release the source location in case it was also a temp. Reason: we
don't increase reference counts when assigning to a ti_const (they are
like const parameters), so the original value must stay alive until
the ti_const temp is freed
o free the original data in the ttempdelete node for the ti_const temp
o don't increase reference counts when assigning composite types to
ti_const temps either
git-svn-id: trunk@34287 -
adjust the parameter comparison code to ignore types of $vmt parameters
when comparing (previously they were all voidpointer and hence also
equal)
git-svn-id: trunk@30955 -
correct type since quite a while)
o exceptions:
o objects/records (self is a "var" parameter there, and sometimes
we want to load it as a pointer to an object instead
o Objective-C class methods, where self has to be id (to have a signature
compatible with what Objective-C compilers generate), but when loading
it we want a classrefdef of the current class for type check and
method resolving
git-svn-id: trunk@30949 -
load a value from a nested context, or to pass a context to a nested
routine. In case a target uses both the stack and frame pointer, this
difference can matter.
git-svn-id: trunk@29937 -
o made all (non-abstract) tdef and tsym constructors virtual
o added c*def/c*sym classref types for every (non-abstract) t*def/t*sym
class
o added cpusym unit for every architecture that derives a tcpu*def/tcpu*sym
class from the base classes, and initialises the c*def/c*sym classes with
them. This is done so that the llvm target will be able to derive from
the tcpu*def/sym classes without umpteen ifdefs, and it also means that
the WPO can devirtualise everything because the c* variables are only
initialised with one class type
o replaced all t*def/t*sym constructor calls with c*def/c*sym constructor
calls
git-svn-id: trunk@27361 -
defcmp.pas:
+ add new function "def_is_related" which combines the "is_related" overloads of "tobjectdef", "trecorddef" and "tstringdef" (it returns "false" for other def types which is what "tdef.is_related" did)
* compare_defs_ext & compatible_childmethod_resultdef: change call from "x.is_related" to "def_is_related(x,...)"
symtype.pas, tdef:
- remove "is_related" method
symdef.pas:
- remove "is_related" in "tobjectdef", "trecorddef" and "tstringdef"
* tobjectdef.needs_inittable: for checking whether a Corba interface somehow inherits from a IInterface don't use "is_related" anymore (we want to avoid the dependency after all), but mimic the necessary functionality of "def_is_related"
htypechk.pas, nadd.pas, ncal.pas, ncnv.pas, ngtcon.pas, nld.pas, optvirt.pas, pdecobj.pas, pdecvar.pas, pexpr.pas, pgenutil.pas:
* change call from "x.is_related" to "def_is_related(x,...)"
symtable.pas
+ use unit defcmp
* change call from "x.is_related" to "def_is_related(x,...)"
jvm/njvmcnv.pas, jvm/njvmflw.pas:
* change call from "x.is_related" to "def_is_related(x,...)"
git-svn-id: trunk@25847 -
e.g. being a const parameter or immutable temp. values), managed types like dyn. arrays, new string types and interfaces can be kept in registers.
git-svn-id: trunk@24953 -
o support for the new codepage-aware ansistrings in the jvm branch
o empty ansistrings are now always represented by a nil pointer rather than
by an empty string, because an empty string also has a code page which
can confuse code (although this will make ansistrings harder to use
in Java code)
o more string helpers code shared between the general and jvm rtl
o support for indexbyte/word in the jvm rtl (warning: first parameter
is an open array rather than an untyped parameter there, so
indexchar(pcharvar^,10,0) will be equivalent to
indexchar[pcharvar^],10,0) there, which is different from what is
intended; changing it to an untyped parameter wouldn't help though)
o default() support is not yet complete
o calling fpcres is currently broken due to limitations in
sysutils.executeprocess() regarding handling unix quoting and
the compiler using the same command lines for scripts and directly
calling external programs
o compiling the Java compiler currently requires adding ALLOW_WARNINGS=1
to the make command line
git-svn-id: branches/jvmbackend@20887 -
When arguments are temps, they are supposed to be released by tcgcallnode.release_para_temps. However, taking address with taddrnode effectively changes original location of temp to some LOC_REGISTER, tcgcallnode.release_para_temps no longer sees the original location and therefore does not free such temps.
git-svn-id: trunk@19959 -
pexpr.pas: Changes in postfixoperators and the base of handle_factor_typenode not yet incorporated (the code from trunk was simply commented for now)
git-svn-id: branches/svenbarth/generics@19676 -
The changes regarding pretty names for generics and token buffer endianess were integrated into my changes. Not every call to generate_specialization is fixed though, so compilation will fail.
