I´m doing my master thesis and I´m having problems configuring the cross compiler. I´ve read lots of articles, faq, tutorial, etc.., but I still think I´m missing something, maybe very stupid and basilar, but without that I can´t have a clear vision of everything and I can´t make it all work…so please feel free to explain me everything. Let´s go to the problem.
I´m using SoClib to design a NoC with many MIPS and many RAMs and I want to load different applications in every RAM, to make every MIPS read from only one RAM. I managed to do that using very simple applications (like hello word with interrupts…) so now I was trying to realize a JPEG decoder to use in a more intense way the MIPS. The problem is that the cross compiler I´m using can´t find the basic stdio functions and so cannot compile the application for the mips. So basically I didn´t managed to install a working full cross compiler.
1) Let´s start with the cross compiler installed with SoClib:
Here there is the guide I´ve followed to install it: http://www.soclib.fr/trac/dev/wiki/CrossCompiler
Now from what I´ve read this is not a full cross compiler. I´ve only installed a first stage gcc that can generate elf code for the mips but that can´t use any C function. With that one, I should compile a C library to create a new cross compiler. So it´s normal that it is not working, although that doesn´t answer to the question: why “Hello world” was working if it uses a printf and includes stdio.h? The answer should be that I´ve seen that SoClib has his header stdio.h with very few function realized (printf is there) so I can´t use the standard C library with that compiler.
2) Because of that I´ve decided to install a full cross compiler and I´ve read so many guides that I will only post the two I´ve used in practice:
http://www.cse.iitb.ac.in/grc/gcc-workshop-11/downloads/slides/gccw11-config-build.pdf from page 108 (before explains the problems with cross compiling)
It uses EGLIBC. The only difference is that I´ve used as target “mipsel-elf”. If I´ve understood correctly, this method requires a 3 stage cross compiler, because eglibc can´t be fully compiled with the first stage cross compiler. Now the problem is that I can´t succeed in installing eglibc with the first stage cross compiler (it fails at page 122). I´have attached the log. I think problems starts when it says: “mipsel-elf-gcc: error: unrecognized option ´-V´” and “mipsel-elf-gcc: fatal error: no input files compilation terminated”. Seems that I´ve some problem with the compiler and all finishes with an unsupported platform…but it should work because I´ve copied the eglibc/ports directory like the guide says…
3) I´ve tried another guide that uses newlib instead of eglibc:
http://www(dot)cygwin(dot)com/ml/crossgcc/2005-08/msg00114/l-cross-ltr.pdf
This guide creates only a 2 stage cross compiler. I think it´s because newlib can be fully compiled with the first stage compiler…am I right? Anyway I get the same problem. I can´t compile newlib and I get similar errors (log attached).
So that´s my problem and here I put some doubt I´ve had trying to resolve it.
4) Can the problem be the target=mipsel-elf? I know from the gnu documentation that the name to configure gnu should be: cpu-manufacturer-os (or cpu-manufacturer-kernel-os) but I´ve read that mipsel-elf is accepted. The elf should be because I don´t want to load an OS on my NoC platform and here the second doubt…
5) Can the problem be the OS thing? The guides uses linux kernel headers…maybe with a mipsel-elf target I shouldn´t configure binutils –with-sysroot? But sounds me strange…what should change?
6) Yet another probably stupid question. The stdio and generally C functions, doesn´t need an OS right? Because all these problems are making me doubt on the basics…so I should be able to run a JPEG decoding C application with a mipsel-elf cross compiler right?
Every consideration, advice and help will be appreciated. If you have some documents about these arguments please let me know, I would like to learn and become more confident in this field.
Thanks
eglibc log:
configure:2426: $? = 0
configure:2433: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -v >&5
Using built-in specs.
