* coretype.h (YASM_PATHSEP): New define for path separator.
* vc8/config.h, vc/config.h, dj/config.h: Override.
* dbgfmts/codeview: New, adds "cv8" dbgfmt.
* dbgfmts/Makefile.inc: Hook into build.
* coff-objfmt.c: Enable cv8 dbgfmt for win32 and win64. Change a few things
to more closely match MASM output.
(coff_section_data): Add isdebug flag so that section flags are set properly.
(coff_objfmt_secton_switch): Initialize to 0 here.
(coff_objfmt_init_remaining_section): Initialize to 1 and set various section
flags (DATA, DISCARD, READ) if section name starts with ".debug".
(coff_objfmt_output): If non-NULL dbgfmt, set object flags to say line numbers
are included.
* dwarfwin64_testhd.hex: Minor update due to coff_objfmt_output() change.
Peter Johnson [Mon, 20 Mar 2006 00:43:18 +0000 (00:43 -0000)]
Add a feature and fix a long-standing bug in Win64 output. The new feature
is support for cross-section relative symbol references using "sym-$".
This generates a 32-bit relative relocation similar to those used for
cross-section jumps and calls.
The bugfix is that in Win64 output, RIP-relative relocations do something
special when there is an immediate value (or anything else) between the
value being relocated and the next instruction. E.g.
"shl dword [sym wrt rip], 5" needs to generate a REL32_1 relocation thanks
to the immediate byte following the RIP-relative value.
* symrec.c (sym_type): add SYM_CURPOS to track labels representing the
current assembly position (e.g. $ in NASM, . in GAS).
(yasm_symtab_define_curpos): New function to create symbols of this type.
(yasm_symrec_is_curpos): Check to see if symbol is SYM_CURPOS type.
(yasm_symrec_get_label, yasm_symrec_print): Update to handle SYM_CURPOS.
* symrec.h (yasm_symtab_define_curpos): Prototype.
(yasm_symrec_is_curpos): Prototype.
* gas-bison.y: Use yasm_symtab_define_curpos when defining '.'.
* nasm-bison.y: Use yasm_symtab_define_curpos when defining '$'.
* value.c (yasm_value_finalize_expr): Look for cross-section
"sym-SYM_CURPOS" combinations and generate curpos-relative value if found.
* coretype.h (yasm_value): Add ip_rel member to designate that curpos_rel
is set due to the value being IP-relative (rather than sym-curpos).
* value.h (yasm_value_initialize, yasm_value_init_sym): Initialize ip_rel.
* x86expr.c (yasm_x86__check_ea): Set ip_rel in addition to curpos_rel if
detected WRT rip.
* x86bc.c (x86_bc_insn_tobytes): Use ip_rel instead of curpos_rel when
adjusting for RIP-relative.
* coff-objfmt.c (coff_objfmt_output_value): Properly adjust output and
generate correct relocations for both curpos_rel and ip_rel. This
includes new generation of REL32_1, REL32_2, etc relocations.
* symrec.c, symrec.h (yasm_symtab_define_label2): Delete.
* coff-objfmt.c (stabs_debgfmt_generate_sections): Change to use
yasm_symtab_define_label() instead.
* win32-curpos.asm, win64-curpos.asm, curpos.asm, curpos-err.asm,
elf_gas64_curpos.asm: New tests for above.
Peter Johnson [Sun, 19 Mar 2006 18:31:00 +0000 (18:31 -0000)]
* value.c (value_finalize_expr): Check for purely -1*symrec cases; these
are also invalid.
* fwdequ64.asm: This test actually had a "4-label" expression; the correct
way to write this is "4-(label-$$)" (same output generated).
Peter Johnson [Sun, 19 Mar 2006 04:18:10 +0000 (04:18 -0000)]
Massive cleanup of relocation and WRT handling. Closes #49 and lays the
groundwork for further features and possible cleanups.
Note: this commit changes the way in which relocations in the
COFF/Win32/Win64 target can be forced to reference a different symbol than
is being pointed to; instead of the ambiguous "trap+(trap.end-trap)" to get
the reloc to point at trap.end but reference the trap symbol, after this
commit "trap.end wrt trap" is the way to say this. This also reads a lot
more clearly and is not ambiguous. This should really only affect people
who write .pdata sections for Win64. See the
objfmts/win64/tests/win64-dataref.asm testcase for an example of usage.
