this is a bit complicated: for i386 and x86-64 we really need to
extend return values ourself, as the common code now does. For arm64
this at least preserves old behaviour. For riscv64 we don't have to
extend ourself but can expect things to be extended up to int (this
matters for var-args tests, when the sign-extension to int64 needs to
happen explicitely). As the extensions are useless, don't do them.
And for arm32 we actually can't express GCC behaviour: the callee side
expects the return value to be correctly extended to int32, but
remembers the original type. In case the ultimate target type for the
call result is only int, no further extension is done. But in case
the target type is e.g. int64 an extension happens, but not from int32
but from the original type. We don't know the ultimate target type,
so we have to choose a type to put into vtop:
* original type (plus VT_MUSTCAST) - this looses when the ultimate
target is int (GCC: no cast, TCC: a cast)
* int (without MUSTCAST) - this looses when the ultimate target is
int64 (GCC: cast from original type, TCC: cast from int)
This difference can only be seen with undefined sources, like the
testcases, so it doesn't seem worthwhile to try an make it work, just
disable the test on arm and choose the second variant as that generates
less code.
This allows creation of TCCStates and operation with API
calls independently from each other, even from threads.
Frontend (option parsing/libtcc.c) and backend (linker/tccelf.c)
now depend only on the TCCState (s1) argument.
Compilation per se (tccpp.c, tccgen.c) is still using
globals for convenience. There is only one entry point
to this section which is tcc_compile() which is protected
by a semaphore.
There are some hacks involved to avoid too many changes,
as well as some changes in order to avoid too many hacks ;)
The test libtcc_test_mt.c shows the feature. Except this
new file the patch adds 87 lines overall.
- libtcc.c/tccpp.c: fix -U option for multiple input files
- libtcc: remove decl of tcc_add_crt() for PE
- tcc.h: define __i386__ and __x86_64__ for msvc
- tcc.h: undef __attribute__ for __TINYC__ on gnu/linux platforms
- tccelf.c: disable prepare_dynamic_rel unless x86/x64
- tccpe.c: construct rather than predefine PE section flags
- tccpp.c: (alt.) fix access of dead stack variable after error/longjmp
- x86_64-gen.c: fix func_alloca chain for nocode_wanted
- tccpp.c/tccgen.c: improve file:line info for inline functions
- winapi/winnt.h: correct position for DECLSPEC_ALIGN attribute
- win32/lib/crt: simplify top exception handler (needed for signal)
- arm64-gen.c: remove dprintf left from VT_CMP commit
- tccgen.c: limit binary scan with gcase to > 8 (= smaller code)
- tccgen.c: call save_regs(4) in gen_opl for cmp-ops (see test in tcctest.c)
A more automatic approach to code suppression (aka. nocode_wanted)
The simple rules are:
- Clear 'nocode_wanted' at (im/explicit) label IF it was used
- Set 'nocode_wanted' after unconditional jumps
Also in order to test this then I did add the "function might
return no value" warning, and then to make that work again I
did add the __attribute__((noreturn)).
Also moved the look ahead label check into the type parser
to gain a little speed.
In gfunc_call, regisger will be saved before gcall_or_jmp. The register
stored the function will be saved too, though in some generator the SValue
of this function will be immediately poped after gcall_or_jmp, and no need to be saved. So I modify some generator to avoid save redundant SValue before gcall_or_jmp.
linkers don't treat relocations using symindex 0 (undefined)
very well, it can't be misused as indicator for an absolute number.
Just don't bother with special casing this, rather emit an indirect
call/jump right away. ARM64 needs the same (and didn't handle
calls via constant absolute func pointers before).