COLLECT_GCC=/home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc
COLLECT_LTO_WRAPPER=/home/bertone/programmazione/crosscompiler/mips/install/libexec/gcc/mipsel-elf/4.6.2/lto-wrapper
Target: mipsel-elf
Configured with: ../../../../../Scaricati/src/gcc-4.6.2/configure --target=mipsel-elf --prefix=/home/bertone/programmazione/crosscompiler/mips/install --without-headers --with-newlib --disable-shared --disable-threads --disable-libssp --disable-libgomp --disable-libmudflap --enable-languages=c --disable-werror
Thread model: single
gcc version 4.6.2 (GCC)
configure:2437: $? = 0
configure:2444: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -V >&5
mipsel-elf-gcc: error: unrecognized option '-V'
mipsel-elf-gcc: fatal error: no input files
compilation terminated.
configure:2448: $? = 1
configure:2452: checking for suffix of object files
configure:2478: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -c conftest.c >&5
configure:2482: $? = 0
configure:2507: result: o
configure:2511: checking whether we are using the GNU C compiler
configure:2540: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -c conftest.c >&5
configure:2547: $? = 0
configure:2564: result: yes
configure:2573: checking whether /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc accepts -g
configure:2603: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -c -g conftest.c >&5
configure:2610: $? = 0
configure:2711: result: yes
configure:2728: checking for /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc option to accept ISO C89
configure:2802: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -c -g -O2 conftest.c >&5
conftest.c:9:19: fatal error: stdio.h: No such file or directory
compilation terminated.
configure:2809: $? = 1
configure: failed program was:
| /* confdefs.h. */
| #define PACKAGE_NAME "GNU C Library"
| #define PACKAGE_TARNAME "glibc"
| #define PACKAGE_VERSION "(see version.h)"
| #define PACKAGE_STRING "GNU C Library (see version.h)"
| #define PACKAGE_BUGREPORT "http://sourceware.org/bugzilla/"
| /* end confdefs.h. */
| #include <stdarg.h>
| #include <stdio.h>
| #include <sys/types.h>
| #include <sys/stat.h>
| /* Most of the following tests are stolen from RCS 5.7's src/conf.sh. */
| struct buf { int x; };
| FILE * (*rcsopen) (struct buf *, struct stat *, int);
| static char *e (p, i)
| char **p;
| int i;
| {
| return p[i];
| }
| static char *f (char * (*g) (char **, int), char **p, ...)
| {
| char *s;
| va_list v;
| va_start (v,p);
| s = g (p, va_arg (v,int));
| va_end (v);
| return s;
| }
|
| /* OSF 4.0 Compaq cc is some sort of almost-ANSI by default. It has
| function prototypes and stuff, but not '\xHH' hex character constants.
| These don't provoke an error unfortunately, instead are silently treated
| as 'x'. The following induces an error, until -std is added to get
| proper ANSI mode. Curiously '\x00'!='x' always comes out true, for an
| array size at least. It's necessary to write '\x00'==0 to get something
| that's true only with -std. */
| int osf4_cc_array ['\x00' == 0 ? 1 : -1];
|
| /* IBM C 6 for AIX is almost-ANSI by default, but it replaces macro parameters
| inside strings and character constants. */
| #define FOO(x) 'x'
| int xlc6_cc_array[FOO(a) == 'x' ? 1 : -1];
|
| int test (int i, double x);
| struct s1 {int (*f) (int a);};
| struct s2 {int (*f) (double a);};
| int pairnames (int, char **, FILE *(*)(struct buf *, struct stat *, int), int, int);
| int argc;
| char **argv;
| int
| main ()
| {
| return f (e, argv, 0) != argv[0] || f (e, argv, 1) != argv[1];
| ;
| return 0;
| }
configure:2802: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -qlanglvl=extc89 -c -g -O2 conftest.c >&5
mipsel-elf-gcc: error: unrecognized
option '-qlanglvl=extc89'
configure:2809: $? = 1
NEWLIB log
configure:4049: $? = 0
configure:4038: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -v >&5
Using built-in specs.
