add some features for more complete 'long' support
tcc.h:
- use LONG_SIZE=4/8 instead of TCC_LONG_ARE_64_BIT
tccgen.c:
- add ptrdiff_type, update size_type
- support shift and ?: operations
- support long enum types
- display 'long' from type_to_str
- nwchar_t is unsigned short on windows
- unrelated: use memcpy in init_putv for long doubles to avoid
random bytes in the image (if tcc was compiled by gcc) for
diff purposes.
tccpp.c:
- make parse_number return correct types
- improve multi-character-constants 'XX' 'abcd'
Changelog:
- update
tccgen.c:
doubles need to be aligned, on ARM. The section_reserve()
in init_putv does not do that.
-D ONE_SOURCE: is now the default and not longer needed. Also,
tcc.h now sets the default native target. These both make
compiling tcc simple as "gcc tcc.c -o tcc -ldl" again.
arm-asm.c:
enable pseudo asm also for inline asm
tests/tests2/Makefile:
disable bitfield tests except on windows and x86_64
and don't generate-always
tcc.c:
fix a loop with -dt on errors
configure:
print compiler version (as recognized)
tccpp.c:
actually define symbols for tcc -dt
clear static variables (needed for -dt or libtcc usage)
96_nodata_wanted.c:
use __label__ instead of asm
lib/files:
use native symbols (__i386__ etc.) instead of TCC_TARGET_...
* check that _Generic don't match unsigned char * with char *
this case is usefull as with -funsigned-char, 'char *' are unsigned
* change VT_LONG so it's now a qualifier
VT_LONG are never use for code generation, but only durring parsing state,
in _Generic we need to be able to make diference between
'long' and 'long long'
So VT_LONG is now use as a type qualifier, it's old behaviour is still
here, but we can keep trace of what was a long and what wasn't
* add TOK_CLONG and TOK_CULONG
tcc was directly converting value like '7171L' into TOK_CLLONG or
TOK_CINT depending of the machine architecture.
because of that, we was unable to make diference between a long and a
long long, which doesn't work with _Generic.
So now 7171L is a TOK_CLONG, and we can handle _Generic properly
* check that _Generic can make diference between long and long long
* uncomment "type match twice" as it should now pass tests on any platforms
* add inside_generic global
the point of this variable is to use VT_LONG in comparaison only
when we are evaluating a _Generic.
problem is with my lastest patchs tcc can now make the diference between
a 'long long' and a 'long', but in 64 bit stddef.h typedef uint64_t as
typedef signed long long int int64_t and stdint.h as unsigned long int, so tcc
break when stdint.h and stddef.h are include together.
Another solution woud be to modifie include/stddef.h so it define uint64_t as
unsigned long int when processor is 64 bit, but this could break some
legacy code, so for now, VT_LONG are use only inside generc.
* check that _Generic parse first argument correctly
* check that _Generic evaluate correctly exresion like "f() / 2"
* -dt now with lowercase t
* test snippets now separated by real preprocessor statements
which is valid C also for other compilers
#if defined test_xxx
< test snippet x >
#elif defined test_yyy
< test snippet y >
#elif ...
#endif
* simpler implementation, behaves like -run if no 'test_...' macros
are seen, works with -E too
* for demonstration I combined some of the small tests for errors
and warnings (56..63,74) in "60_errors_and_warnings.c"
Also:
* libtcc.c:
put tcc_preprocess() and tcc_assemble() under the setjmp clause
to let them return to caller after errors. This is for -dt -E.
* tccgen.c:
- get rid of save/restore_parse_state(), macro_ptr is saved
by begin_macro anyway, now line_num too.
- use expr_eq for parsing _Generic's controlling_type
- set nocode_wanted with const_wanted. too, This is to keep
VT_JMP on vtop when parsing preprocessor expressions.
* tccpp.c: tcc -E: suppress trailing whitespace from lines with
comments (that -E removes) such as
NO_GOTPLT_ENTRY,\t /* never generate ... */
The existing variable 'nocode_wanted' is now used to control
output of static data too. So...
(nocode_wanted == 0)
code and data (normal within functions)
(nocode_wanted < 0)
means: no code, but data (global or static data)
(nocode_wanted > 0)
means: no code and no data (code and data suppressed)
(nocode_wanted & 0xC0000000)
means: we're in declaration of static data
Also: new option '-dT' to be used with -run
tcc -dT -run file.c
This will look in file.c for certain comment-boundaries:
/*-* test-xxx: ...some description */
and then for each test below run it from memory. This way
various features and error messages can be tested with one
single file. See 96_nodata_wanted.c for an example.