The testcase as is doesn't fail without the patch, it actually
needs separate compilation (to -fPIC .o file, then to shared lib)
to fail.
which requires being able to emit an arbitrary number of NOP
instructions, which is also implemented here. For x86 we
could emit other sequences but these are the easiest.
tcc.h:
* cleanup struct 'Sym'
* include some 'Attributes' into 'Sym'
* in turn get rid of VT_IM/EXPORT, VT_WEAK
* re-number VT_XXX flags
* replace some 'long' function args by 'int'
tccgen.c:
* refactor parse_btype()
- configure:
- add --config-uClibc,-musl switch and suggest to use
it if uClibc/musl is detected
- make warning options magic clang compatible
- simplify (use $confvars instead of individual options)
- Revert "Remove some unused-parameter lint"
7443db0d5f
rather use -Wno-unused-parameter (or just not -Wextra)
- #ifdef functions that are unused on some targets
- tccgen.c: use PTR_SIZE==8 instead of (X86_64 || ARM64)
- tccpe.c: fix some warnings
- integrate dummy arm-asm better
Also:
- on windows i386 and x86-64, structures of size <= 8 are
NOT returned in registers if size is not one of 1,2,4,8.
- cleanup: put all tv-push/pop/swap/rot into one place
tccgen.c: remove any 'nocode_wanted' checks, except in
- greloca(), disables output elf symbols and relocs
- get_reg(), will return just the first suitable reg)
- save_regs(), will do nothing
Some minor adjustments were made where nocode_wanted is set.
xxx-gen.c: disable code output directly where it happens
in functions:
- g(), output disabled
- gjmp(), will do nothing
- gtst(), dto.
C standard specifies that array should be declared with a non null size
or with * for standard array. Declaration of relocs_info in tcc.h was
not respecting this rule. This commit add a R_NUM macro that maps to the
R_<ARCH>_NUM macros and declare relocs_info using it. This commit also
moves all linker-related macros from <arch>-gen.c files to <arch>-link.c
ones.
With -run the call instruction and a defined function can be
far away, if the function is defined in the executable itself,
not in the to be compiled code. So we always need PLT slots
for -run, not just for undefined symbols.
This is a work-around for TCC's linker, on AArch64, not building a PLT
when TCC is invoked with "-run". Fixing the linker should be possible:
it works on arm and x86_64, apparently.
* Documentation is now in "docs".
* Source code is now in "src".
* Misc. fixes here and there so that everything still works.
I think I got everything in this commit, but I only tested this
on Linux (Make) and Windows (CMake), so I might've messed
something up on other platforms...
Jsut for testing. It works for me (don't break anything)
Small fixes for x86_64-gen.c in "tccpp: fix issues, add tests"
are dropped in flavor of this patch.
Pip Cet:
Okay, here's a first patch that fixes the problem (but I've found
another bug, yet unfixed, in the process), though it's not
particularly pretty code (I tried hard to keep the changes to the
minimum necessary). If we decide to actually get rid of VT_QLONG and
VT_QFLOAT (please, can we?), there are some further simplifications in
tccgen.c that might offset some of the cost of this patch.
The idea is that an integer is no longer enough to describe how an
argument is stored in registers. There are a number of possibilities
(none, integer register, two integer registers, float register, two
float registers, integer register plus float register, float register
plus integer register), and instead of enumerating them I've
introduced a RegArgs type that stores the offsets for each of our
registers (for the other architectures, it's simply an int specifying
the number of registers). If someone strongly prefers an enum, we
could do that instead, but I believe this is a place where keeping
things general is worth it, because this way it should be doable to
add SSE or AVX support.
There is one line in the patch that looks suspicious:
} else {
addr = (addr + align - 1) & -align;
param_addr = addr;
addr += size;
- sse_param_index += reg_count;
}
break;
However, this actually fixes one half of a bug we have when calling a
function with eight double arguments "interrupted" by a two-double
structure after the seventh double argument:
f(double,double,double,double,double,double,double,struct { double
x,y; },double);
In this case, the last argument should be passed in %xmm7. This patch
fixes the problem in gfunc_prolog, but not the corresponding problem
in gfunc_call, which I'll try tackling next.
The common code to move a returned structure packed into
registers into memory on the caller side didn't take the
register size into account when allocating local storage,
so sometimes that lead to stack overwrites (e.g. in 73_arm64.c),
on x86_64. This fixes it by generally making gfunc_sret also return
the register size.
__clear_cache is defined in lib-arm64.c with a single call to
__arm64_clear_cache, which is the real built-in function and is
turned into inline assembler by gen_clear_cache in arm64-gen.c
Michael Matz