This cleanup adds a new data structure, yasm_value, which is used for all
expressions that can be potentially relocatable. This data structure
splits the absolute portion of the expression away from the relative
portion and any modifications to the relative portion (SEG, WRT,
PC-relative, etc). A large amount of code in the new value module breaks
a general expression into its absolute and relative parts
(yasm_value_finalize_expr) and provides a common set of code for writing
out non-relocated values (yasm_value_output_basic).
All bytecode handling in both libyasm and the architecture modules was
rewritten to use yasm_values when appropriate (e.g. data values,
immediates, and effective addresses). The yasm_output_expr_func is now
yasm_output_value_func and all users and implementors (mainly in object
formats) have been updated to handle yasm_values.
Simultaneously with this change, yasm_effaddr and yasm_immval full
structure definitions have been moved from bc-int.h to bytecode.h.
The data hiding provided by bc-int.h was relatively minimal and probably
overkill. Also, great simplifications have been made to x86 effective
address expression handling.
Peter Johnson [Sat, 18 Mar 2006 22:36:21 +0000 (22:36 -0000)]
Eliminate some signed/unsigned character mismatches in GAP build.
* phash.c, phash.h (phash_checksum): Change to take signed char.
* perfect.h (key): Change name_k to char.
* perfect.c (initinl): Cast name_k.
Peter Johnson [Tue, 14 Mar 2006 08:52:41 +0000 (08:52 -0000)]
More gracefully handle absolute section refernce expansion, and allow for
correct detection of absolute section reference loops (fixing a crash case).
This is also needed for an ongoing rewrite of reloc/value handling.
* expr.c (expr_xform_bc_dist): Remove transformation of absolute section
references; move in changed form into...
(yasm_expr__level_tree): Here. The new code doesn't immediately calculate
the distance from the start of the absolute section to the referenced symbol;
rather it generates an expression for this quantity. As this actually adds
new absolute section refs to the tree, we can't expand with YASM_EXPR_SYMs,
otherwise we would expand multiple times. Thus we need a new
YASM_EXPR_SYMEXP type that thus does not get expanded. Unfortunately this
ripples changes a bit because everywhere *else* we look for YASM_EXPR_SYM,
we now need to look for YASM_EXPR_SYMEXP as well...
(expr_xform_bc_dist): Here.
(yasm_expr__copy_except): Here.
(yasm_expr_extract_symrec): Here.
(yasm_expr_get_symrec): Here.
(yasm_expr_print): Here.
* bin-objfmt.c (bin_objfmt_expr_xform): And here.
* expr-int.h (yasm_expr__type): Define new YASM_EXPR_SYMEXP.
* section.h (yasm_section_abs_get_sym): To implement above, we need to get
a symbol referencing the first bytecode in the absolute section. To avoid
creating redundant symrecs, one is generated for us now. This function
lets us get it in yasm_expr__level_tree().
* section.c (yasm_section_abs_get_sym): Implement.
(yasm_section): Add necessary SECTION_ABSOLUTE data.
(yasm_section_create_absolute): Create the symrec here.
* absloop-err.asm: New test for absolute section reference loops.
Peter Johnson [Sun, 5 Mar 2006 21:15:59 +0000 (21:15 -0000)]
* hamt.c: Add stopgap fix for GAP in cross-build situations by typedefing
uintptr_t to unsigned long for the build platform (instead of trying to
pull in the host platform's _stdint.h).
Peter Johnson [Sun, 5 Mar 2006 01:07:38 +0000 (01:07 -0000)]
* phash.c (phash_lookup), perfect.c (initnorm): Mask upper bits in hash
calculations so that these functions behave identically on 32-bit and >32-bit
architectures.
* gap.c: Fix warnings.
* perfect.c: Change K&R to prototypes to avoid warnings.
Peter Johnson [Sat, 4 Mar 2006 22:09:26 +0000 (22:09 -0000)]
Rewrite x86 identifier recognition to use a minimal perfect hash table
instead of re2c-generated code. This gives identifier recognition a
significant speedup and also drastically shortens compilation time of yasm
itself. This rewrite encouraged combining instruction and prefix
recognition into one function and register and target modifier
recognition into a second function (rather than having 5 or so separate
functions).
Also created a state in the NASM parser (as was done in the GAS parser),
so instructions/prefixes are only looked for until an instruction is
recognized. This avoids search time in the instructions hash for operands.
The tool used to generate the new identifier recognition is called GAP.
Someday we might extend this to generate more code than just the perfect
hash lookup.
* tools/gap: New tool to Generate Architecture Parser (aka perfect hashes).
* phash.c, phash.h: Helper functions used by GAP-generated code.
* x86id.re: Delete. Split into..