Also: tccgen.c: one more bitfield fix
tccpp.c:
* #pragma comment(option,"-some-option")
to set commandline option from C code. May work only
for some options.
libtcc.c:
* option "-d1..9": sets a 'g_debug' global variable.
(for development)
tests2/Makefile:
* new make targets: tests2.37 / tests2.37+
run single test from tests2, optionally update .expect
* new variable GEN-ALWAYS to always generate certain .expects
* bitfields test
tccgen.c:
* bitfields: fix a bug and improve slightly more
* _Generic: ignore "type match twice"
Like returned local variables also labels local to a statement expression
can be returned, and so their symbols must not be immediately freed
(though they need to be removed from the symbol table).
Use 2 level strategy to access packed bitfields cleanly:
1) Allow to override the original declaration type with
an auxilary "access type". This solves cases such as
struct {
...
unsigned f1:1;
};
by using VT_BYTE to access f1.
2) Allow byte-wise split accesses using two new functions
load/store_packed_bf. This solves any cases, also ones
such as
struct __attribute((packed)) _s {
unsigned x : 12;
unsigned char y : 7;
unsigned z : 28;
unsigned a: 3;
unsigned b: 3;
unsigned c: 3;
};
where for field 'z':
- VT_INT access from offset 2 would be unaligned
- VT_LLONG from offset 0 would go past the total
struct size (7)
and for field 'a' because it is in two bytes and
aligned access with VT_SHORT/INT is not possible.
Also, static bitfield initializers are stored byte-wise always.
Also, cleanup the struct_layout function a bit.
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
* use aarch64 instead of arm64
- Makefile
* rename the custom include file to "config-extra.mak"
* Also avoid "rm -r /*" if $(tccdir) is empty
- pp/Makefile
* fix .expect generation with gcc
- tcc.h
* cleanup #defines for _MSC_VER
- tccgen.c:
* fix const-propagation for &,|
* fix anonymous named struct (ms-extension) and enable
-fms-extension by default
- i386-gen.c
* clear VT_DEFSIGN
- x86_64-gen.c/win64:
* fix passing structs in registers
* fix alloca (need to keep "func_scratch" below each alloca area on stack)
(This allows to compile a working gnu-make on win64)
- tccpp.c
* alternative approach to 37999a4fbf
This is to avoid some slowdown with ## token pasting.
* get_tok_str() : return <eof> for TOK_EOF
* -funsigned-char: apply to "string" literals as well
- tccpe/tools.c: -impdef: support both 32 and 64 bit dlls anyway
That means: we do not macro-expand the argument of 'defined'
Also: store the last token position into the TokenString
structure in order to get rid of the tok_last function.
- 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
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.
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).
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.
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.
The O(xxx) stuff in i386-asm.c had me scratching my head. Extracting
the macro and trying it out in a separate program doesn't give
me any warnings, so I'm confused about what could be going on there.
Any cast will make things happy. I used a uint64_t to catch actual
cases of overflow, which will still cause a -Wconstant-conversion
warning.
Signed-off-by: Andrei Warkentin <andrey.warkentin@gmail.com>
Also, retain storage qualifiers in type_decl, in particular
also for function pointers. This allows to get rid of this
very early hack in decl()
type.t |= (btype.t & VT_STATIC); /* Retain "static". */
which was to fix the case of
int main() { static int (*foo)(); ...
Also:
- missing __declspec(dllimport) is an error now
- except if the symbol is "_imp__symbol"
- demonstrate export/import of data in the dll example (while
'extern' isn't strictly required with dllimport anymore)
- new function 'patch_storage()' replaces 'weaken_symbol()'
and 'apply_visibility()'
- new function 'update_storage()' applies storage attributes
to Elf symbols.
- put_extern_sym/2 accepts new pseudo section SECTION_COMMON
- add -Wl,-export-all-symbols as alias for -rdynamic
- add -Wl,-subsystem=windows for mingw compatibility
- redefinition of 'sym' error for initialized global data
supports building cross compilers on the fly without need
for configure --enable-cross
$ make cross # all compilers
$ make cross-TARGET # only TARGET-compiler & its libtcc1.a
with TARGET one from
i386 x86_64 i386-win32 x86_64-win32 arm arm64 arm-wince c67
Type 'make help' for more information
since configure supports only native configuration
a file 'cross-tcc.mak' needs to be created manually.
