Commit Graph

2647 Commits

Author SHA1 Message Date
Michael Matz
53c5fc2246 x86-64: Rewrite linux parameter passing
This fixes two ABI testcases involving large arguments when there
are still registers available for later args.
2017-05-27 21:23:13 +02:00
Aron BARATH
307b7b183d the R_X86_64_GOTOFF64 relocation was missing 2017-05-16 07:03:26 +02:00
grischka
28435ec58c configure: --config-musl/-uClibc switch & misc cleanups
- 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
2017-05-13 08:59:06 +02:00
Aron BARATH
3a9d6b3655 added 64-bit relocation types that required to linking against 64-bit (large model) libraries 2017-05-12 09:31:09 +02:00
grischka
1ed20a01c9 bitfields: promote to signed int
For integer promotion with for example arithmetics or
expr_cond (x ? y : z), integral types need to be promoted
to signed if they fit.

According to latest standards, this also applies to bit-field
types taking into account their specific width.

In tcc, VT_BITFIELD set means width < original type width
Field-widths between 33 and 63 are promoted to signed long long
accordingly.

    struct { unsigned long long ullb:35; } s = { 1 };
    #define X (s.ullb - 2)

    int main (void)
    {
        long long Y = X;
        printf("%d %016llx %016llx\n", X < 0, -X, -Y);
        return 0;
    }

Results:
    GCC 4.7 : 0 0000000000000001 FFFFFFF800000001
    MSVC    : 1 0000000000000001 0000000000000001
    TCC     : 1 0000000000000001 0000000000000001

Also, gcc would promote long long bitfields of size < 32
to int as well.  Example:

    struct { unsigned long long x:20; } t = { 123 };
    /* with gcc: */ printf("%d %d\n", t.x, 456);
    /* with tcc: */ printf("%lld %d\n", t.x, 456);
2017-05-09 18:36:24 +02:00
grischka
d242706f3b bitfields: one more hack and a "-Wgcc-compat" switch
bit_pos + bit_size > type_size * 8

must NEVER happen because the code generator can read/write
only the basic integral types.

Warn if tcc has to break GCC compatibility for that reason
and if -Wgcc-compat is given.

Example:
    struct __attribute__((packed)) _s
    {
        unsigned int x  : 12;
        unsigned char y :  7;
        unsigned int z  : 28;
    };

Expected (GCC) layout (sizeof struct = 6)
.xxxxxxxx.xxxxyyyy.yyyzzzzz.zzzzzzzz.zzzzzzzz.zzzzzzz0.

But we cannot read/write 'char y'from 2 bytes in memory.
So we have to adjust:
.xxxxxxxx.xxxx0000.yyyyyyyz.zzzzzzzz.zzzzzzzz.zzzzzzzz.zzz00000

Now 'int z' cannot be accessed from 5 bytes.  So we arrive
at this (sizeof struct = 7):
.xxxxxxxx.xxxx0000.yyyyyyy0.zzzzzzzz.zzzzzzzz.zzzzzzzz.zzzz0000

Otherwise the bitfield load/store generator needs to be
changed to allow byte-wise accesses.

Also we may touch memory past the struct in some cases
currently.  The patch adds a warning for that too.

   0:	55                   	push   %ebp
   1:	89 e5                	mov    %esp,%ebp
   3:	81 ec 04 00 00 00    	sub    $0x4,%esp
   9:	90                   	nop

    struct __attribute__((packed)) { unsigned x : 5; } b = {0} ;

   a:	8b 45 ff             	mov    -0x1(%ebp),%eax
   d:	83 e0 e0             	and    $0xffffffe0,%eax
  10:	89 45 ff             	mov    %eax,-0x1(%ebp)

