A test program:
//////////////
int main()
{
void *p = ({ 0 ; ((void *)1); });
}
/////////////
Porblem is introduced in a commit a80acab: Display error on statement expressions with complex return type
This error is exposed when compiling a linux 2.4.26. tcc 0.9.23 can sucessfully compile
this version of the linux.
Current tcc don't understand an initialization of the empty struct
This problem was found trying to compile a linux kernel 2.4.26
which can be compiled by tcc 0.9.23
A test program:
////////////////////
// ./tcc -c test_3.c
// test_3.c:31: error: too many field init
#undef __GNUC__
#undef __GNUC_MINOR__
#define __GNUC__ 2
#define __GNUC_MINOR__ 95
typedef struct { } rwlock_t;
struct fs_struct {
int count;
rwlock_t lock;
int umask;
};
#define INIT_FS { \
1, \
RW_LOCK_UNLOCKED, \
0022, \
}
#if (__GNUC__ > 2 || __GNUC_MINOR__ > 91)
typedef struct { } rwlock_t;
#define RW_LOCK_UNLOCKED (rwlock_t) { }
#else
typedef struct { int gcc_is_buggy; } rwlock_t;
#define RW_LOCK_UNLOCKED (rwlock_t) { 0 }
#endif
static struct fs_struct init_fs = INIT_FS;
// static struct fs_struct init_fs = { { (1) }, (rwlock_t) { 0 }, 0022, };
// ^ with this all Ok
// static struct fs_struct init_fs = { { (1) }, (rwlock_t) { }, 0022, };
// ^ current tcc don't understand, but tcc 0.9.23 can
int main()
{
return 0;
}
////////////////////
A regression is detected after a patch 69fdb57edd
////////////////////
// A test for patch 69fdb57edd
// Author: grischka <grischka>
// Date: Wed Jun 17 02:09:07 2009 +0200
// unions: initzialize only one field
// struct {
// union {
// int a,b;
// };
// int c;
// } sss = { 1,2 };
// This had previously assigned 1,2 to a,b and 0 to c which is wrong.
//
// Expected: sss.a=1 sss.b=1 sss.c=2
int main()
{
struct {
union {
int a,b;
};
int c;
} sss = { 1, 2 };
printf ("sss.a=%d sss.b=%d sss.c=%d\n", sss.a, sss.b, sss.c);
return 0;
}
////////////////////
A regression was found trying to compile a linux kernel 2.4.26
which can be compiled by tcc 0.9.23
///////////////////
#include <stdio.h>
// test for a bug:
// compiler don't understand am extern array of structs
// $ tcc test_1.c
// test_1.c:8: error: unknown struct/union/enum
extern struct FILE std_files[4];
int main()
{
return 0;
}
//////////////////
tcc-current
/* enum/struct/union declaration. u is either VT_ENUM or VT_STRUCT */
static void struct_decl(CType *type, int u, int tdef)
...
if (tok != '{') {
v = tok;
next();
/* struct already defined ? return it */
if (v < TOK_IDENT)
expect("struct/union/enum name");
s = struct_find(v);
if (s) {
if (s->type.t != a)
tcc_error("invalid type");
goto do_decl;
} else if (tok >= TOK_IDENT && !tdef)
tcc_error("unknown struct/union/enum");
} else {
v = anon_sym++;
}
tcc-0.9.23 which don't have such error
/* enum/struct/union declaration. u is either VT_ENUM or VT_STRUCT */
static void struct_decl(CType *type, int u)
....
if (tok != '{') {
v = tok;
next();
/* struct already defined ? return it */
if (v < TOK_IDENT)
expect("struct/union/enum name");
s = struct_find(v);
if (s) {
if (s->type.t != a)
error("invalid type");
goto do_decl;
}
} else {
v = anon_sym++;
}
libtcc.c: Add greloca, a generalisation of greloc that takes an addend.
tcc.h: Add greloca and put_elf_reloca.
tccelf.c: Add put_elf_reloca, a generalisation of put_elf_reloc.
tccgen.c: On x86_64, use greloca instead of greloc in init_putv.