It is included in the Makefile if present.
# ----------------------------------------------------
# Example config-cross.mak:
#
# windows -> i386-linux cross-compiler
# (it expects the linux files in <prefix>/i386-linux)
ROOT-i386 = {B}/i386-linux
CRT-i386 = $(ROOT-i386)/usr/lib
LIB-i386 = $(ROOT-i386)/lib:$(ROOT-i386)/usr/lib
INC-i386 = {B}/lib/include:$(ROOT-i386)/usr/include
DEF-i386 += -D__linux__
# ----------------------------------------------------
Also:
- use libtcc1-<target>.a instead of directories
- add dummy arm assembler
- remove include dependencies from armeabi.c/lib-arm64.c
- tccelf/ld_add_file: add SYSROOT (when defined) to absolute
filenames coming from ld-scripts
Forgot about it. It allows to compile several
sources (and other .o's) to one single .o file;
tcc -r -o all.o f1.c f2.c f3.S o4.o ...
Also:
- option -fold-struct-init-code removed, no effect anymore
- (tcc_)set_environment() moved to tcc.c
- win32/lib/(win)crt1 minor fix & add dependency
- debug line output for asm (tcc -c -g xxx.S) enabled
- configure/Makefiles: x86-64 -> x86_64 changes
- README: cleanup
usage:
tcc -ar [rcsv] lib files...
tcc -impdef lib.dll [-v] [-o lib.def]
also:
- support more files with -c: tcc -c f1.c f2.c ...
- fix a bug which caused tcc f1.c f2.S to produce no asm
- allow tcc -ar @listfile too
- change prototype: _void_ tcc_set_options(...)
- apply -Wl,-whole-archive when a librariy is given
as libxxx.a also (not just for -lxxx)
Today by accident i had to deal with linker problems of some
software and found an issue that mentioned DT_RUNPATH, which
mentioned that DT_RPATH is legacy and searched for
$LD_LIBRARY_PATH, whereas the newer DT_RUNPATH is searched
thereafter. Completely unencrypted! Well. For what's it worth,
i for one am astonished because of course i want to override
$LD_LIBRARY_PATH, but it surely has its merites, smart people came
to the conclusion, did they.
The attached diff below seems to be sufficient to support
DT_RUNPATH instead of DT_RPATH with tcc(1). But i have no insight
in what --enable-new-dtags is supposed to change in addition, so
i wonder.
Ciao!
--steffen
libtcc.c | 2 ++
tcc-doc.texi | 4 ++++
tcc.h | 1 +
tccelf.c | 3 ++-
4 files changed, 9 insertions(+), 1 deletion(-)
- tccgen.c/tcc.h: allow function declaration after use:
int f() { return g(); }
int g() { return 1; }
may be a warning but not an error
see also 76cb1144ef
- tccgen.c: redundant code related to inline functions removed
(functions used anywhere have sym->c set automatically)
- tccgen.c: make 32bit llop non-equal test portable
(probably not on C67)
- dynarray_add: change prototype to possibly avoid aliasing
problems or at least warnings
- lib/alloca*.S: ".section .note.GNU-stack,"",%progbits" removed
(has no effect)
- tccpe: set SizeOfCode field (for correct upx decompression)
- libtcc.c: fixed alternative -run invocation
tcc "-run -lxxx ..." file.c
(meant to load the library after file).
Also supported now:
tcc files ... options ... -run @ arguments ...
- we're now exporting tcc_prefixed symbols from libtcc only
- On windows, the msvcrt startup code would remove backslashes
from commandline arguments such as
-DFOO=\"foo\"
which would appear in argv as
-DFOO="foo"
Therefor before passing these to spawnvp, we need to restore
the backslashes.
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
tests/Makefile: fix out-of-tree build issues
Also:
- win64: align(16) MEM_DEBUG user memory
on win64 the struct jmp_buf in the TCCState structure which we
allocate by tcc_malloc needs alignment 16 because the msvcrt
setjmp uses MMX instructions.
- libtcc_test.c: win32/64 need __attribute__((dllimport)) for
extern data objects
- tcctest.c: exclude stuff that gcc does not compile
except for relocation_test() the other issues are mostly ASM
related. We should probably check GCC versions but I have
no idea which mingw/gcc versions support what and which don't.