This touches -0x1 ... +0x3(%ebp), hence 3 bytes beyond
stack space.  Since the data is not changed, nothing
else happens here.
2017-05-09 18:10:02 +02:00
Larry Doolittle
3f5fd305af Touch up previous "s"[i<2] patch
The real equivalence is between the original "s" + (i<2)
and &"s"[i<2].
2017-05-09 07:08:43 -07:00
Larry Doolittle
3c0a73822d Give clang one less thing to complain about
"s"[i<2] and "s" + (i<2) are literally identical, but the latter triggers
a warning from clang because it looks so much like a noob is trying to
concatenate an integer and a string.  The former is arguably more clear.
2017-05-08 21:56:38 -07:00
Michael Matz
dfb75d9b6c Remove some unused arguments
these ones are really superfluous.
2017-05-08 20:26:59 +02:00
Michael Matz
377e8e5e68 bitfields: fix long bitfields
now the testcase works on i386-linux as well.
2017-05-08 19:58:31 +02:00
Marc Vertes
1094891e67 Improve musl and uclibc detection
Do not enable musl or uclibc native support if a GNU linker is
already present. This avoids interference for example on a
Debian platform with musl-dev installed. More work is required
to select musl libc in that case, with additional configure flags.
2017-05-08 16:14:35 +00:00
Larry Doolittle
70fca7f799 Fix spelling in help message
"seach" becomes "search" in help2 when not TCC_TARGET_PE
2017-05-08 08:44:52 -07:00
Larry Doolittle
7443db0d5f Remove some unused-parameter lint
Mark TCCState parameter as unused in tcc_undefine_symbol(), tcc_add_symbol(),
tcc_print_stats(), asm_get_local_label_name(), use_section1(), tccpp_delete(),
tcc_tool_ar(), tcc_tool_impdef(), and tcc_tool_cross().
Also mark it unused in tcc_add_bcheck() unless CONFIG_TCC_BCHECK.
Remove it entirely in ld_next().
2017-05-08 08:30:43 -07:00
Christian Jullien
56df27502c C string litteral is const, fix return type of default_elfinterp. Remove #ifndef TCC_IS_NATIVE in arm-gen.c as suggested by grischka. 2017-05-08 09:18:27 +02:00
Larry Doolittle
19d8b8a173 Spelling fixes in C comments only 2017-05-07 21:38:09 -07:00
Larry Doolittle
1b9935cf36 Add missing const and add warning flags 2017-05-07 21:28:05 -07:00
Michael Matz
524f6dff17 Fix a warning
"missing intitializer for field op_type" with gcc -Wextra.  It's
zero-initialized anyway, and not that much a hassle to explicitely
initialize either, so let's be nice and make it not warn.
2017-05-08 05:29:54 +02:00
grischka
44abffe33a more minor fixes
* tccgen: re-allow long double constants for x87 cross
  sizeof (long double) may be 12 or 16 depending on host platform
  (i386/x86_64 on unix/windows).
  Except that it's 8 if the host is on windows and not gcc
  was used to compile tcc.

* win64: fix builtin_va_start after VT_REF removal
  See also a8b83ce43a

* tcctest.c: remove outdated limitation for ll-bitfield test
  It always worked, there is no reason why it should not work
  in future.

* libtcc1.c: exclude long double conversion on ARM

* Makefile: remove CFLAGS from link recipes

* lib/Makefile: use target DEFINES as passed from main Makefile

* lib/armflush.c lib/va_list.c: factor out from libtcc1.c

* arm-gen.c: disable "depreciated" warnings for now
2017-05-07 12:41:29 +02:00
Michael Matz
94ac9f2b49 Accept extern initialized file-scope variables
'extern int i = 42;' at file scope (but not in function scope!) is
allowed and is a proper definition, even though questionable style;
some compilers warn about this.
2017-05-07 08:10:06 +02:00
Michael Matz
5732a1882e elf: Fix last commit
Forgot to mark tccelf.c as to be committed in last commit :-/
2017-05-07 04:52:21 +02:00
Michael Matz
680e84fe42 elf: Support STB_LOCAL dynamic symbols
local symbols can be resolved statically, they don't have to be
done dynamically, so this is a slight speedup at load time for
produced executables and shared libs.  The musl libc also rejects
any STB_LOCAL symbols for dynamic symbol resolution, so there it
also fixes use of shared libs created by tcc.
2017-05-07 04:41:40 +02:00
Michael Matz
600018ce47 elf: Ignore SHF_COMPRESSED sections
some newer systems have debug sections compressed by default, which
includes those in the crt[1in].o startup files.  These can't simply
be concatenated like all others (which leads to invalid section contents
ultimately making gdb fail) but need special handling.