COLLECT_GCC=/home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc
COLLECT_LTO_WRAPPER=/home/bertone/programmazione/crosscompiler/mips/install/libexec/gcc/mipsel-elf/4.6.2/lto-wrapper
Target: mipsel-elf
Configured with: ../../../../../Scaricati/src/gcc-4.6.2/configure --target=mipsel-elf --prefix=/home/bertone/programmazione/crosscompiler/mips/install --without-headers --with-newlib --disable-shared --disable-threads --disable-libssp --disable-libgomp --disable-libmudflap --enable-languages=c --disable-werror
Thread model: single
gcc version 4.6.2 (GCC)
configure:4049: $? = 0
configure:4038: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -V >&5
mipsel-elf-gcc: error: unrecognized option '-V'
mipsel-elf-gcc: fatal error: no input files
compilation terminated.
configure:4049: $? = 1
configure:4038: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc -qversion >&5
mipsel-elf-gcc: error: unrecognized option '-qversion'
mipsel-elf-gcc: fatal error: no input files
compilation terminated.
configure:4049: $? = 1
configure:4069: checking for C compiler default output file name
configure:4091: /home/bertone/programmazione/crosscompiler/mips/install/bin/mipsel-elf-gcc conftest.c >&5
/home/bertone/programmazione/crosscompiler/mips/install/lib/gcc/mipsel-elf/4.6.2/../../../../mipsel-elf/bin/ld: cannot find crti.o: No such file or directory
/home/bertone/programmazione/crosscompiler/mips/install/lib/gcc/mipsel-elf/4.6.2/../../../../mipsel-elf/bin/ld: cannot find crtbegin.o: No such file or directory
/home/bertone/programmazione/crosscompiler/mips/install/lib/gcc/mipsel-elf/4.6.2/../../../../mipsel-elf/bin/ld: cannot find -lgcc
/home/bertone/programmazione/crosscompiler/mips/install/lib/gcc/mipsel-elf/4.6.2/../../../../mipsel-elf/bin/ld: cannot find -lgcc
/home/bertone/programmazione/crosscompiler/mips/install/lib/gcc/mipsel-elf/4.6.2/../../../../mipsel-elf/bin/ld: cannot find crtend.o: No such file or directory
/home/bertone/programmazione/crosscompiler/mips/install/lib/gcc/mipsel-elf/4.6.2/../../../../mipsel-elf/bin/ld: cannot find crtn.o: No such file or directory
collect2: ld returned 1 exit status
configure:4095: $? = 1
configure:4132: result:
configure: failed program was:
| /* confdefs.h */
| #define PACKAGE_NAME ""
| #define PACKAGE_TARNAME ""
| #define PACKAGE_VERSION ""
| #define PACKAGE_STRING ""
| #define PACKAGE_BUGREPORT ""
| #define PACKAGE_URL ""
| /* end confdefs.h. */
|
| int
| main ()
| {
|
| ;
| return 0;
| }
configure:4138: error: in `/home/bertone/programmazione/crosscompiler/mips/build/newlib':
configure:4142: error: C compiler cannot create executables
First, off, do you need these stdio functions?
Can you replace these stdio functions. For example do you really really need a printf (one of the more painful ones), or might you be willing to replace those with some thing else, maybe a string_display() and hex_display() functions perhaps that are quite easy to implement and library and system independent.
Are these file I/O functions? I seem to remember one of the popular open source jpeg decoders wanted to do the file I/O for you. How hard is it really to implement an fopen() (just return with anything valid), an fread, which just keeps track of a pointer in some const data you have included in the binary, and does a mem copy, and fwrite, same deal keeps track of a pointer in an array and does a mem copy.
I have embedded jpeg decoders, mp3, etc that make system calls, both by using newlib() which makes it a bit easier but is these days very painful to build a cross compiler with, and also with replacing the system calls with different functions or by implementing simulations of those functions.
If you use the free codesourcery lite (now called codebench or something as it is mentor graphics and not codesourcery anymore) and replace the system calls with simulated ones you dont have to mess with compilers anymore.
I have some very simple mips code, but also some instructions for building the kind of gnu cross compiler I am talking about (works great if you dont have a use for system calls, C library calls, gcc library calls because you implemented everything yourself).
https://github.com/dwelch67/pic32_samples
It is not a far stretch from what I am building to a complete cross compiler. Gcc/gnu is not so stable that you can just cross compile any version from any version and have it work, likewise the cross compilers for any target from any version on any host using any version is not stable. From time to time a version of binutils+gcc+newlib for a target happen to just work. for example, at this period of time/history this worked, both on linux and with mingw on windows.
http://www.dwelch.com/ipod/gccarm.txt
but unless you use a virtual machine (vmware, etc) and install an old linux on it with an old enough build system, you wont be able to build the above.