* x86parse.gap: Contains all identifier recognition portions.
* x86id.c: Contains instruction operand tables and code and higher-level
entry points into x86parse.gap perfect hash recognizers. Chose to flow
history of x86id.re into this file.
* arch.h: Combine instruction/prefix entry points and register/target
modifier entry points.
* lc3barch.c, lc3bid.re, lc3barch.h: Update to match.
* x86arch.c, x86arch.h: Update to match.
* Makefile.am, various Makefile.inc: Update.
* POTFILES.in: Update due to numerous file changes (not just this commit).
* Mkfiles: Update. VC build files untested at the moment.
Peter Johnson [Thu, 2 Mar 2006 02:53:11 +0000 (02:53 -0000)]
* x86id.re (retnf_insn, yasm_x86__parse_check_insn): Fix handling of retf
(NASM syntax) in 64-bit mode. While I'm here, make all ret forms in GAS mode
match GAS output.
Peter Johnson [Tue, 28 Feb 2006 07:59:29 +0000 (07:59 -0000)]
* coff-objfmt.c: Fix crash when sections are generated inside of the
assembler that don't call section_switch(), e.g. for dwarf2 internally
generated debug sections.
Peter Johnson [Sat, 25 Feb 2006 19:39:57 +0000 (19:39 -0000)]
Fix #70 by allowing overrides on the default (usually ".text") section.
The fix for this rippled into a lot of places, and I'm starting to see some
opportunities for cleaning up some of the object and objfmt structures.
* objfmt.h (yasm_objfmt_add_default_section): Move from standalone function
into objfmt-specific function.
(yasm_objfmt_module): Remove default_section_name string, and add objfmt
specific add_default_section function.
* yasm.c (main): Use slightly updated parameters when calling.
* xdf-objfmt.c, bin-objfmt.c, dbg-objfmt.c, coff-objfmt.c: Implement.
Usually this required refactoring the objfmt-specific section data creation
into a separate function that could be used by both section_switch() and
the new add_default_section() functions, and changing section_switch() to
update changes to the section data if section was new or previously just a
default section, instead of the previous behavior of warning if the section
was not new.
* objfmt.c: Delete (no longer needed).
* Makefile.inc, Makefile.flat, libyasm.vcproj
Makefile.dj: Update to reflect removal.
Peter Johnson [Fri, 24 Feb 2006 04:29:39 +0000 (04:29 -0000)]
Fix #69 by making the NASM preproc and parser use the yasm built-in
alignment bytecode rather than just times'ing a NOP. This generates better
NOP code.
The new align only triggers when the NASM align directive is used unadorned
or with nop as the parameter (e.g. "align 16" or "align 16, nop"). Other
uses, including all uses of balign, maintain their old NASM behavior. This
is somewhat useful if you still want a string of NOPs rather than more
optimized instruction patterns: just use "balign X, nop" rather than
"align X". The new align also follows the GAS behavior of increasing the
section's alignment to be the specified alignment (if not already larger).
While I was in here, I found and fixed a bug in 16-bit alignment generation
(typo). I also changed the x86 32-bit code alignment fill pattern per
suggestions in the AMD x86 code optimization manual.
* nasm-bison.y: Implement a new [align] directive that can take a single
parameter (the alignment) and generate a nop-generating align bytecode.
* standard.mac: Change align macro to generate [align] if the second
macro parameter is nonexistent or "nop".
* x86arch.c (x86_get_fill): Update 32-bit fill pattern and fix bug in 16-bit
fill pattern.
Peter Johnson [Sun, 12 Feb 2006 23:12:10 +0000 (23:12 -0000)]
* intnum.c (yasm_intnum_create_leb128): Create an intnum from a LEB128
encoded value.
* intnum.h (yasm_intnum_create_leb128): Prototype.
* leb128_test.c: Test above.
Peter Johnson [Sat, 11 Feb 2006 21:58:35 +0000 (21:58 -0000)]
Generate DWARF2 information from asm source, closing #43. Asm-level source
debug information is generated if no file/loc directives are used. Also will
generate basic DWARF2 info/abbrev/aranges if not specified in source. What's
not handled is multiple code sections; I need to figure out how DWARF2 expects
these to be generated.
The implementation of this refactors all the line generation into
dwarf2-line.c and makes dwarf2-dbgfmt.c contain only the core DWARF2
functions.
One challenge yet to be taken care of is how to test the automatic generation,
as the current working directory is saved into the output object file.
* dwarf64_2loc: Update so built-in info generation doesn't happen.