git-svn-id: branches/svenbarth/generics@19674 -
Conflicts:
compiler/pdecl.pas
compiler/pexpr.pas
compiler/pgenutil.pas
compiler/ptype.pas
The original log messages as git was a bit forgetting here :( (newest at the top):
commit 7ef252de8023494ee6d39910e289f9e31658d47b
Author: Sven Barth <pascaldragon@minerva>
Date: Mon Nov 21 17:13:36 2011 +0100
Fix the compilation of inline specializations of which the generic is derived from another generic.
pgenutil.pas, generate_specialization:
* Set the "block_type" to "bt_type" when parsing the type parameters, so that the nodes are returned as "ttypenode" instead of e.g. "tloadvmtaddrnode" in case of classes outside of type sections.
* Set the "block_type" to "bt_type" before calling "read_name_type", so that no unexpected sideeffects happen, because types like classes normally only are declared inside type sections (e.g. for the case a generic class is derived from another generic class a classrefdef for the specialized parent class will be created inside the derived specialized class if the block type is not a type one).
commit 1041a8f7a3a41f4fdf2975ce40055c698281ce71
Author: Sven Barth <pascaldragon@minerva>
Date: Fri Nov 18 19:03:50 2011 +0100
Improve inline specializations a bit, so now expressions like "TSomeGeneric<TSomeType>.SomeClassProc OP SomeNonGeneric" is possible. Using another class function of a generic as the right side is not yet working (that still needs some thinking).
To achive this the generalization code must basically continue directly after the "factor" call, so that the operator and the right side are correctly parsed when walking up the call stack. This is done by jumping from the end of the specialization code in the "<"-case to the start of "sub_expr". The freshly generated node (in the above example a callnode) will be passed down the callstack through a new parameter "factornode". If that is set (currently only in the case of a specialization on the left side) "factor" won't be called and the right side will be parsed with the "factornode" as the left side. If it is not set (which is the case for all other calls to "sub_expr" in the unit) then the usual call to "factor" will be done and the result will be used as the left side.
commit a01ccd265f8d6cc5a2f3e88e23afbcd3d5960afb
Author: Sven Barth <pascaldragon@minerva>
Date: Fri Nov 18 18:37:04 2011 +0100
Fix compilation of ppudump.
symconst.pas:
* Remove sto_has_generic, which was the last remainer of my "overloaded type symbols" approach.
* Remove df_methods_specialized, as it isn't needed anymore with the recent "temporary symtable" solution.
psub.pas, specialize_objectdefs, process_abstractrecorddef:
Remove the checks for/inclusion of df_methods_specialized.
utils/ppudump.pp:
Add "sp_generic_dummy" to the symbol options.
commit d16deac060e65d4b53e8fe9c27fe7e1f6d00a416
Author: Sven Barth <pascaldragon@minerva>
Date: Wed Nov 16 16:34:51 2011 +0100
Fix compilation of "gset.pp" from fcl-stl.
nld.pas:
Extend ttypenode by a reference to the type symbol. Normally this is simply the typesym of the given def, but for specializations in type sections of generics this is not the case, because generate_specialization will return a reference to the generic definition and not the new one (thus the symbol will be wrong).
ppu.pas:
Increase PPU version because of the extension of ttypenode.
pexpr.pas:
* handle_factor_typenode: Extend the function by a "sym" parameter which will normally be "nil". In that case it is set to the def's typesym. The "typesym" field of the created type node is then set to this sym.
* For now pass nearly always "nil" for the above mentioned sym except inside factor_read_id when we've encountered a typesym.
ptype.pas, read_named_type, expr_type:
Exchange the "is_owned_by" check with a "sym_is_owned_by" check so that we can correctly detect that we are using a specialized type declaration inside a generic (once nested generic are allowed this condition needs to be checked).
commit 23668d2fc9070afc26b4288ed0db9a8eaf6f40e6
Author: Sven Barth <pascaldragon@minerva>
Date: Wed Nov 16 07:51:12 2011 +0100
psub.pas:
* tcgprocinfo.parse_body: Methods of generic classes need to set "parse_generic" as well, so that variables for "stacked generics" (generic array => generic record) inside the method body are handled correctly.