Not long ago, and perhaps still this worked on a modern system. it is for arm but maybe it will also work for mips.
http://www.cowlark.com/2009-07-04-building-gcc/
I dont think I have any of the code posted with the mp3 and zlib and jpeg libraries used without system calls or with simulated system calls in a place where I can publish them. mallocs are easy to implement for things like zlib if you want to use zlib as a performance test or something, one time compress or decompress or both, not a general purpose thing but a controlled test environment with data in and data out. Same goes for the others controlled environment jpeg encode or decode, mp3 encode or decode (they are lossy so you cant compare input to output necessarily to verify the system produced the right result, a checksum or crc is easy though compared against a known good system/result).
If you examine the files in newlib that I null out and replace later you can see what newlib boils the stdio calls down into, you would then implement each of those functions. I have a couple of old sites that have at least dhrystone examples which pretty much just wanted a printf http://www.dwelch.com/ipod/ http://www.dwelch.com/gba/dhry.htm where you can see what I did with the newlib functions, mostly just neutered them by returning a passing value, and figuring out what newlib does when it wants to output to stdout.
Or an example of replacing some stdio calls with alternate functions resulting in something that still works. not saying it is pretty but you can play the game.
https://github.com/dwelch67/stm32f4d/blob/master/adventure/
Basically what I am reading is that you have identified a problem but are only pursuing one path, solving it on the compiler side. When you could be solving it by creating an abstraction layer or shim, or by removing the offending code completely or replacing it with less offensive code. If you look at the source code for adventure above and similar era games empire, etc, even small c compilers and other code that has been ported a number of times, you can see they did exactly that, they solved the problem by making non-language, non-system specific calls in the bulk of the code, and then the last mile was to implement those calls in one or a few system specific places. Even if you were to get a gcc+newlib or gcc+glibc cross compiler you are still going to have to replace the default stdio to operating system layer with your own. (that is what the compiler side of this is as well, a large bulky abstraction layer, glibc, newlib, etc, that you still have to implement the backend).
Why not just download the free prebuilt GCC Codesourcery Lite cross compiler toolchain for MIPS?
6) The stdio and generally C functions, doesn´t need an OS right?
Usually, libc translates stdio calls into system calls to the underlying OS, so you would need to have an OS for stdio to work. Your JPEG decoder application will work fine if it doesn't call anything that needs an OS, but you will need to create a method to get the results from each individual CPU across your NoC to the outside world.
Related
I'm learning x86 assembly Linux from the book "Programming from the Ground up", and currently I'm learning how to open a file and read from or write to it.
I'm having trouble with the options for opening files. I know 0 is for read-only, 03101 is for write and truncate, but where can I get the full documentation to all the open options?
It's not about WSL, it's about what Linux distro you chose to install in your WSL. Different distros will put things in different places in the filesystem.
locate '*fcntl*.h' is a good way to find the appropriate headers.
You can always compile a C program that includes the documented headers (which will pull in the "real" headers), and look at its gcc -E -dM macro defines. Or even
gcc -E -dM /usr/include/fcntl.h | grep ' O_'
to filter just the O_ macro constants. (That fcntl.h is I think likely to be in the plain /usr/include, not buried somewhere, but maybe that's just my Arch GNU/Linux distro keeping it simple. It keeps Linux-specific libc headers like <asm/unistd.h> in /usr/include/asm/, where you can find unistd_32.h and unistd_64.h for 32 and 64-bit call numbers, respectively.) Or let the usual include-path search happen:
echo '#include <fcntl.h>' | gcc -E -dM - | grep ' O_'
Or write code that does printf("%x, %x\n", O_CREAT, O_TRUNC) or whatever to print out some constants you're interested in, whatever header they came from. (Or print out their bitwise OR, like O_CREAT|O_TRUNC).