Instead of that special handling (decompressing, which in turn requires
linking against zlib) let's just ignore such sections, even though that
means to also ignore all other debug sections from that particular input
file.  Our own generated files of course don't have the problem.
2017-05-06 07:30:44 +02:00
Michael Matz
ff998900b1 struct-init: Fix zero initialization with multi-level designators
See the added testcase.  When one used designators like .a.x to initialize
sub-members of members, and didn't then initialize all of them the
required zero-initialization of the other sub-members wasn't done.
The fix also enables tiny code cleanups.
2017-05-06 05:28:13 +02:00
Michael Matz
0757234560 Fix unsigned enum bit-fields
See testcase.  If an enum has only positive values, fits N bits,
and is placed in a N-bit bit-field that bit-fields must be treated
as unsigned, not signed.
2017-05-05 23:16:43 +02:00
Michael Matz
a7a3627ab2 Fix segfault with invalid function def
This invalid function definition:
  int f()[] {}
was tried to be handled but there was no testcase if it actually worked.
This fixes it and adds a TCC only testcase.
2017-05-05 22:01:02 +02:00
Michael Matz
d4878985e8 Rebuild cross compilers when sources change
ONE_SOURCE=yes cross-compilers currently only depend on tcc.c, which
itself has no further deps.  So e.g. changing tccgen.c or tcctok.h
don't automatically rebuild cross compilers.  Let's go over
the intermediate $(X)tcc.o file which automatically depends on
LIBTCC_INC, which are all relevant source files.
2017-05-02 03:14:05 +02:00
Michael Matz
149678e888 Tidy decl_designator
Removing the need for the notfirst local variable, and tidy loop structure
and error messaging a bit.
2017-05-02 03:14:05 +02:00
Michael Matz
9839b60177 Remove label_or_decl
The only use of this function can be rewritten in terms of
is_label (if one other use of that one are a bit amended).
2017-05-02 03:14:05 +02:00
Michael Matz
c7dbc900c8 Cleanups (float consts, sections, symbols)
introduce common_section (SHN_COMMON), factorize some handling
in decl_initializer_alloc, add section_add and use it to factorize
some code that allocates stuff in sections (at the same time also fixing
harmless bugs re section alignment), use init_putv to emit float consts
into .data from gv() (fixing an XXX).
2017-05-02 03:13:55 +02:00
Michael Matz
25522e4799 Merge func_decl_list into decl0
Removes some code duplication and also implements one feature:
checking for duplicate decls for old style parameters.
2017-05-02 03:07:37 +02:00
Michael Matz
7aef0522b0 Tidy decl_designator
The fixme therein is long solved.
2017-05-02 03:07:37 +02:00
Michael Matz
8ca98e23c4 Tidy unary() a bit
factor code a bit for transforming tokens into SValues.  This revealed
a bug in TOK_GET (see testcase), which happened to be harmless before.
So fix that as well.
2017-05-02 03:07:37 +02:00
Michael Matz
21b12ea10d Factor some code
Three places that skip (and store) tokens in some fashion can
be factored a bit.
2017-05-02 03:07:37 +02:00
Michael Matz
328b06a3fc Extend type_to_str
to also print storage-class specifiers.
2017-05-02 03:07:36 +02:00
Michael Matz
182367e232 Reorganize type parsing
Various corner cases for declarator parsing were incorrect.  This
reorganizes and fixes it, and somewhat simplifies it as well.
2017-05-02 03:07:36 +02:00
Michael Matz
5891fbc0c8 Tidy typename parsing a bit 2017-05-02 03:07:36 +02:00
Michael Matz
51314932e3 Tidy arg parsing for builtins
Saves some lines of code.
2017-05-02 03:07:36 +02:00
Michael Matz
10551d6961 Clarify that the CIL target code is dead
This backend doesn't work anymore since a long time.
Don't remove it yet, it might be an inspiration.
2017-05-02 03:07:36 +02:00
Michael Matz
a8b83ce43a Remove VT_REF
The canonical way to describe a local variable that actually holds
the address of an lvalue is VT_LLOCAL.  Remove the last user of VT_REF,
and handling of it, thereby freeing a flag for SValue.r.
2017-05-02 03:07:36 +02:00
Michael Matz
8b9697ca6b Fix bogus check for VT_LLOCAL types
VT_LLOCAL is a flag on .r, not on type.t.  Fixing this requires
minor surgery for compound literals which accidentally happened
to be subsumed by the bogus test.
2017-05-02 03:07:36 +02:00
Michael Matz
24420bb5c0 elf: Tidy section headers
Don't emit useless section headers and also sort them in allocated
order.  Doesn't change behaviour except makes the resulting files
a tiny bit smaller (though at the expense of some very tiny compile
time and code size increase of tcc itself; not 100% it's worth it).
2017-05-02 03:07:36 +02:00
Michael Matz
3c39cb5cd8 fix __builtin_expect
the second argument can be an arbitrary expression (including
side-effects), not just a constant.  This removes the last user
of expr_lor_const and hence also that function (and expr_land_const).
Also the argument to __builtin_constant_p can be only a non-comma
expression (like all functions arguments).
2017-05-02 03:07:36 +02:00
Michael Matz
8d9dd3c008 Fix more bitfield corner cases
Our code generation assumes that it can load/store with the
bit-fields base type, so bit_pos/bit_size must be in range for this.
We could change the fields type or adjust offset/bit_pos; we do the
latter.
2017-05-01 06:18:48 +02:00
Michael Matz
4ce73354fc Fix last change
Skipping anonymous bit-fields is correct, but not other
anonymous ones like unions or structs.
2017-05-01 06:04:19 +02:00
Michael Matz
f775d68725 Remove a bit-field TODO
Checked the lcc testsuite for bitfield stuff (in cq.c and fields.c),
fixed one more error in initializing unnamed members (which have
to be skipped), removed the TODO.
2017-04-29 22:09:10 +02:00
Michael Matz
28084420fe Fix char bitfields corner case
See testcase.
2017-04-29 21:25:31 +02:00
Thomas Stalder
76e16465bf arm: Fix build_got_entries 2017-04-27 14:16:49 +02:00
Christian Jullien
f12851e388 Add -static to be in sync with Windows bat. 2017-04-27 07:01:46 +02:00
grischka
a35752d233 configure: fix permissions
There must be something with git on windows that it
messes up x permissions sometimes.
2017-04-26 00:01:23 +02:00
Larry Doolittle
37999a4fbf Simple-minded fix for bug #50847
bug #50847: #line directive corrupts #include search path
Keep a second copy of the filename, that isn't changed by the #line directive,
and use that on the initial search path for #include files.
2017-04-25 12:32:41 -07:00