* specialize_objectdefs: Don't try to generate method bodies for abstract methods.
pdecvar.pas, read_property_dec:
Allow specializations for the return types of properties (should they be allowed for index types as well?).
symtable.pas:
Add a new class "tspecializesymtable" which is basically a globalsymtable but is always assuming to be the current unit. This symtable is used in "generate_specializations" (see below) and is needed to allow visibilty checks for "private", etc. to succeed.
pgenutil.pas, generate_specializations:
Instead of hackily pushing a symtable that may contain conflicting symbols onto the symtable stack for the specialization, a temporary global symtable using the above mentioned "tspecializesymtable" is created and pushed. After the specialization is done all symbols and defs that were added to the temporary symtable are moved to their final symtable (either the global- or localsymtable of the unit, depending on the current position of compilation). This way symbols are correctly added to a top level symtable, but without potential side effects like resolving the wrong symbol.
git-svn-id: branches/svenbarth/generics@19671 -
- warn if compiler finds implicit or explicit string type conversion from a form of ansi string to a form of unicode string and vice versa
- mark explicit warnings as OFF by default
- warn if compiler finds an assignment of an unicode char const to ansi string or char type (warn for assignment of unicode string const to be implemented)
- revert a piece of code from r19457 regards shortstring handling because shortstring handling should not differ from ansistring in this paticular case
git-svn-id: trunk@19574 -
- don't pass CP_NONE encoding to internal functions. They handled it as 0 encoding. This will optimize the generated code a bit.
- convert all king of string/char/pchar constants to local ansistring def if they needs to be passed to rawbytestring type. They should not get a CP_NONE codepage (delphi compatible)
- don't convert left and right arguments of string concatenation to ansistring type if they are already ansistrings but with different codepage - RTL already handles different codepages in concat routine
- fix resultdef for ansistring concatenations inside assignments - return def of left assignment operand if it is already ansistring - this reduces amount of unneeded condepage conversions since concat functions can return result in any desired codepage
rtl: remove CP_NONE comparisions from most of RTL functions, replace 0 constant with CP_ACP
tests: add test to check various conversions to RawByteString type
git-svn-id: trunk@19519 -
- add helper function getansistringcodepage which returns explicitly set codepage or 0 in other case
- add helper function getansistringdef which return a def with explicitly set codepage or cansistringtype in other case
- change tstoreddef.createnai constructor to allow set codepage in constructor
- don't convert string constants to rawbytestring. if string constant already has a codepage - preserve it or convert to ansistring codepage (delphi compatible)
- don't perform string conversion from ansistring to strings with explicitly set codepage (by directive or by compiler switch) and vice versa (delphi compatible)
+ test which covers most of the cases
git-svn-id: trunk@19510 -
assignment-nodes. For global typed constants and typed constants/
local variable initialisers in regular functions/procedurs, the
assignments are performed in the unit initialisation code. For
those in object/record definitions and their methods, it's done
in the class constructor. Since we may not yet have parsed all
method implementations when the class constructor is parsed, part
of these may be initialised in a helper routine called from the
class constructor. The ones known when the class constructor is
parsed are inited there, because the ones marked as "final" and
declared as static class fields must be initialised in the class
constructor for Java
o new set systems_typed_constants_node_init in systems unit that
indicates that a target uses node trees to initialise typed consts
instead of an initialised data section
o mark typed constants in {$j-} mode as "final" for JVM
o mangle the name of staticvarsyms inside localtables a bit to avoid
name clashes (only with procedure names for now, no parameters yet
so can still cause problems with overloaded routines)
o after a routine has been parsed, it is now processed by
cnodeutils.wrap_proc_body(), which can add extra nodes before code
generation (used for injected the typed constant node trees)
git-svn-id: branches/jvmbackend@18475 -
o support for copying value parameters at the callee side if they were
passed by reference in hlcg
o JVM g_concatcopy() implementation for arrays
o moved code to get length of an array from njvminl to hlcgcpu so it can
be reused elsewhere as well
o export array copy helpers from system unit for use when assigning one
array to another
o some generic support for types that are normally not implicit pointers,
but which are for the JVM target (such as normal arrays)
* handle assigning nil to a dynamic array by generating a setlength(x,0)
node instead of by hardcoding a call to fpc_dynarray_clear, so
target-specific code can handle it if required
* hook up gethltemp() for JVM ttgjvm so array temps are properly
allocated
git-svn-id: branches/jvmbackend@18388 -
since the definition-specific adorning of JVM mangled names is Jasmin-
specific, and such code has no place in symdef
* moved code to adorn JVM mangled names for Jasmin definitions to agjasmin
git-svn-id: branches/jvmbackend@18346 -
