|ead725||2003-11-13||Martin Stjernholm|| ||File overview
System interface glue
Emulation of alloca(3) if it doesn't exist.
Implementation of BSD style crypt(3) if it doesn't exist.
dlopen(3) implementation for NT.
Support for setting fd modes (blocking, nonblocking, close-on-exec).
NT emulation of libc style I/O on fd's.
CPU specific code for implementing atomic locks. Used only when
running without the interpreter lock.
dlfcn.h replacement for the dlopen(3) implementation in dlopen.c.
Support stuff for dealing with floats on different systems.
Support stuff for dealing with ip addresses on different systems.
Support for getrusage(2) and look-alikes on different systems, and
support for querying virtual cpu time.
Glue for various stuff that doesn't work quite the same on
Macro package for defining an allocator that lumps together fixed
size structs into larger blocks. Also contains a simple hash table
implementation for lookup of the structs.
Tool for handling lists of functions to call at specific times.
Callbacks may be added and removed at runtime.
Support for detecting and breaking cyclic call chains.
Functions for dmalloc (DEBUG_MALLOC). The code is in
Tool for building strings in streaming fashion.
A generic sort implementation using fsort_template.h.
Macro package for defining a sort function for an arbitrary data
type using an arbitrary compare test.
Initially included in just about every file. Contains macros for
the chosen Pike integer and float types, and various other basic
A simple hash table for pike strings. Used by cpp.c.
Miscellaneous support macros.
The wrappers for malloc(3) etc and the dmalloc implementation.
Also contain some utility functions for handling of generic memory
blocks, e.g. hashing and some wide string tools.
A memory search implementation that handles wide strings. It's
used by search() and others.
Contains the needle specific parts of the memory searcher.
Included by pike_search.c once for each string width of the
Contains the haystack specific parts of the memory searcher.
Included by pike_search_engine.c once for each string width of the
FIFO and LIFO queues for functions to call. The difference from
callback.* is that the queue is emptied when the callbacks have
been called, and that it's designed for large amounts of
callbacks. Used primarily by the gc.
Red/black tree implementation. Used to implement the multisets.
Miscellaneous smaller utility functions, primarily numerical.
Some tools to handle timeval structs.
Wrap together the preprocessor from preprocessor.h.
Macros for the fields in the program_state struct. Used to unify
code that does operations uniformly across all of them.
Convert node trees to opcode sequences. Also call assemble in
peep.c to convert it to byte/machine code.
The yacc/bison language productions. These primarily build node
trees (i.e. abstract syntax trees) from the language constructs.
Node tree handling. Wrap together the tree optimizer. Also contain
an evaluator for constant expressions and dooptcode which wraps
together the conversion of a node tree for a function body to
optimized byte code.
Wrap together the lexer. For mysterious reasons, they also contain
basic opcode table stuff.
The actual lexer. Included by lex.c for different string widths.
Wrap together the peephole optimizer, handle the conversion of
labels from opcodes to offsets, and the conversion from opcode
sequences to machine code. Also have some primitives for opcodes
and insertion into the opcode list.
The core of the peephole optimizer. peep_engine.c is generated
from transformation rules for opcode sequences in peep.in.
peep_engine.c is included by peep.c.
The bulk of the preprocessor. Included by cpp.c once for each
Implement the data type for programs (aka classes). The compiler
is wrapped together here. Also contain the lfun support, the C
level parts of the resolver, and the trampoline program.
Macros for the different areas in the program struct. Used to
unify code that does operations uniformly across all of them.
The node tree optimizer. treeopt.h is generated from tree
transformation rules in treeopt.in. treeopt.h is included by
The deeper parts of the apply function. Included by interpret.c to
make different apply variants.
Contains the different byte code formats, e.g. machine code for
The core interpreter, i.e. apply and eval functions and runtime
scope handling. Also have tools to handle the pike stack, e.g.
pushing and popping various data types.
Implements all the opcodes. Included by interpret.c and
Macros for the opcode prototypes. Extracted from
Wraps up the opcodes in interpret_functions.h into an eval
functions with a loop and a gigantic switch. This is only used
when machine code isn't enabled. The opcodes become separate
functions otherwise. Included by interpret.c.
Import the chosen byte code format from the code subdirectory.
Built-in data types
Implement arrays. Also contain the compare functions for some
orders used in the core.
The magic to hook in Gmp integers as bignums and to convert
between 64 bit ints and bignums on systems where the int type is
The interface class for iterators, the built-in iterators, and the
bulk of the foreach opcode.
Implement objects. Contain the code that initializes and destruct
objects, the implementations of the "magically inherited" lfuns,
and the code that compiles and instantiates the master.
Implement pike strings.
Implement the loading of dynamic modules (the load_module
Standard modules that don't need the recently built pike in order
to be built.
Maintain the entry and exit hooks for the statically compiled
Compatibility header for old modules that don't use
PIKE_MODULE_INIT and PIKE_MODULE_EXIT.
Tools for importing and exporting symbols between modules. Also
contain helpers for checking arguments on the pike stack.
Standard modules that are built after the pike binary. These need
the recently built pike in order to be built.
Core runtime support
Implement Pike.Backend, i.e. callback driven I/O and call outs.
Implement various core functions (mostly efuns).
builtin_functions.c primarily contains the functions that predate
the precompile system.
Case tables for lower_case and upper_case in builtin_functions.c.
Template for combine_path in builtin_functions.c.
The global constants mapping, along with the support for efuns.
Implement encode_value and decode_value. Contain low-level
compile/resolve stuff to decode dumped programs.
Exception handling and debug style fatal exit.
Built-in exception objects. Included by error.c.
The garbage collector. Also contain tools for locating references
to a memory object and to identify and print out the common types
of memory objects in debug style.
The entry point. Contain parsing of low-level command arguments
(i.e. those not handled in the master), initialization and
Contain the opcode enum but nothing else regarding opcodes (see
lex.* instead). Instead we find the cast operator support here,
along with the sscanf implementation.
Implement the runtime behavior for all operators except cast.
Macros for choosing C compiler integer types for INT64, INT32,
INT16 and INT8.
Support for threads on different systems. Also has the header
stuff for swapping pike threads and releasing/taking the
interpreter lock. The code is in threads.c.
Handling of types in pike.
The pike security system.
Signal and process support.
Handling of normal and short svalues, e.g. functions to free,
assign, compare and print svalues. Also contain the parts of the
gc that follow references.
Thread support. Provides a generic thread farm too.
FIXME: Document these.
Call out test. Used by testsuite.in.
GC destruct order test. Used by testsuite.in.
Testsuite for the core.
Statically allocated program id numbers.
The standard unicode database. Data from this is extracted to
build tables for various functions, e.g. the Unicode module and