It's not connected up yet because I'm unsure of just where to put things
(it gets messy fast), but just being able to see the structure of
complex types is nice.
This eases type unaliasing on functions a little.
Still more to to go, but this fixes a really hair-pulling bug: linux's
heap randomiser was making the typedef test fail randomly whenever
typedef.qfo was compiled.
When a type is aliased, the alias has two type chains: the simple type
chain with all other aliases stripped, and the full type chain. There
are still plenty of bugs in it, but having the clean type chain takes
care of the major issue that was in the previous attempt as only the
head of the type-chain needs to be skipped for type comparison.
Most of the bugs are in finding the locations where the head needs to be
skipped.
All simple type checks are now done using is_* helper functions. This
will help hide the implementation details of the type system from the
rest of the compiler (especially the changes needed for type aliasing).
They take a pointer to a free-list used for hashlinks so the hashlink
pools can be per-thread. However, hash tables that are not updated are
always thread-safe, so this affects only updates. progs_t has been set
up such that it is easy for multiple progs within one thread can share
hashlinks.
and its usage. The parts of flow_analyze_statement that use it know
where the returned operand needs to go. Unfortunately, this breaks dags
pretty hard, but that's because dags needs to learn about the fancy
assignment-type statements.
I noticed that pointer math is currently incorrect in qfcc, but it would
be nice for fixing it to not break anonstruct since it is testing
something else.
This fixes the technically correct but horrible mess of temps and
addressing when dealing with ivars, and the resulting uninitialized
temps due to the non-constant pointers (do need statement level constant
folding, though).
This is part of what messed up float_val in the encoding for @params.
The other part is something in the linker type encoding merge code: it
may be too aggressive. It's also what messed up the size of @params.
That is, those created by operand_address. The dag code needs the
expression that is attached to the statement to have the correct
expression type in order to do the right thing with the operands and
aliasing, especially when generating temps. This fixes assignchain when
optimizing (all tests pass again).
This reverts commit c78d15b331.
While a block expression's result may be an l-value, block expressions
are not (and their results may not be), thus taking the address of one
is not really correct. It seems the only place that tries to do so is
the assignment code when dealing with structures.
This reverts commit b49d90e769.
I suspect this was a workaround for the mess in assignment chains.
However, it caused compile errors with the new implementation, and is
just bogus anyway.
While I still hate ".=", at least it's more hidden, and the new
implementation is a fair bit cleaner (hah, goto a label in an if (0) {}
block).
Most importantly, the expression tree code knows nothing about it. Now
just to figure out what broke func-epxr. A bit of whack-a-mole, but yay
for automated tests.
Doing it in the expression trees was a big mistake for a several
reasons. For one, expression trees are meant to be target-agnostic, so
they're the wrong place for selecting instruction types. Also, the move
and memset expressions broke "a = b = c;" type expression chains.
This fixes most things (including the assignchain test) with -Werror
turned off (some issues in flow analysis uncovered by the nil
migration: memset target not extracted).
Now convert_nil only assigns the nil expression a type, and nil makes
its way down to the statement emission code (where it belongs, really).
Breaks even more things :)
This bug drove me nuts for several hours until I figured out what was
going on.
The assignment sub-tree is being generated, then lost. It works for
simple assignments because a = b = c -> (= a (= b c)), but for complex
assignments (those that require move or memset), a = b = c -> (b = c) (a
= c) but nothing points to (b = c). The cause is using binary
expressions to store assignments.
It's not possible to take the address of constants (at this stage) and
trying to use a move instruction with .zero as source would result in
the VM complaining about null pointer access when bounds checking is on.
Thus, don't convert a nil source expression until it is known to be
safe, and use memset when it is not.
This fixes the problem of using the return value of a function as an
element in a compound initializer. The cause of the problem is that
compound initializers were represented by block expressions, but
function calls are contained within block expressions, so def
initialization saw the block expression and thought it was a nested
compound initializer.
Technically, it was a bug in the nested element parsing code in that it
wasn't checking the result value of the block expression, but using a
whole new expression type makes things much cleaner and the work done
paves the way for labeled initializers and compound assignments.
Not that it really makes any difference for labels since they're
guaranteed unique, but it does remove the question of "why nva instead
of save_string?". Looking at history, save_string came after I changed
it from strdup (va()) to nva(), and then either didn't think to look for
nva or thought it wasn't worth changing.
Multi-line calls (especially messages) got rather confusing to read as
the lines jumped back and forth. Now the binding is better but the dags
code is reordering the parameters sometimes.
