.pkg) to your local directory (Pacifist can do that) and pass the path with the missing headers to your compiler. I haven't tried that though.
I downloaded
Ruben’s build of
Cygwin GCC.
However upon running it seems unable to compile any files
$ touch foo.c
$ gcc foo.c
gcc: error: spawn: No such file or directory
As a workaround, I found this to work
i686-w64-mingw32-gcc foo.c
I had the same problem and solved it by installing the g++ package in addition to gcc-core
I had this same problem on Cygwin64, and the solution was PATH related..kinda.
Turns out, there are copies of gcc in /usr/bin and /bin (at least, there is in my install).
Executing /bin/gcc failed with the error above -- I'm guessing due to incorrectly assumed relative paths???
Executing /usr/bin/gcc works as expected!
In my case, the "problem" was that I had inadvertently injected "/bin" into my PATH environment variable, resulting in /bin/gcc being executed, instead of /usr/bin/gcc. Removing the "/bin" from the path solved the problem.
Still unclear why there are two gcc binaries (which appear to be identical) in different places... but maybe the Cygwin gurus can answer that; or maybe my installation is just foo-barred.
Ruben's builds are not Cygwin GCC packages, rather they are cross-compilers which run on various platforms but target native Windows using the MinGW-w64 toolchain.
In any case, you shouldn't be using them on Cygwin. If you want to compile Cygwin executables, install the gcc4 packages; if you want to cross-compile for Windows, install the mingw64-i686-gcc (for Win32) or mingw64-x86_64-gcc (for Win64) packages instead.
Gcc isn't really the compiler. It's a front end program that orchestrates the execution of any necessary compiler, assembler, and linker components. Typically these others are separately compiled programs.
So, gcc is trying (kind of) to tell you that it can't find the compiler. I guess it needs to be on your PATH or in an expected location.
If you are executing this from a Windows DOS box then it definitely needs a windows PATH setting.
I like to install Cygwin, making sure to include rxvt. At that point, you can configure a purely sh(1) path and your environment is rather more civilized.
I had the same error when I tried to extract a couple of executables from cygwin installation dirctory and copied them into another location.
strace shows me the file which was not found by spawn:
/lib/gcc/x86_64-pc-cygwin/6.4.0/cc1.exe
When I copied cc1.exe into the location relative to
<dir with sh.exe and cpp.exe>/../lib/gcc/x86_64-pc-cygwin/6.4.0/cc1.exe
it works fine.
This error occurs whenever cygwin cc can't find a required file.
For those running stuff within cygwin's bin directly from a Windows shell, a gotcha to watch out for is that Windows allow you to run programs from the command line like this:
e:cyg/bin/gcc -flags
Notice that there is no slash between e: and cyg.
So this command would successfully start cygwin gcc from the Windows shell, but halfway through the run it will error out because some component(s) of gcc will utilize the first argument of the input e:cyg/bin/gcc and unlike mingw, this is not a valid path for cygwin gcc.
This can be fixed simply by changing the command to:
e:/cyg/bin/gcc -flags
Notice the slash in between e: and cyg.
A similar gotcha is due to Windows allowing paths like e:/../folder1 as an alternative to e:/folder1. Windows does not give you an error if you are at the root folder and try to go up another folder using ...
So you could start running cygwin gcc using the command:
e:/../cyg/bin/gcc -flags
..or even:
e:/../../../../../../../../../cyg/bin/gcc -flags
However, it would fail halfway with gcc: error: spawn: No such file or directory because some component(s) of cygwin gcc would attempt to run gcc using the first argument of the command input itself, and unlike mingw, e:/../cyg/bin/gcc is not recognized as a valid path by cygwin because you are going up a folder when there's no folder to go up to.
As like above, this can be fixed by keeping the path valid:
e:/cyg/bin/gcc -flags
Make sure the source file extension is in lowercase (i.e. main.c, not main.C):
$ gcc -o main main.C
$ gcc: error: spawn: No such file or directory
$ gcc -o main main.c
$ # all good
This only refers to the case of the extension as given to the gcc, the actual source file can have the extension in whatever case you want.
Explanation: This is from my experimenting with cygwin and gcc, I don't know the actual reason for this behavior.
I have a problem with the implicit LIBRARY_PATH modification of g++ between two versions (g++ -v gives this info). I set the LIBRARY_PATH to a single local directory where I have custom libraries. However, it turned out that only one version of the g++ (let's call it version A) correctly linked, the other (version B) linked to the system-default, which was not desired. Apparently, the order of directories was mixed up and my specification was not properly respected. It is a similar issue to LIBRARY_PATH not used before /usr/lib anymore in gcc 4.2 and later? although not with these versions.