The back end functions gen_op(comparison) and gtst() might allocate
registers so case_reg should be left on the value stack while they
are called and set again afterwards.
This for example suppresses string constants such as with
int main()
{
return sizeof "foo";
}
Actually, setting
nocode_wanted = 1;
in libtcc.c for the initial global level seemed wrong, since
obviously "nocode_wanted" means code as any side effects, also
such as string constants.
This reverts a part of 2de1b2d14c
(documented as "Some in-between fixes" in Changelog)
This adds parsing of (GCC compatible) visibility attribute
in order to mark selected global symbols as hidden. The generated
.o files contain hidden symbols already, the TCC linker doesn't
yet do the right thing.
*** UNCONDITIONALLY ***
Esp. sihce tinycc winapi headers are not as complete as people might
expect this can otherwise lead to obscure problems that are difficult
to debug.
(Originally 'warn_implicit_function_declaration' was set to 1
always for windows but someone must have deleted that line)
This was going wrong (case TOK_LAND in unary: computed labels)
- vset(&s->type, VT_CONST | VT_SYM, 0);
- vtop->sym = s;
This does the right thing and is shorter:
+ vpushsym(&s->type, s);
Test case was:
int main(int argc, char **argv)
{
int x;
static void *label_return = &&lbl_return;
printf("label_return = %p\n", label_return);
goto *label_return; //<<<<< here segfault on linux X86_64 without the memset on vset
printf("unreachable\n");
lbl_return:
return 0;
}
Also::
- Rename "void* CValue.ptr" to more usable "addr_t ptr_offset"
and start to use it in obvious cases.
- use __attribute__ ((noreturn)) only with gnu compiler
- Revert CValue memsets ("After several days searching ...")
commit 4bc83ac393
Doesn't mean that the vsetX/vpush thingy isn't brittle and
there still might be bugs as to differences in how the CValue
union was set and is then interpreted later on.
However the big memset hammer was just too slow (-3% overall).
For program manipulating argv or arge as pointer with construct such as:
(while *argv++) {
do_something_with_argv;
}
it is necessary to have argv and arge inside a region. This patch create
regions argv and arge) if main is declared with those parameters.
I found the problem it was because CValue stack variables have rubish as it inital values
and assigning to a member that is smaller than the big union item and trying to
recover it later as a different member gives bak garbage.
ST_FUNC void vset(TCCState* tcc_state, CType *type, int r, int v)
{
CValue cval;
memset(&cval, 0, sizeof(CValue));
cval.i = v; //,<<<<<<<<<<< here is the main bug that mix with garbage
vsetc(tcc_state, type, r, &cval);
}
/* store a value or an expression directly in global data or in local array */
static void init_putv(TCCState* tcc_state, CType *type, Section *sec, unsigned long c,
int v, int expr_type)
{
...
case VT_PTR:
if (tcc_state->tccgen_vtop->r & VT_SYM) {
greloc(tcc_state, sec, tcc_state->tccgen_vtop->sym, c, R_DATA_PTR);
}
//<<< on the next line is where we try to get the assigned value to cvalue.i as cvalue.ull
*(addr_t *)ptr |= (tcc_state->tccgen_vtop->c.ull & bit_mask) << bit_pos;
break;
Also this patch makes vla tests pass on linux 32 bits
When checking for exact compatibility between types (such as in
__builtin_types_compatible_p) consider the case of default signedness to
be incompatible with both of the explicit signedness for char. That is,
char is incompatible with signed char *and* unsigned char, no matter
what the default signedness for char is.
negate(x) is subtract(-0,x), not subtract(+0,x), which makes
a difference with signed zeros. Also +x was expressed as x+0,
in order for the integer promotions to happen, but also mangles signed
zeros, so just don't do that with floating types.