The permission mode bit constants like S_IRUSR are defined in terms of other constants like __S_IREAD so it's a bit of a rats nest to follow; probably just printing it out is a good idea. Or simply write permission bits in octal, like mov edx, 0o666 (NASM) or mov $0666, %edx (GAS). (Letting umask clear the write-for-other bit on file creation).
The names of the constants to look for can be found in the man page, open(2).
Using: Windows Subsystem
I have found the correct file for x86 asm linux in /usr/include/asm-generic/fcntl.h
I read the following manual:
http://linux.die.net/man/3/clock_gettime
and I wrote the following code:
#include <time.h>
int main() {
struct timespec clk;
clock_gettime(CLOCK_REALTIME, &clk);
return 0;
}
Surprisingly, I get the following errors:
Symbol CLOCK_REALTIME could not be resolved
undefined reference to clock_gettime
I still don't understand what is the problem. I included the header, and these names show in this header.
maybe you should use#define _POSIX_TIMERS,#define _REENTRANT
besides, when you compile the code, make sure to link the real-time library which is cc filename.c -o filename -lrt
Update 1.0:
sometimes in windows or mac os, C ide may not include real-time library automatically, or we may not used the posix directly without _POSIX_TIMES, therefore you have to link the real-time library manually. In Linux, you can just type in cc filename.c -o filename -lrt to compile the c file.
I have a C program that tries to modify a const string literal. As now I learned that this is not allowed.
When I compile the code with clang test.c the compiler gives no warning. But when I compile it with clang++ test.c it gives a warning:
test.c:6:15: warning: conversion from string literal to 'char *' is deprecated
[-Wdeprecated-writable-strings]
char *s = "hello world";
^
The problem is that it turns out clang++ is just a symbol link of clang:
ll `which clang++`
lrwxr-xr-x 1 root admin 5 Jan 1 12:34 /usr/bin/clang++# -> clang
So my question is how could clang++ behaves differently from clang given that it's a symbol link of clang?
Clang is looking at its argv[0] and altering its behavior depending on what it sees. This is an uncommon and discouraged, but not rare, trick going at least as far back as 4.2BSD ex and vi, which were the same executable, and probably farther.
In this case, clang is compiling your .c file as C, and clang++ is compiling it as C++. This is a historical wart which you should not rely on; use the appropriate compiler command and make sure that your file extension reflects the true contents of the file.
By convention, the name by which a command is invoked is passed as argv[0]; it is not especially unusual for programs to change their behavior based on this. (Historically, ln, cp, and mv were hardlinks to the same executable on Research Unix and used argv[0] to decide which action to do. Also, most shells look for a leading - in argv[0] to decide if they should be a login shell.) Often there is also some other way to get the same effect (options, environment variables, etc.); you should in general use this instead of playing argv[0] games.
There are reasons to do this, but in most cases it's not a good idea to rely on it or to design programs around it.
Problem when compiling gcc 4.6.2:
checking for avr-gcc... /data/data6/soft/src_build/gcc-4.6.2/host-x86_64-unknown-linux-gnu/gcc/xgcc -B/data/data6/soft/src_build/gcc-4.6.2/host-x86_64-unknown-linux-gnu/gcc/ -B/data/data6/soft/programming/gcc-avr/avr/bin/ -B/data/data6/soft/programming/gcc-avr/avr/lib/ -isystem /data/data6/soft/programming/gcc-avr/avr/include -isystem /data/data6/soft/programming/gcc-avr/avr/sys-include
checking for suffix of object files... configure: error: in `/data/data6/soft/src_build/gcc-4.6.2/avr/libgcc':
configure: error: cannot compute suffix of object files: cannot compile
See `config.log' for more details.