Peter Johnson [Sat, 11 Feb 2006 21:52:20 +0000 (21:52 -0000)]
* elf-objfmt.c (elf_objfmt_output): Create all missing section headers using
initial traversal of object sections. This ensures any symbols generated by
elf_objfmt_create_dbg_secthead() are properly accounted for.
(elf_objfmt_create_dbg_secthead): Update to be properly called from traversal.
Peter Johnson [Sat, 11 Feb 2006 21:33:51 +0000 (21:33 -0000)]
* linemgr.h (yasm_linemap_traverse_filenames): New function to go through all
filenames used by linemgr.
* linemgr.c (yasm_linemap_traverse_filenames): Implement.
Peter Johnson [Fri, 10 Feb 2006 04:24:12 +0000 (04:24 -0000)]
Fix #67 by autosizing push immediate with no specified size to either 8 bit
or the current BITS-derived operation size.
* x86id.re (OPS_BITS): New match rule for size == BITS setting.
(push_insn): Use OPS_BITS for 16-bit/32-bit mode, and reuse GAS rule for
64-bit mode.
(yasm_x86__finalize_insn): Set size_lookup[7] to BITS setting. Set opersize
based on this value if OPS_BITS is used.
* pushnosize.asm: New test for push imm, both sized and unsized, in all BITS
modes.
Peter Johnson [Fri, 10 Feb 2006 04:09:38 +0000 (04:09 -0000)]
* intnum.c (yasm_intnum_get_sized): Honor negative warn setting (signed).
* intnum.h (yasm_intnum_check_size): Remove reference to nonexistent param
"is_signed" (this was replaced a long time ago with the rangetype param).
Peter Johnson [Thu, 9 Feb 2006 05:00:09 +0000 (05:00 -0000)]
* intnum.h (yasm_get_sleb128, yasm_size_sleb128): New functions to handle
signed LEB128-encoded numbers straight from long values (rather than going
through intnum).
(yasm_get_uleb128, yasm_size_uleb128): Likewise, for unsigned LEB128.
* intnum.c (get_leb128, size_leb128): Move most of functionality here from:
(yasm_intnum_get_leb128, yasm_intnum_size_leb128): Here (refactor).
(yasm_get_sleb128, yasm_size_sleb128, yasm_get_uleb128, yasm_size_uleb128):
Implement using refactored functionality.
Peter Johnson [Sun, 5 Feb 2006 22:49:10 +0000 (22:49 -0000)]
Match yasm_symtab_traverse() implementation to doxygen documentation by
fixing underlying HAMT implementation to match. Change is to reflect that
traversal stops when subfunction return is nonzero.
* hamt.h (HAMT_traverse): Update doxygen comment for stop on nonzero instead
of stop on zero.
* hamt.c (HAMT_traverse): Implement.
* symrec.c, xdf-objfmt.c, elf-objfmt.c, coff-objfmt.c: Update return values
for subfunctions.
Peter Johnson [Sun, 5 Feb 2006 02:47:56 +0000 (02:47 -0000)]
* bytecode.c (yasm_bc_tobytes): *really* fix handling of bytecodes larger than provided buffer.
The only way to cause this to happen is with incbin, and while we do have a test intended for this,
it doesn't actually check this functionality due to a change made to unbreak distcheck; next up is
figuring out a good test that will work for different source/build directories.
Peter Johnson [Fri, 3 Feb 2006 06:38:30 +0000 (06:38 -0000)]
* dwarf2-dbgfmt.c (generate_section): Split off large chunks into gen_line_op
and finalize_locs functions, preparing for later asm-source dbg generation.
Peter Johnson [Wed, 1 Feb 2006 07:01:47 +0000 (07:01 -0000)]
Implement DWARF2 .file "foo.c" (sans file number) so that it actually sets
the source filename stored in the object file (as GAS does). The internals
of this move the source and object filename data into the yasm_object
structure and simplify the dbgfmt and objfmt interfaces to remove this
information from function parameters.
Peter Johnson [Tue, 31 Jan 2006 03:12:59 +0000 (03:12 -0000)]
* dwarf2-dbgfmt.c (dwarf2_dbgfmt_generate_section): Propogate last-label refs
to following locations in case there is no intermediate label.
* dwarf64_2loc.asm: Test for this.
Peter Johnson [Sun, 29 Jan 2006 21:11:03 +0000 (21:11 -0000)]
Add DWARF2 debugging format. Note: at the moment, this ONLY implements
passthrough information for line numbers through the use of the .loc and
.file directives in the GAS parser and does NOT generate line numbers for
raw assembly files. This will be added next, after the other Mkfiles are
updated.