Applying 64bit relocs assumes that the CVal is initialized to zero
for the whole 64bit. Consolidate this a bit, at the same time
zeroing the .ull member more consistently when needed. Fixes segfault
on x86_64-linux using global vars in tcctest.c.
Refactoring (no logical changes):
- use memcpy in tccgen.c:ieee_finite(double d)
- use union to store attribute flags in Sym
Makefile: "CFLAGS+=-fno-strict-aliasing" basically not necessary
anymore but I left it for now because gcc sometimes behaves
unexpectedly without.
Also:
- configure: back to mode 100755
- tcc.h: remove unused variables tdata/tbss_section
- x86_64-gen.c: adjust gfunc_sret for prototype
- tccgen: error out for cast to void, as in
void foo(void) { return 1; }
This avoids an assertion failure in x86_64-gen.c, also.
also fix tests2/03_struct.c accordingly
- Error: "memory full" - be more specific
- Makefiles: remove circular dependencies, lookup tcctest.c from VPATH
- tcc.h: cleanup lib, include, crt and libgcc search paths"
avoid duplication or trailing slashes with no CONFIG_MULTIARCHDIR
(as from 9382d6f1a0)
- tcc.h: remove ";{B}" from PE search path
in ce5e12c2f9 James Lyon wrote:
"... I'm not sure this is the right way to fix this problem."
And the answer is: No, please. (copying libtcc1.a for tests instead)
- win32/build_tcc.bat: do not move away a versioned file
The procedure calling standard for ARM architecture mandate the use of
the base standard for variadic function. Therefore, hgen float aggregate
must be returned via stack when greater than 4 bytes and via core
registers else in case of variadic function.
This patch improve gfunc_sret() to take into account whether the
function is variadic or not and make use of gfunc_sret() return value to
determine whether to pass a structure via stack in gfunc_prolog(). It
also take advantage of knowing if a function is variadic or not move
float result value from VFP register to core register in gfunc_epilog().
Move the logic to do a test of an integer value (ex if (0)) out of
arch-specific code to tccgen.c to avoid code duplication. This also
fixes test of long long value which was only testing the bottom half of
such values on 32 bits architectures.
- avoid assumption "ret_align == register_size" which is
false for non-arm targets
- rename symbol "sret" to more descriptive "ret_nregs"
This fixes commit dcec8673f2
Also:
- remove multiple definitions in win32/include/math.h
On ARM with hardfloat calling convention, structure containing 4 fields
or less of the same float type are returned via float registers. This
means that a structure can be returned in up to 4 double registers in a
structure is composed of 4 doubles. This commit adds support for return
of structures in several registers.
TLS support in tinyCC is absolutely not ready:
- segment register not select in load and store
- no relocation added for computing offset of per-thread symbol
- no support for TLS-specific relocations
- no program header added as per Drepper document about TLS
This reverts commit 1c4afd1350.
Prevent the following code from compiling:
enum color {RED, GREEN, BLUE};
enum color {R, G, B};
int main()
{
return R;
}
Reported-by: John Haque <j.eh@mchsi.com>
Use one more bit in AttributeDef to differenciate between declared
function (only its prototype is known) and defined function (its body is
also known). This allows to generate an error in cases like:
int f(){return 0;}
int f(){return 1;}
VLA storage is now freed when it goes out of scope. This makes it
possible to use a VLA inside a loop without consuming an unlimited
amount of memory.
Combining VLAs with alloca() should work as in GCC - when a VLA is
freed, memory allocated by alloca() after the VLA was created is also
freed. There are some exceptions to this rule when using goto: if a VLA
is in scope at the goto, jumping to a label will reset the stack pointer
to where it was immediately after the last VLA was created prior to the
label, or to what it was before the first VLA was created if the label
is outside the scope of any VLA. This means that in some cases combining
alloca() and VLAs will free alloca() memory where GCC would not.
Also made XMM0-7 available for use as temporary registers, since they
are not used by the ABI. I'd like to do the same with RSI and RDI but
that's trickier since they can be used by gv() as temporary registers
and there isn't a way to disable that.