- lib/Makefile: use tcc to compile libtcc1.a (except on arm
which needs arm-asm
handle mms-bitfields as sub-options of -m. (-mfloat-abi
is still special because it requires arguments)
tcc.c: help():
- list -mms-bitfields under 'Target specific options'
libtcc.c/MEM_DEBUG
- add check for past buffer writes
Also ...
tcctest.c:
- exclude stuff that gcc doesn't compile on windows.
libtcc.c/tccpp.c:
- use unsigned for memory sizes to avoid printf format warnings
- use "file:line: message" to make IDE error parsers happy.
tccgen.c: fix typo
The linux kernel has some structures that are page aligned,
i.e. 4096. Instead of enlarging the bit fields to specify this,
use the fact that alignment is always power of two, and store only
the log2 minus 1 of it. The 5 bits are enough to specify an alignment
of 1 << 30.
See testcases. We now support 64bit case constants. At the same time
also 64bit enum constants on L64 platforms (otherwise the Sym struct
isn't large enough for now). The testcase also checks for various
cases where sign/zero extension was confused.
See testcase. We must always paste tokens (at least if not
currently substing a normal argument, which is a speed optimization
only now) but at the same time must not regard a ## token
coming from argument expansion as the token-paste operator, nor
if we constructed a ## token due to pasting itself (that was already
checked by pp/01.c).
When intializing members where the initializer needs relocations
and the member is initialized multiple times we can't allow
that to lead to multiple relocations to the same place. The last
one must win.
Similar to GCC a local asm register variable enforces the use of a
specified register in asm operands (and doesn't otherwise
matter). Works only if the variable is directly mentioned as
operand. For that we now generally store a backpointer from
an SValue to a Sym when the SValue was the result of unary()
parsing a symbol identifier.
Some routines were using the wrong type (int) in passing addends,
truncating it. This matters when bit 31 isn't set and the high
32 bits are set: the truncation would make it unsigned where in
reality it's signed (happen e.g. on the x86-64 with it's load
address at top-2GB).
This happens when e.g. string constants (or other static data)
are passed as operands to inline asm as immediates. The produced
symbol ref wouldn't be found. So tighten the connection between
C and asm-local symbol table even more.
Our preprocessor throws away # line-comments in asm mode.
It did so also inside preprocessor directives, thereby
removing stringification. Parse defines in non-asm mode (but
retain '.' as identifier character inside macro definitions).
The return value of statement expressions might refer to local
symbols, so those can't be popped. The old error message always
was just a band-aid, and since disabling it for pointer types it
wasn't effective anyway. It also never considered that also the
vtop->sym member might have referred to such symbols (see the
testcase with the local static, that used to segfault).
For fixing this (can be seen better with valgrind and SYM_DEBUG)
simply leave local symbols of stmt exprs on the stack.
But like GCC do warn about changes in signedness. The latter
leads to some changes in gen_assign_cast to not also warn about
unsigned* = int*
(where GCC warns, but only with extra warnings).
In particular subtracting a defined symbol from current section
makes the value PC relative, and .org accepts symbolic expressions
as well, if the symbol is from the current section.
This option includes a file as if '#include "file"' is the first
line of compiled files. It's processed after all -D/-U options
and is processed per input file.
- generate and use SYM@PLT for plt addresses
- get rid of patch_dynsym_undef hack (no idea what it did on FreeBSD)
- use sym_attrs instead of symtab_to_dynsym
- special case for function pointers into .so on i386
- libtcc_test: test tcc_add_symbol with data object
- move target specicic code to *-link.c files
- add R_XXX_RELATIVE (needed for PE)
MSVC does not support array designator so cannot compile source using
relocs_info. This commit replace the relocs_info array into a set of
functions, each returning the value given by a given field of the struct
reloc_info.
Last use for pltoff_addend field of relocs_info array was removed in
commit 25927df3b7. It is now useless so
this commit removes it and all initialization related to it.
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.
Change alloc_sym_attr into get_sym_attr and add a parameter to control
whether to allocate a new symattr structure or return NULL if symbol is
not found;
add_elf_sym is a confusing name because it is not clear what the
function does compared to put_elf_sym. As a matter of fact, put_elf_sym
also adds a symbol in a symbol table. Besides, "add_elf_sym" fails to
convey that the function can be used to update a symbol (for instance
its value). "set_elf_sym" seems like a more appropriate name: it will
set a symbol to a given set of properties (value, size, etc.) and create
a new one if non exist for that name as one would expect.