The server code is not yet ready for doubles, especially in its varargs
builtins: they expect only floats. When float promotion is enabled
(default for advanced code, disabled for traditional or v6only),
"@float_promoted@" is written to the prog's strings.
That was a fair bit trickier than I thought, but now .return and .paramN
are handled correctly, too, especially taking call instructions into
account (they can "kill" all 9 defs).
This reverts commit a2f203c840.
There is indeed a world of difference between "any" and "only", and it
helps if I read the rest of the docs AND the code :P.
As expected, this does not fix the mangled pointer problem in
struct-init-param.r, but it does improve the ud-chains. There's still a
problem with .return, but it's handling in flow_analyze_statement is a
bit "special" :P.
Doing the same thing at the end of two branches of an if/else seems off.
And doing an associative(?) set operation every time through a loop is
wasteful.
This fixes the ICE when attempting to compile address-cast without
optimization (just realized why, too: the assignment was optimized out
of existence).
This the fixes the incorrect flow analysis caused by the def being seen
to have the wrong size (structure field of structure def seen through a
constant pointer). Fixes the ICE, but the pointer constant is broken
somewhere in dags, presumably.
This fixes the problem of using nil for two different compound types
within the one expression. The problem is all compound types have the
same low-level type (ev_invalid) and this caused the two different nils
to have the same type when taken back up to expression level.
While expression symbols worked for what they are, they weren't so good
for ivar access because every ivar of a class (and its super classes)
would be accessed at method scope creation, generating spurious access
errors if any were private. That is, when the access checks worked at
all.
The end goal was to fix erroneous non-constant initializer errors for
the following (ie, nested initializer blocks):
typedef struct { int x; int y; } Point;
typedef struct { int width; int height; } Extent;
typedef struct Rect_s { Point offset; Extent extent; } Rect;
Rect makeRect (int xpos, int ypos, int xlen, int ylen)
{
Rect rect = {{xpos, ypos}, {xlen, ylen}};
return rect;
}
However, it turned out that nested initializer blocks for local
variables did not work at all in that the relocations were lost because
fake defs were being created for the generated instructions.
Thus, instead of creating fake defs, simply record the offset relative
to the base def, the type, and the basic type initializer expression,
then generate instructions that all refer to the correct def but with a
relative offset.
Other than using the new element system, static initializers are largely
unaffected.
This is for adding methods to classes and protocols via their interface,
not for adding methods by adding protocols (they still get copied).
Slightly more memory efficient.
Copying methods is done when adding protocols to classes (the current
use for adding regular methods is an incorrect solution to a different
problem). However, when a method is added to a class, the type of its
self parameter is set to be a pointer to the class. Thus, not only does
the method need to be copied, the self parameter does too, otherwise
the self parameter of methods added via protocols will have their type
set to be a pointer to the last class seen adding the protocol.
That is, if, while compiling the implementation for class A, but the
interface for class B is comes after the interface for class A, and both
A and B add protocol P, then all methods in protocol P will have self
pointing to B rather than A.
@protocol P
-method;
@end
@interface A <P>
@end
@interface B <P>
@end
@implementation A
-method {} // self is B, not A!
@end
Duplicate methods in an interface (especially across protocols and
between protocols and the interface) are both harmless and even to be
expected. They certainly should not cause the compiler to demand
duplicate method implementations :)
This is actually a double issue: when a class implementing a protocol
used the protocol in @protocol(), not only would the protocol get
emitted as part of the class data specifying that the class conforms to
the protocol, a second instance would be emitted again when @protocol()
was used. On top of that, only the instance referenced by @protocol()
would be initialized. Now, both class emission and @protocol() get their
protocol def from the same place and thus only one, properly
initialized, protocol instance is emitted.
The problem was an erroneous assumption that the methods had to be
defined. Any class implementing a protocol must implement (and thus
define) the methods, but a protocol declaration cannot: it merely
declares the methods, and it's entirely possible for a module to see
only the protocol definition and not any classes implementing the
protocol.
Unlike gcc, qfcc requires foo to be defined, not just declared (I
suspect this is a bug in gcc, or even the ObjC spec), because allowing
forward declarations causes an empty (no methods) protocol to be
emitted, and then when the protocol is actually defined, one with
methods, resulting in two different versions of the same protocol, which
comments in the gnu objc runtime specifically state is a problem but is
not checked because it "never happens in practice" (found while
investigating gcc's behavior with @protocol and just what some of the
comments about static instance lists meant).
It proved to be too fragile in its current implementation. It broke
pointers to incomplete structs and switch enum checking, and getting it
to work for other things was overly invasive. I still want the encoding,
but need to come up with something more robust.a