All tests pass. I think I've caught all the cases assuming only XMM0 is
used. I expect that Win64 is horribly broken by this point though,
because I haven't altered it to cope with XMM1.
I've had to introduce the XMM1 register to get the calling convention
to work properly, unfortunately this has broken a fair bit of code
which assumes that only XMM0 is used.
There are probably still issues on x86-64 I've missed.
I've added a few new tests to abitest, which fail (2x long long and 2x double
in a struct should be passed in registers).
abitest now passes; however test1-3 fail in init_test. All other tests
pass. I need to re-test Win32 and Linux-x86.
I've added a dummy implementation of gfunc_sret to c67-gen.c so it
should now compile, and I think it should behave as before I created
gfunc_sret.
I expect that Linux-x86 is probably fine. All other architectures
except ARM are definitely broken since I haven't yet implemented
gfunc_sret for these, although replicating the current behaviour
should be straightforward.
Should fix some warnings wrt. access out of array bounds.
tccelf.c: fix "static function unused" warning
x86_64-gen.c: fix "ctype.ref uninitialzed" warning and cleanup
tcc-win32.txt: remove obsolete limitation notes.
This reverts commit 63193d1794.
Had some problems (_STATIC_ASSERT) and was too ugly anyway.
For retry, I'd suggest to implement a general function
static inline void memswap (void *p1, void* p2, size_t n);
and then use that. If you do so, please keep the original code
as comment.
This replaces commit 3d409b0889
- revert old fix in libtcc.c
- #include_next: look up the file in the include stack to see
if it is already included.
Also:
- streamline include code
- remove 'type' from struct CachedInclude (obsolete because we check
full filename anyway)
- remove inc_type & inc_filename from struct Bufferedfile (obsolete)
- fix bug with TOK_FLAG_ENDIF not being reset
- unrelated: get rid of an 'variable potentially uninitialized' warning
For vstack Fabrice used the trick to initialize vtop to &vstack[-1], so
that on first push, vtop becomes &vstack[0] and a value is also stored
there - everything works.
Except that when tcc is compiled with bounds-checking enabled, vstack - 1
returns INVALID_POINTER and oops...
Let's workaround it with artificial 1 vstack slot which will not be
used, but only serve as an indicator that pointing to &vstack[-1] is ok.
Now, tcc, after being self-compiled with -b works:
$ ./tcc -B. -o tccb -DONE_SOURCE -DCONFIG_MULTIARCHDIR=\"i386-linux-gnu\" tcc.c -ldl
$ cd tests
$ ../tcc -B.. -run tcctest.c >1
$ ../tccb -B.. -run tcctest.c >2
$ diff -u 1 2
and note, tcc's compilation speed is not affected:
$ ./tcc -B. -bench -DONE_SOURCE -DCONFIG_MULTIARCHDIR=\"i386-linux-gnu\" -c tcc.c
before: 8270 idents, 47221 lines, 1527730 bytes, 0.152 s, 309800 lines/s, 10.0 MB/s
after: 8271 idents, 47221 lines, 1527733 bytes, 0.152 s, 310107 lines/s, 10.0 MB/s
But note, that `tcc -b -run tcc` is still broken - for example it crashes
on
$ cat x.c
double get100 () { return 100.0; }
$ ./tcc -B. -b -DTCC_TARGET_I386 -DCONFIG_MULTIARCHDIR=\"i386-linux-gnu\" -run \
-DONE_SOURCE ./tcc.c -B. -c x.c
Runtime error: dereferencing invalid pointer
./tccpp.c:1953: at 0xa7beebdf parse_number() (included from ./libtcc.c, ./tcc.c)
./tccpp.c:3003: by 0xa7bf0708 next() (included from ./libtcc.c, ./tcc.c)
./tccgen.c:4465: by 0xa7bfe348 block() (included from ./libtcc.c, ./tcc.c)
./tccgen.c:4440: by 0xa7bfe212 block() (included from ./libtcc.c, ./tcc.c)
./tccgen.c:5529: by 0xa7c01929 gen_function() (included from ./libtcc.c, ./tcc.c)
./tccgen.c:5767: by 0xa7c02602 decl0() (included from ./libtcc.c, ./tcc.c)
that's because lib/bcheck.c runtime needs more fixes -- see next
patches.
