has been upcasted to int64, but convert it back to cardinal (mantis
#15015)
* put the "remove unnecessary 64 bit type conversions" code between
{$ifndef cpu64bitalu} instead of {$ifndef cpu64bitaddr}
git-svn-id: trunk@14396 -
+ RTL support:
o VFP exceptions are disabled by default on Darwin,
because they cause kernel panics on iPhoneOS 2.2.1 at least
o all denormals are truncated to 0 on Darwin, because disabling
that also causes kernel panics on iPhoneOS 2.2.1 (probably
because otherwise denormals can also cause exceptions)
* set softfloat rounding mode correctly for non-wince/darwin/vfp
targets
+ compiler support: only half the number of single precision
registers is available due to limitations of the register
allocator
+ added a number of comments about why the stackframe on ARM is
set up the way it is by the compiler
+ added regtype and subregtype info to regsets, because they're
also used for VFP registers (+ support in assembler reader)
+ various generic support routines for dealing with floating point
values located in integer registers that have to be transferred to
mm registers (needed for VFP)
* renamed use_sse() to use_vectorfpu() and also use it for
ARM/vfp support
o only superficially tested for Linux (compiler compiled with -Cpvfpv6
-Cfvfpv2 works on a Cortex-A8, no testsuite run performed -- at least
the fpu exception handler still needs to be implemented), Darwin has
been tested more thoroughly
+ added ARMv6 cpu type and made it default for Darwin/ARM
+ ARMv6+ implementations of atomic operations using ldrex/strex
* don't use r9 on Darwin/ARM, as it's reserved under certain
circumstances (don't know yet which ones)
* changed C-test object files for ARM/Darwin to ARMv6 versions
* check in assembler reader that regsets are not empty, because
instructions with a regset operand have undefined behaviour in that
case
* fixed resultdef of tarmtypeconvnode.first_int_to_real in case of
int64->single type conversion
* fixed constant pool locations in case 64 bit constants are generated,
and/or when vfp instructions with limited reach are present
WARNING: when using VFP on an ARMv6 or later cpu, you *must* compile all
code with -Cparmv6 (or higher), or you will get crashes. The reason is
that storing/restoring multiple VFP registers must happen using
different instructions on pre/post-ARMv6.
git-svn-id: trunk@14317 -
from any ordinal constant, and give a warning for this one allowed
conversion like Kylix does (mantis #14713)
* this required calling simplify from typecheckpass in ttypeconvnode for
cord_to_pointer, because this simplification prevents the type checking
from happening (but the typecheck itself also does that simplification)
git-svn-id: trunk@13919 -
starting with a previous 2.3.1 or compiler built from the objc branch
+ added basic objcprotocol support (only for external protocols
currently)
o use in type declaration: "type xp = objcprotocol ... end;"
o when defining a root class that implements it:
"type yc = objcclass(xp) ... end" (note: no support yet
for something like "objcclass(id,xp)" or so)
o when defining a non-root class that implements a protocol:
"type zc = objcclass(nsobject,xp) ... end"
o includes support for "required" and "optional" sections
o no support yet for the objcprotocol(<protocol>) expression
that enables getting a class instance representing the
protocol (e.g., for use with "conformsToProtocol:")
o message names have to specified in protocol declarations,
but if an objcclass implements a protocol, the message names do
not have to be repeated (but if they are, they have to match;
the same goes when overriding inherited methods)
+ allow specifying the external name of Objective-C classes and
protocols, since classes and protocols can have the same name
(and you cannot use the same Pascal identifier in such caseq)
+ added NSObject protocol, and make the NSObject class use it
+ added missing NSObject class methods that have the same name
as instance methods (added "class" name prefix to avoid clashes)
* fixed several cases where the compiler did not treat Objective-C
classes/protocols the same as Object Pascal classes/interfaces
(a.o., forward declarations, alignment, regvars, several type
conversions, ...)