In my gcc-4.6.2/avr/libgcc/config.log I found:
configure:3268: /data/data6/soft/src_build/gcc-4.6.2/host-x86_64-unknown-linux-gnu/gcc/xgcc -B/data/data6/soft/src_build/gcc-4.6.2/host-x86_64-unknown-linux-gnu/gcc/ -B/data/data6/soft/programming/gcc-avr/avr/bin/ -B/data/data6/soft/programming/gcc-avr/avr/lib/ -isystem /data/data6/soft/programming/gcc-avr/avr/include -isystem /data/data6/soft/programming/gcc-avr/avr/sys-include -c -g -O2 conftest.c >&5
exec: 89: -o: not found
configure:3272: $? = 1
configure: failed program was:
| /* confdefs.h */
| #define PACKAGE_NAME "GNU C Runtime Library"
| #define PACKAGE_TARNAME "libgcc"
| #define PACKAGE_VERSION "1.0"
| #define PACKAGE_STRING "GNU C Runtime Library 1.0"
| #define PACKAGE_BUGREPORT ""
| /* end confdefs.h. */
|
| int
| main ()
| {
|
| ;
| return 0;
| }
configure:3286: error: in `/data/data6/soft/src_build/gcc-4.6.2/avr/libgcc':
configure:3289: error: cannot compute suffix of object files: cannot compile
See `config.log' for more details.
If this command run directly (after creating conftest.c) output will be same.
What is exec: 89: -o: not found? Which program write it? Where is 89 line?
Solutions like exporting LD_CONFIG_PATH or adding lines in /etc/ld.so.conf with path to GMP, MPFR, MPC not work for me.
Where find solution for this problem?
Update. It seems is configuration problem. Gcc has very good option -v :-) And i see that is assembler call from gcc, which fails. With option -save-temps i saved assembler source, and run assembler, but it doesnn't matter.
Instead of calling real assembler it calls shell-script wrapper, in my case /data/data6/soft/src_build/avr-gcc/gcc/as, which begins from:
ORIGINAL_AS_FOR_TARGET=""
ORIGINAL_LD_FOR_TARGET=""
ORIGINAL_PLUGIN_LD_FOR_TARGET=""
ORIGINAL_NM_FOR_TARGET=""
And when it form command-line call, we have this:
exec -o conftest.o conftest.s
which fails as described above.
Where is necessary fix this? To get call like this:
exec as -o conftest.o conftest.s
Where good place to set variable ORIGINAL_AS_FOR_TARGET?
I had the same problem cross-compiling for mips.
Step one: Cross-compile the binutils for your target. Install them somewhere sane. I use /usr/local/[target]-gcc
Make sure you configure them with --program-prefix=[target]-
Something like (untested):
configure --prefix=/usr/local/avr-gcc --program-prefix=avr- --target=avr
Then you need to set up some environment variables so GCC can find them:
export AR_FOR_TARGET=/usr/local/avr-gcc/bin/avr-ar
export LD_FOR_TARGET=/usr/local/avr-gcc/bin/avr-ld
export OBJDUMP_FOR_TARGET=/usr/local/avr-gcc/bin/avr-objdump
export NM_FOR_TARGET=/usr/local/avr-gcc/bin/avr-nm
export RANLIB_FOR_TARGET=/usr/local/avr-gcc/bin/avr-ranlib
export READELF_FOR_TARGET=/usr/local/avr-gcc/bin/avr-readelf
export STRIP_FOR_TARGET=/usr/local/avr-gcc/bin/avr-strip
export AS_FOR_TARGET=/usr/local/avr-gcc/bin/avr-as
Now you can configure and compile GCC. Ensure you start from a completely empty build directory, otherwise gcc/as etc won't get recreated.
"*Building GCC is not trivial, but is not difficult if you follow the instructions carefully.
Many people rush into trying to build it without reading the installation docs properly and make one or more of these common mistakes:
1) do not run ./configure from gcc src dir (this is not supported) => you need to run configure from outside the gcc source directory
2) Note: if GCC links dynamically to the prerequisite libs (GMP/MPFR/MPC) then the shared libraries must be in the dynamic linker's path (LD_LIBRARY_PATH), both when building gcc and when using the installed compiler.*"
Simple example (without dynamic link to GMP/MPFR/MPC):
tar xzf gcc-4.8.0.tar.gz
cd gcc-4.8.0
./contrib/download_prerequisites
cd ..
mkdir objdir
cd objdir
$PWD/../gcc-4.8.0/configure --prefix=/opt/gcc-4.8.0
make
make install
Sources:
Advogato Doc -
GNU Doc
This question is related to this one as well as its answer.