Continuing d6072d37 (Add __builtin_frame_address(0)) implement
__builtin_frame_address for levels greater than zero, in order for
tinycc to be able to compile its own lib/bcheck.c after
cffb7af9 (lib/bcheck: Prevent __bound_local_new / __bound_local_delete
from being miscompiled).
I'm new to the internals, and used the most simple way to do it.
Generated code is not very good for levels >= 2, compare
gcc tcc
level=0 mov %ebp,%eax lea 0x0(%ebp),%eax
level=1 mov 0x0(%ebp),%eax mov 0x0(%ebp),%eax
level=2 mov 0x0(%ebp),%eax mov 0x0(%ebp),%eax
mov (%eax),%eax mov %eax,-0x10(%ebp)
mov -0x10(%ebp),%eax
mov (%eax),%eax
level=3 mov 0x0(%ebp),%eax mov 0x0(%ebp),%eax
mov (%eax),%eax mov (%eax),%ecx
mov (%eax),%eax mov (%ecx),%eax
But this is still an improvement and for bcheck we need level=1 for
which the code is good.
For the tests I had to force gcc use -O0 to not inline the functions.
And -fno-omit-frame-pointer just in case.
If someone knows how to improve the generated code - help is
appreciated.
Thanks,
Kirill
Cc: Michael Matz <matz@suse.de>
Cc: Shinichiro Hamaji <shinichiro.hamaji@gmail.com>
Currently, VLA are not forbidden for static variable. This leads to
problems even if for fixed-size array when the size expression uses the
ternary operator (cond ? then-value : else-value) because it is parsed
as a general expression which leads to code generated in this case.
This commit solve the problem by forbidding VLA for static variables.
Although not required for the fix, avoiding code generation when the
expression is constant would be a nice addition though.
The code for shifts is now similar to code for binary arithmetic operations,
except that only the first argument is considered, as required by the ISO C
standard.
On 2012-06-26 15:07:57 +0200, Vincent Lefevre wrote:
> ISO C99 TC3 says: [6.5.7#3] "The integer promotions are performed on
> each of the operands. The type of the result is that of the promoted
> left operand."
I've written a patch (attached). Now the shift problems no longer
occur with the testcase and with GNU MPFR's "make check".
--
Vincent Lefèvre <vincent@vinc17.net> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
Sometimes the result of a comparison is not directly used in a jump,
but in arithmetic or further comparisons. If those further things
do a vswap() with the VT_CMP as current top, and then generate
instructions for the new top, this most probably destroys the flags
(e.g. if it's a bitfield load like in the example).
vswap() must do the same like vsetc() and not allow VT_CMP vtops
to be moved down.
This matters when sizeof is directly used in arithmetic,
ala "uintptr_t t; t &= -sizeof(long)" (for alignment). When sizeof
isn't size_t (as it's specified to be) this masking will truncate
the high bits of the uintptr_t object (if uintptr_t is larger than
uint).
This needs to be accepted:
typedef int foo;
void f (void) { foo: return; }
namespaces for labels and types are different. The problem is that
the block parser always tries to find a decl first and that routine
doesn't peek enough to detect this case. Needs some adjustments
to unget_tok() so that we can call it even when we already called
it once, but next() didn't come around restoring the buffer yet.
(It lazily does so not when the buffer becomes empty, but rather
when the next call detects that the buffer is empty, i.e. it requires
two next() calls until the unget buffer gets switched back).
Removes a premature optimization of char/short loads
rewriting the source type. It did so also for bitfield
loads, thereby removing all the shifts/maskings.
