On Sun, Nov 22, 2009 at 05:43:14PM +0100, Luigi Rizzo wrote:
> Hi,
> there is a well known problem with tcc and FreeBSD in the generation
> of elf objects -- see
> http://lists.gnu.org/archive/html/tinycc-devel/2005-07/msg00070.html
>
> Apparently Sergey Lyubka has tried a partial fix to the problem.
> I was wondering if Sergey or someone can post some more detail on
> what needs to be done so we can try to help fixing this issue
I think i have managed to solve the problem and produce
almost valid elf files on FreeBSD. The two patches attached
address a few problems (trying to explain to the
best of my knowledge; i am not very familiar with ELF and
the FreeBSD ELF conventions):
1. ELF file format
tcc produces an ELF executable which is good for linux but
not for FreeBSD. It misses the PHDR section which is almost
mandatory for shared executables, puts in the .dynsym section
some relocation info that FreeBSD expects to be in .got,
and expect the relocation sections to be contiguous.
patch-tccelf.c tries to address the above problem using
conditional sections (so hopefully can be imported upstream)
and also adds the ability to override the name of the dynamic
loader through an environment variable (this is important to
debug tcc).
2. predefined macros
patch-libtcc.c adds/fixes some predefined macros when compiling
on FreeBSD: these are __FreeBSD__ and the usual set of
__i386__ and __unix__ variants.
It also sets __INTEL_COMPILER so we can grab the __aligned
macro from cdefs.h , otherwise many programs would fail
The resulting elf file is still not 100% correct -- if you strip it,
the program will not run (presumably there is some dangling reference).
Other than that, program do seem to run correctly.
It would be nice to integrate these patches in the main repository.
The FreeBSD specific code is in #ifdef so it should not harm
linux users
cheers
luigi
TCC produces code which is incompatible with GCC for the following code:
printf("%lld\n", (long long)-2147483648);
printf("%lld\n", (long long)-2147483649);
For now, just avoid using the corner value.
- Now we can run tcc -run tcc.c successfully, though there are some bugs.
- Remove jmp_table and got_table and use text_section for got and plt entries.
- Combine buffers in tcc_relocate().
- Use R_X86_64_64 instead of R_X86_64_32 for R_DATA_32 (now the name R_DATA_32 is inappropriate...).
- Add got_table in TCCState. This approach is naive and the distance between executable code and GOT can be longer than 32bit.
- Handle R_X86_64_GOTPCREL properly. We use got_table for TCC_OUTPUT_MEMORY case for now.
- Fix load() and store() so that they access global variables via GOT.
- Add a macro TCC_OUTPUT_DLL_WITH_PLT.
-- Now, the DLL with PLT support works only on x86-64, but we may be able to support it on all architectures eventually.
- Define TCC_OUTPUT_DLL_WITH_PLT when target architecture is x86-64.
- Current status (x86-64):
-- Main program should be able to call functions in shared objects.
-- Main program should be able to use global variables in shared objects.
-- Shared objects should be able to call functions in main program.
-- Shared objects can NOT use global variables in main program.
- To fix the last issue, we may need to add support of -fPIC option in our code generator.
Most change was done in #ifdef TCC_TARGET_X86_64. So, nothing should be broken by this change.
Summary of current status of x86-64 support:
- produces x86-64 object files and executables.
- the x86-64 code generator is based on x86's.
-- for long long integers, we use 64bit registers instead of tcc's generic implementation.
-- for float or double, we use SSE. SSE registers are not utilized well (we only use xmm0 and xmm1).
-- for long double, we use x87 FPU.
- passes make test.
- passes ./libtcc_test.
- can compile tcc.c. The compiled tcc can compile tcc.c, too. (there should be some bugs since the binary size of tcc2 and tcc3 is differ where tcc tcc.c -o tcc2 and tcc2 tcc.c -o tcc3)
- can compile links browser. It seems working.
- not tested well. I tested this work only on my linux box with few programs.
- calling convention of long-double-integer or struct is not exactly the same as GCC's x86-64 ABI.
- implementation of tcc -run is naive (tcc -run tcctest.c works, but tcc -run tcc.c doesn't work). Relocating 64bit addresses seems to be not as simple as 32bit environments.
- shared object support isn't unimplemented
- no bounds checker support
- some builtin functions such as __divdi3 aren't supported
- Use REL_SECTION_FMT instead of ".rel%s".
- Use PTR_SIZE instead of sizeof(int) for GOT entries.
- Use sizeof(ElfW(Dyn)) instead of magic number 8.
- Use TCC_ELFCLASS instead of ELFCLASS32.
This patch adds a switch --with-libgcc to configure.
When passed it prevents libtcc1.a from being built and links to
/lib/libgcc_s.so.1 instead of PREFIX/lib/tcc/libtcc1.a.
It will work on ARM when using libgcc from GCC >= 4.2.0.
Prior versions don't have the __floatun[sd]i[sdx]f functions.
It won't work on i386 because of two missing symbols emitted when
floats are cast to integers, but users can provide those symbols
(global short constants) in their code if needed.
Daniel