- call RtlDeleteFunctionTable
(important for multiple compilations)
- the RUNTIME_FUNCTION* is now at the beginning of the
runtime memory. Therefor when tcc_relocate is called
with user memory, this should be done manually before
it is free'd:
RtlDeleteFunctionTable(*(void**)user_mem);
[ free(user_mem); ]
- x86_64-gen.c: expand char/short return values to int
Also:
- regenerate all tests/pp/*.expect with gcc
- test "insert one space" feature
- test "0x1E-1" in asm mode case
- PARSE_FLAG_SPACES: ignore \f\v\r better
- tcc.h: move some things
Makefile :
- do not 'uninstall' peoples /usr/local/doc entirely
libtcc.c :
- MEM_DEBUG : IDE-friendly output "file:line: ..."
- always ELF for objects
tccgen.c :
- fix memory leak in new switch code
- move static 'in_sizeof' out of function
profiling :
- define 'static' to empty
resolve_sym() :
- replace by dlsym()
win32/64: fix R_XXX_RELATIVE fixme
- was fixed for i386 already in
8e4d64be2f
- do not -Lsystemdir if compiling to .o
tccgen.c:gv() when loading long long from lvalue, before
was saving all registers which caused problems in the arm
function call register parameter preparation, as with
void foo(long long y, int x);
int main(void)
{
unsigned int *xx[1], x;
unsigned long long *yy[1], y;
foo(**yy, **xx);
return 0;
}
Now only the modified register is saved if necessary,
as in this case where it is used to store the result
of the post-inc:
long long *p, v, **pp;
v = 1;
p = &v;
p[0]++;
printf("another long long spill test : %lld\n", *p);
i386-gen.c :
- found a similar problem with TOK_UMULL caused by the
vstack juggle in tccgen:gen_opl()
(bug seen only when using EBX as 4th register)
Except
- that libtcc1.a is now installed in subdirs i386/ etc.
- the support for arm and arm64
- some of the "Darwin" fixes
- tests are mosly unchanged
Also
- removed the "legacy links for cross compilers" (was total mess)
- removed "out-of-tree" build support (was broken anyway)
-- Not a fix
This reverts commit 089ce6235c.
Revert "handle a -s option by executing sstrip/strip program"
-- related, not a fix.
This reverts commit 5cd4393a54.
- "utf8 in identifiers"
from 936819a1b9
- CValue: remove member str.data_allocated
- make tiny allocator private to tccpp
- allocate macro_stack objects on heap
because otherwise it could crash after error/setjmp
in preprocess_delete():end_macro()
- mov "TinyAlloc" defs to tccpp.c
- define_push: take int* str again
Also:
- allow more than one item per line
- respect "quoted items" and escaped quotes \"
(also for LIBTCCAPI tcc_setoptions)
- cleanup some copy & paste
after several "fixes" and "improvements"
b3782c3cf55fb57bead4
feature did not work at all
- Use 'once' flag, not 'ifndef_macro'
- Ignore filename letter case on _WIN32
- Increment global pp_once for each compilation
- would parse linker args in two different places
- would mess up "tcc -v ..." output:
tcc -v test.c
-> test.c
+> test.c
- would use function "tcc_load_alacarte()" to do the contrary of
what its name suggests.
This reverts commit 19a169ceb8.
A patch is implemented as suggested in tinycc-devel mail list.
From: Reuben Thomas
Date: Thu, 31 Jul 2014 16:52:53 +0100
Subject: [PATCH] Add --{no,}-whole-archive support
I resurrected the patch supplied to the mailing list in 2009
Since --whole-archive is a useful flag to get tcc working with
autotools, and of course in its own right, I suggest you have a look
at the patch and see if it is acceptable. I cannot see any suggestion
that it was actively rejected last time round, just no evidence that
it was ever added.
Fix it to actually be able to parse 64bit immediates (enlarge
operand value type). Then, generally there's no need for accepting
IM64 anywhere, except in the 0xba+r mov opcodes, so OP_IM is
unnecessary, as is OPT_IMNO64. Improve the generated code a bit
by preferring the 0xc7 opcode for im32->reg64, instead of the
im64->reg64 form (which we therefore hardcode).
... for fast redeclaration checks
Also, check function parameters too:
void foo(int a) { int a; ... }
Also, try to fix struct/union/enum's on different scopes:
{ struct xxx { int x; };
{ struct xxx { int y; }; ... }}
and some (probably not all) combination with incomplete
declarations "struct xxx;"
Replaces 2bfedb1867
and 07d896c8e5
Fixes cf95ac399c