(cond ? 0 : ptr)->member wasn't handled correctly. If one arm
is a null pointer constant (which also can be a pointer) the result
type is that of the other arm.
This fixes a bug introduced in commit
8d107d9ffd
that produced wrong code because of interference between
0x10 bits VT_CONST and x86_64-gen.c:TREG_MEM
Also fully zero-pad long doubles on x86-64 to avoid random
bytes in output files which disturb file comparison.
This allows passing colon separated paths to
tcc_add_library_path
tcc_add_sysinclude_path
tcc_add_include_path
Also there are new configure variables
CONFIG_TCC_LIBPATH
CONFIG_TCC_SYSINCLUDE_PATHS
which define the lib/sysinclude paths all in one and can
be overridden from configure/make
For TCC_TARGET_PE semicolons (;) are used as separators
Also, \b in the path string is replaced by s->tcc_lib_path
(CONFIG_TCCDIR rsp. -B option)
Since no code should be generated outside a function, force expr_cond to
only consider constant expression when outside a function since the
generic code can generate some code.
Basically, with:
typedef __attribute__((aligned(16))) struct _xyz {
...
} xyz, *pxyz;
we want the struct aligned but not the pointer.
FIXME: This patch is a hack, waiting for someone in the knowledge
of correct __attribute__ semantics.
VLA inserts a call to alloca via enum TOK_alloca, but TOK_alloca
only exists on I386 and X86_64 targets. This patch just emits an
error at compile-time if someone tries to compile some VLA code
for a TOK_alloca-less target. The best solution might be to just
push the problem to link-time, since the existence-or-not of a
alloca implementation can only be determined by linking. It seems
like just declaring TOK_alloca unconditionally would achieve that,
but for now, this at least gets the cross compilers to build.
I don't know if it makes a difference to gen_op(TOK_PDIV) or not,
but logically the ptr1_is_vla test in TP's VLA patch seems out of
order, where the patch to fix it would be:
------------------------------------------------------------------
@@ -1581,15 +1581,15 @@ ST_FUNC void gen_op(int op)
u = pointed_size(&vtop[-1].type);
}
gen_opic(op);
+ if (ptr1_is_vla)
+ vswap();
/* set to integer type */
#ifdef TCC_TARGET_X86_64
vtop->type.t = VT_LLONG;
#else
vtop->type.t = VT_INT;
#endif
- if (ptr1_is_vla)
- vswap();
- else
+ if (!ptr1_is_vla)
vpushi(u);
gen_op(TOK_PDIV);
} else {
------------------------------------------------------------------
Instead of that patch, which increases the complexity of the code,
this one fixes the problem by just rolling back and retrying with
a simpler approach.
A VLA is not really an array, it's a pointer-to-an-array.
Making this explicit allows us to back out a few parts
of the original VLA patch and paves the way for the next
part of the fix, where a VLA will be stored on the runtime
stack as a pointer-to-an-array, rather than on the compile-
time stack as a Sym*.
test target in Makefile does not depend on tcc.
i'm not sure why, but i can think of at least one
good reason. in my local tree I have it modified
to do so, but somehow inadvertently reverted that
so when i did "make test" before committing, it
didn't actually test my changes. sorry.
previously, tcc would accept a prototype of a function returning
an array, but not giving those functions bodies nor calling them.
it seems that gcc has never supported them, so we should probably
just error out... but it's possible that someone already using
tcc includes some header that contains an unused prototype for
one, so let's continue to support that.
Add support for asm labels for variables, that is the ability to rename
a variable at assembly level with __asm__ ("newname") appended in
its declaration.
See http://gcc.gnu.org/onlinedocs/gcc-4.4.4/gcc/Asm-Labels.html for more
details.
- Fix function assembly label mechanism introduced in commit
9b09fc376e to only accept alternative
name for function declaration.
- merge the code with the one introduced in commit
264a103610.
- Don't memorize token for asm label but directly the asm label.
Error out on static function without file scope and give an explaination
to the user
This is a rewrite of e9406c09a3 but
considering problems raised about static local function pointers in
632ee5a540.