I just discovered some ugliness in a build I'm working on. The situation looks somewhat like the following (written in gmake format); note, this specifically applies to a 32-bit memory model on sparc and x86 hardware:
OBJ_SET1 := some objects
OBJ_SET2 := some objects
# note: OBJ_SET2 doesn't get this flag
${OBJ_SET1} : CCFLAGS += -PIC
${OBJ_SET1} ${OBJ_SET2} : %.o : %.cc
${CCC} ${CCFLAGS} -m32 -o ${#} -c ${<}
obj1.o : ${OBJ_SET1}
obj2.o : ${OBJ_SET2}
sharedlib.so : obj1.o obj2.o
obj1.o obj2.o sharedlib.so :
${LINK} ${LDFLAGS} -m32 -PIC -o ${#} ${^}
Clearly it can work to mix objects compiled with and without PIC in a shared object (this has been in use for years). I don't know enough about PIC to know whether it's a good idea/smart, and my guess is in this case it's not needed but rather it's happening because someone didn't care enough to find out the right way to do it when tacking on new stuff to the build.
My question is:
Is this safe
Is it a good idea
What potential problems can occur as a result
If I switch everything to PIC, are there any non-obvious gotchas that I might want to watch out for.
Forgot I even wrote this question.
Some explanations are in order first:
Non-PIC code may be loaded by the OS into any position in memory in [most?] modern OSs. After everything is loaded, it goes through a phase that fixes up the text segment (where the executable stuff ends up) so it correctly addresses global variables; to pull this off, the text segment must be writable.
PIC executable data can be loaded once by the OS and shared across multiple users/processes. For the OS to do this, however, the text segment must be read-only -- which means no fix-ups. The code is compiled to use a Global Offset Table (GOT) so it can address globals relative to the GOT, alleviating the need for fix-ups.
If a shared object is built without PIC, although it is strongly encouraged it doesn't appear that it's strictly necessary; if the OS must fix-up the text segment then it's forced to load it into memory that's marked read-write ... which prevents sharing across processes/users.
If an executable binary is built /with/ PIC, I don't know what goes wrong under the hood but I've witnessed a few tools become unstable (mysterious crashes & the like).
The answers:
Mixing PIC/non-PIC, or using PIC in executables can cause hard to predict and track down instabilities. I don't have a technical explanation for why.
... to include segfaults, bus errors, stack corruption, and probably more besides.
Non-PIC in shared objects is probably not going to cause any serious problems, though it can result in more RAM used if the library is used many times across processes and/or users.
update (4/17)
I've since discovered the cause of some of the crashes I had seen previously. To illustrate:
/*header.h*/
#include <map>
typedef std::map<std::string,std::string> StringMap;
StringMap asdf;
/*file1.cc*/
#include "header.h"
/*file2.cc*/
#include "header.h"
int main( int argc, char** argv ) {
for( int ii = 0; ii < argc; ++ii ) {
asdf[argv[ii]] = argv[ii];
}
return 0;
}
... then:
$ g++ file1.cc -shared -PIC -o libblah1.so
$ g++ file1.cc -shared -PIC -o libblah2.so
$ g++ file1.cc -shared -PIC -o libblah3.so
$ g++ file1.cc -shared -PIC -o libblah4.so
$ g++ file1.cc -shared -PIC -o libblah5.so
$ g++ -zmuldefs file2.cc -Wl,-{L,R}$(pwd) -lblah{1..5} -o fdsa
# ^^^^^^^^^
# This is the evil that made it possible
$ args=(this is the song that never ends);
$ eval ./fdsa $(for i in {1..100}; do echo -n ${args[*]}; done)
That particular example may not end up crashing, but it's basically the situation that had existed in that group's code. If it does crash it'll likely be in the destructor, usually a double-free error.
Many years previous they added -zmuldefs to their build to get rid of multiply defined symbol errors. The compiler emits code for running constructors/destructors on global objects. -zmuldefs forces them to live at the same location in memory but it still runs the constructors/destructors once for the exe and each library that included the offending header -- hence the double-free.