Somehow I came along the idea to have a look at the specs files of the two different versions (got them via g++ -dumpspecs > specs). I then tried to see if running the version of g++ (B; that was producing the *un*expected modifications) with the specs file of the other version (A) would still yield that modification and to my relief the LIBRARY_PATH was now exactly as I expected it (matching version A)!
I further traced down the place of this weird modification to happen at the following line:
. !m64 !m32;.:../lib64 m64 !m32;.:../lib32 !m64 m32;
Besides appearing to affect the setting/modification of LIBRARY_PATH, I sadly have no clue what this line means.
Therefore I hope that some of you is able to "decipher" this line and explain what it means so that I can try to modify it according to my requirements.
Thank you!
That line affects how libraries are found relative to GCC's $PREFIX/lib directory (where $PREFIX is the directory GCC was installed to.)
There are three parts to it:
$PREFIX/lib/. is used when neither -m32 or -m64 is used on the command-line.
$PREFIX/lib/.:$PREFIX/lib/../lib64 is used when -m64 is used.
$PREFIX/lib/.:$PREFIX/lib/../lib32 is used when -m32 is used.
This suggests to me you are using Debian or Ubuntu, I don't think a vanilla GCC build from the FSF sources would have that in the specs. Were both your GCC versions from .deb packages or did you install them yourself? (The Multi Arch support in recent Debian/Ubuntu releases moves library directories and so breaks vanilla GCC, I think Debian and Ubuntu patch the GCC code for their .deb packages.)
Could you add the output of linking with g++ -v for each of your versions, to see the exact library search paths used by each version?
Also, why not just use -L instead of LIBRARY_PATH? Directories specified with -L always come first, before the system dirs or GCC's own dirs or the ones specified in LIBRARY_PATH.
There is a laptop on which I have no root privilege.
onto the machine I have a library installed using configure --prefix=$HOME/.usr .
after that, I got these files in ~/.usr/lib :
libXX.so.16.0.0
libXX.so.16
libXX.so
libXX.la
libXX.a
when I compile a program that invokes one of function provided by the library with this command :
gcc XXX.c -o xxx.out -L$HOME/.usr/lib -lXX
xxx.out was generated without warning, but when I run it error like this was thrown:
./xxx.out: error while loading shared libraries: libXX.so.16: cannot open shared object file: No such file or directory , though libXX.so.16 resides there.
my clue-less assumption is that ~/.usr/lib wasn't searched when xxx.out is invoked.
but what can I do to specify path of .so , in order that xxx.out can look there for .so file?
An addition is when I feed -static to gcc, another error happens like this:
undefined reference to `function_proviced_by_the_very_librar'
It seems .so does not matter even though -L and -l are given to gcc.
what should I do to build a usable exe with that library?
For other people who has the same question as I did
I found a useful article at tldp about this.
It introduces static/shared/dynamic loaded library, as well as some example code to use them.
There are two ways to achieve that:
Use -rpath linker option:
gcc XXX.c -o xxx.out -L$HOME/.usr/lib -lXX -Wl,-rpath=/home/user/.usr/lib
Use LD_LIBRARY_PATH environment variable - put this line in your ~/.bashrc file:
export LD_LIBRARY_PATH=/home/user/.usr/lib
This will work even for a pre-generated binaries, so you can for example download some packages from the debian.org, unpack the binaries and shared libraries into your home directory, and launch them without recompiling.
For a quick test, you can also do (in bash at least):
LD_LIBRARY_PATH=/home/user/.usr/lib ./xxx.out
which has the advantage of not changing your library path for everything else.
Should it be LIBRARY_PATH instead of LD_LIBRARY_PATH.
gcc checks for LIBRARY_PATH which can be seen with -v option
I have installed the latest version of Cygwin, selecting the following packages during setup:
libgcc1
gcc
gcc-core
And created a file (test.c) with only this line:
#include <link.h>
Then ran the following from my Cygwin bash:
$ gcc test.c
... but got this error:
test.c:1:18: link.h: No such file or directory
Any ideas how I can fix it?
Cygwin is based on the Win32 subsystem, which means it uses Windows' executable format (COFF) and dynamic linker, i.e. it does not have an ELF dynamic linker. Hence providing the ELF-specific <link.h> would probably make little sense.