* Disable C99 VLA detection when alloca is unavailable and protect the
new reference to TOK_alloca in decl_initializer in order to compile
and run for architecture without working alloca.
Not all code of C99 VLA is commented as it would required many ifdef
stanza. Just the detection is commented so that VT_VLA is never set
any type and the C99 VLA code is compiled but never called. However
vpush_global_sym(&func_old_type, TOK_alloca) in decl_initializer needs
to be protected by an ifdef stanza as well because it uses TOK_alloca.
* include alloca and C99 VLA tests according to availability of
TOK_alloca instead of relying on the current architecture
Implement C99 Variable Length Arrays in tinycc:
- Support VLA with multiple level (nested vla)
- Update documentation with regards to VT_VLA
- Add a testsuite in tcctest.c
- add __builtin_va_arg_types to check how arguments were passed
- move most code of stdarg into libtcc1.c
- remove __builtin_malloc and __builtin_free
- add a test case based on the bug report
(http://www.mail-archive.com/tinycc-devel@nongnu.org/msg03036.html)
Add support for asm labels for functions, that is the ability to rename
a function at assembly level with __asm__ ("newname") appended in
function declaration.
This reverts commit 433ecdfc9d.
The patch breaks e.g. with
for ((i = 10); --i;);
In particular to check for a type decl. this is not sufficient:
if (tok < TOK_UIDENT) {
A future approach to c99 loop variables might instead use:
if (parse_btype(...)) {
plus refactor function decl() accordingly.
Unary expression can start with a parenthesis. Thus, the current test
to detect which sizeof form is being parsed is inaccurate. This patch
makes tcc able to handle things like sizeof (x)[1] where x is declared
as char x[5]; wich is a valid unary expression
Error out with an explicit message when trying to initialize a
character array with something that's not a literal (optionally
enclosed in braces) as per C99 6.7.8:14; thanks to Antti-Juhani
Kaijanaho <ajk@debian.org> who did all the work.
When storing structs with a memcpy call in vstore(),
so far a needless entry remaining on the vstack
sometimes resulted in an useless store generated by
save_regs() in gfunc_call() for the memcpy routine.
Date: Mon, 8 Jun 2009 19:06:56 +0800
From: Soloist Deng <soloist.deng-gmail-com>
Subject: [Tinycc-devel] trying to fix the bug of unclean FPU st(0)
Hi all:
I am using tcc-0.9.25, and the FPU bug brought a big trouble to
me. I read the source and tried to fix it.
Below is my solution.
There are two places where program(`o(0xd9dd)') will generates `fstp
%st(1)': vpop() in tccgen.c:689 and save_reg() in tccgen.c:210.
We should first change both of them to `o(0xd8dd) // fstp %st(0)'.
But these changes are not enough. Let's check the following code.