* allow "override" directive in objcclass declarations, and print
a hint if it's forgotten in an external declaration (because it
doesn't really matter there, and may make automated header
conversion harder than necessary) and an error if will be used in
a non-external declaration (because it is not possible to start
a new vmt entry-tree in Objective-C, you can only override parent
methods)
* reject objcclasses/protocols as parameters to typeof()
* don't try to test VMT validity of objcclasses/protocols
git-svn-id: branches/objc@13375 -
instantiated (for wpo)
* also replace vmt-entries for classes for which we don't have any
information at all with FPC_ABSTRACTERROR (since that means they
certainly are not instantiated), except for their published and
virtual class methods
* fixed check for published methods in wpo
git-svn-id: trunk@13219 -
* const s: string = icorbainterface; is possible now
* as operator is working now with corba interfaces
* supports helper function is working now with corba interfaces
git-svn-id: trunk@12729 -
convert it into a widechar at compile time using the current source file's
code page (and in case it was to a non-widestring, then convert it at run
time into the current run time ansi code page) (mantis #12993)
Previously, the characters were either stuffed into the string without
conversion (widestring), or passed to the run time "ansi-character to
stringtype" routine (other string types; which was also wrong, since
that routine expects the character to be in ansi-encoding, which is
not necessarily the same as whatever the source file encoding was)
git-svn-id: trunk@12580 -
whether it's assignable if 1) the size/sign changes (was already the
case) and 2) in case something actually can be assigned to it
git-svn-id: trunk@12039 -
before optimising function result assignment, because at this point the
hidden self parameter is not yet inserted (mantis #12597)
* changed ttypeconvnode.actualtargetnode to use the same logic as what
is used to determine whether something can be assigned to the result
of a type conversion (so the above check also works if the methodpointer
contains a typecast to a different object type)
git-svn-id: trunk@12038 -
size must happen without any mangling of the value (mantis #11027)
* moved checking for signed-to-unsigned (or vice versa) type
conversions in assignments from htypechk to ncnv (where there
was already code for similar checks) and added support for
bool_to_bool there as well
git-svn-id: trunk@10521 -
it doesn't have to duplicate any type checking code, and so constant
expressions get the same resultdefs as non-constant expressions
* properly fixed resultdef determination of "set + setelementn" (follows
same rules now as "set + set")
* also convert "longint or/xor cardinal" to int64 (needed for correct
results with negative numbers and Delphi-compatible) + test
* extended 64-to-32 type conversion simplification to also handle
or/xor nodes (so if the result is typecasted back to 32 bit, the
evaluation can still be done entirely in 32 bit). These changes also
enable that optimization in some extra cases (not just anymore for
expressions containing only uint32)
git-svn-id: trunk@10418 -
a) cpu64bitaddr, which means that we are generating a compiler which
will generate code for targets with a 64 bit address space/abi
b) cpu64bitalu, which means that we are generating a compiler which
will generate code for a cpu with support for 64 bit integer
operations (possibly running in a 32 bit address space, depending
on the cpu64bitaddr define)
All cpus which had cpu64bit set now have both the above defines set,
and none of the 32 bit cpus have cpu64bitalu set (and none will
compile with it currently)
+ pint and puint types, similar to aint/aword (not pword because that
that conflicts with pword=^word)
* several changes from aint/aword to pint/pword
* some changes of tcgsize2size[OS_INT] to sizeof(pint)
git-svn-id: trunk@10320 -
for qwordbool) + test:
o assigning true to such a variable now sets them to $ff/$ffff/$ffffffff
o these types are now all signed
o converting an integer type to a byte/word/long/qwordbool using an
explicit type cast keeps the integer's original value stored in the
bool, instead of forcing it to ord(true)/ord(false)
(mantis #10233 and #10613, implemented for all architectures, testsuite
tested for ppc32, sparc and x86)
* fixed some places where the rtl depended on longbool(true) having the
value 1
* extended several boolean tests (and adapted some to no longer assume
that byte/word/long/qwordbool(true)=1)
+ support for converting to qwordbool in second_int_to_bool for x86, ppc
and sparc
git-svn-id: trunk@9898 -
of u32bit values on 32 bit platforms (after the int64 values have
already been used for overload selection etc, i.e., semantically
nothing changes)
+ test which checks that not too many typecasts are optimized away
git-svn-id: trunk@9664 -
o it's a by-reference parameter, but ignore that since it's
guaranteed to be safe because of the escape analysis
o it's wrapped in an absolute type conversion -> added
actualtargetnode method to tnode which digs through that
git-svn-id: trunk@9412 -
- check for procedure and class pointers as well;
- do not warn if typecasted pointer is passed as parameter or directly assigned to a variable.
git-svn-id: trunk@9264 -