void foo()
{
double var = 2.7;
var++;
}
Using the changed tcc will generate following machine code:
.text:08000000 public foo
.text:08000000 foo proc near
.text:08000000
.text:08000000 var_18 = qword ptr -18h
.text:08000000 var_10 = qword ptr -10h
.text:08000000 var_8 = qword ptr -8
.text:08000000
.text:08000000 push ebp
.text:08000001 mov ebp, esp
.text:08000003 sub esp, 18h
.text:08000009 nop
.text:0800000A fld L_0
.text:08000010 fst [ebp+var_8]
.text:08000013 fstp st(0)
.text:08000015 fld [ebp+var_8]
.text:08000018 fst [ebp+var_10]
.text:0800001B fstp st(0)
.text:0800001D fst [ebp+var_18]
.text:08000020 fstp st(0)
.text:08000022 fld L_1
.text:08000028 fadd [ebp+var_10]
.text:0800002B fst [ebp+var_8]
.text:0800002E fstp st(0)
.text:08000030 leave
.text:08000031 retn
.text:08000031 foo endp
.text:08000031
.text:08000031 _text ends
--------------------------------------------------
.data:08000040 ; Segment type: Pure data
.data:08000040 ; Segment permissions: Read/Write
.data:08000040 ; Segment alignment '32byte' can not be represented in assembly
.data:08000040 _data segment page public 'DATA' use32
.data:08000040 assume cs:_data
.data:08000040 ;org 8000040h
.data:08000040 L_0 dq 400599999999999Ah
.data:08000048 L_1 dq 3FF0000000000000h
.data:08000048 _data ends
Please notice the code snippet from 0800000A to 08000020
// double var = 2.7; load constant to st(0)
.text:0800000A fld L_0
// double var = 2.7; store st(0) to `var'
.text:08000010 fst [ebp+var_8]
// double var = 2.7; poping st(0) will empty the floating registers stack
.text:08000013 fstp st(0)
After that ,tcc will call `void inc(int post, int c)" in
tccgen.c:2150, and produce 08000015 to 0800001B through the calling
chain (inc ->gv_dup)
// load from `var' to st(0)
.text:08000015 fld [ebp+var_8]
// store st(0) to a temporary location
.text:08000018 fst [ebp+var_10]
// poping st(0) will empty the floating registers stack
.text:0800001B fstp st(0)
And the calling chain
(gen_op('+')->gen_opif('+')->gen_opf('+')->gv(rc=2)->get_reg(rc=2)->save_reg(r=3))
will produce 0800001D to 08000020 .
// store st(0) to a temporary location, but floating stack is empty!
.text:0800001D fst [ebp+var_18]
// poping st(0) will empty the floating registers stack
.text:08000020 fstp st(0)
The `0800001D fst [ebp+var_18]' will store st(0) to a memory
location, but st(0) is empty. That will cause FPU invalid operation
exception(#IE).
Why does tcc do that? Please read `gv_dup' called by `inc' carefully.
Notice these lines:
(1): r = gv(rc);
(2): r1 = get_reg(rc);
(3): sv.r = r;
sv.c.ul = 0;
(4) load(r1, &sv); /* move r to r1 */
(5) vdup();
/* duplicates value */
(6) vtop->r = r1;
(1) let the vtop occupy TREG_ST0, and `r' will be TREG_ST0. (2)
try to get a free floating register,but tcc assume
there is only one, so it wil force vtop goto memory and assign `r1'
with TREG_ST0. When executing (3), it will do nothing
because `r' equals `r1'. (5) duplicates vtop. Then (6) let the new
vtop occupy TREG_ST0, but this will cause problem
because the old vtop has been moved to memory, so the new duplicated
vtop does not reside in TREG_ST0 but also
in memory after that. TREG_ST0 is not occupied but freely availabe
now. `gen_op('+')' need at least one oprand in register,
so it will incorrectly think TREG_ST0 is occupied by vtop and produce
instructions(0800001D and 08000020) to store it to
a temporary memory location.
According program above, if `r' == `r1' it is impossible for the old
vtop to still occupy the `r' register . And `load' will do nothing
too at this condition.
So the `gv_dup' can not promise the semantics that old vtop in one
register and the new duplicated vtop in another register at the same
time.
I changed (6) to
if (r != r1)
{
vtop->r = r1;
}
Then the new generated machine code will be :
.text:08000000 push ebp
.text:08000001 mov ebp, esp
.text:08000003 sub esp, 10h
.text:08000009 nop
.text:0800000A fld L_0
.text:08000010 fst [ebp+var_8]
.text:08000013 fstp st(0)
.text:08000015 fld [ebp+var_8]
.text:08000018 fst [ebp+var_10]
.text:0800001B fstp st(0)
.text:0800001D fld L_1
.text:08000023 fadd [ebp+var_10]
.text:08000026 fst [ebp+var_8]
.text:08000029 fstp st(0)
.text:0800002B leave
.text:0800002C retn
It works well, and will clean the floating registers stack when return.
Finally, I want to know there is any potential problem of this fixing ?
soloist