* dwarf2-dbgfmt.c: New.
* elf-objfmt.c (elf_objfmt_dbgfmt_keywords): Add to list of supported dbgfmts.
* coff-objfmt.c (elf_objfmt_dbgfmt_keywords): Likewise.
* elf-objfmt.c (elf_objfmt_create_dbg_secthead): Support all DWARF2 debug_*
sections.
* dwarf2/tests: Tests for passthrough cases for ELF32, ELF64, and Win64.
Peter Johnson [Sun, 29 Jan 2006 21:06:31 +0000 (21:06 -0000)]
* arch.h (yasm_arch_min_insn_len): Get the minimum instruction length in
bytes.
(yasm_arch_get_address_size): Get the active address size in bits.
* lc3barch.c, x86arch.c: Implement.
These are needed for the DWARF2 dbgfmt, but may be useful for other things
in the future.
Peter Johnson [Sun, 29 Jan 2006 21:04:00 +0000 (21:04 -0000)]
* dbgfmt.h (yasm_dbgfmt_directive): Add section parameter to specify currently
active section at the time of the directive.
* null-dbgfmt.c, stabs-dbgfmt.c: Implement.
* gas-bison.y: Pass.
Peter Johnson [Sun, 29 Jan 2006 01:17:49 +0000 (01:17 -0000)]
Add functions to split a path into a directory and base filename.
* splitpath.c (yasm__splitpath_unix, yasm__splitpath_win): New.
* coretype.h: Prototype above and alias yasm__splitpath.
* splitpath_test.c: New test for above.
Peter Johnson [Mon, 23 Jan 2006 00:16:38 +0000 (00:16 -0000)]
Implement GAS syntax ".local" directive. This is used by GCC to generate
.local sym; .comm sym; rather than directly using .lcomm sym. Handle this
usage as well. While we're here, also implement alignment for .lcomm and
refactor .lcomm handling and alignment handling.
* gas-token.re: Recognize .local.
* gas-bison.y (DIR_LOCAL): Implement .local.
(DIR_COMM): Recognize .local ; .comm case and call define_lcomm().
(DIR_LCOMM): Move functionality into..
(define_lcomm): Here, and implement alignment with..
(gas_parser_align): That now takes raw exprs. The valparam part of that is
now implemented in..
(gas_parser_dir_align): Formerly gas_parser_align.
(DIR_ALIGN): Use gas_parser_dir_align() instead.
* symrec.c (yasm_symtab_get): New function to just get a symbol based on name
without actually "referencing" it.
* symrec.h (yasm_symtab_get): Prototype.
* coretype.h (yasm_sym_vis): Add YASM_SYM_DLOCAL for flagging that a symbol
is explicitly flagged as a local symbol (rather than just default that way).
Peter Johnson [Mon, 23 Jan 2006 00:07:00 +0000 (00:07 -0000)]
* intnum.c (yasm_intnum_set_uint): New function to set intnum to any
arbitrary integer.
(yasm_intnum_zero): Reimplement using yasm_intnum_set_uint.
* intnum.h (yasm_intnum_set_uint): Prototype.
Peter Johnson [Wed, 18 Jan 2006 02:20:06 +0000 (02:20 -0000)]
x86id.re: Fix #66 (fstp generated invalid opcode). Cleaned up fldstpt_insn
a bit as well. We were testing this but for some reason didn't check the
"golden" outputs for correctness.
Peter Johnson [Wed, 18 Jan 2006 01:51:08 +0000 (01:51 -0000)]
* yasm.c (opt_objfmt_handler): Lowercase the objfmt keyword. This forces
the __YASM_OBJFMT__ value to always be lowercase, e.g. for "-f Win32" on
the command line, __YASM_OBJFMT__ becomes "win32".
Requested by: Brian Gladman <brg@gladman.plus.com>
Peter Johnson [Tue, 17 Jan 2006 04:47:49 +0000 (04:47 -0000)]
VC8 build files cleanup. Also combines yasm32.rules and yasm64.rules into single yasm.rules.
Works for native Win32 builds, Win32->x64 cross builds, and native x64 builds.
Submitted by: Brian Gladman <brg@gladman.plus.com>
Peter Johnson [Sun, 15 Jan 2006 22:48:55 +0000 (22:48 -0000)]
* coff-objfmt.c: Add "x64" as an alias for "win64" (when selecting object
formats). This makes integration with the VC8 build tools much nicer, as
VC8 uses "x64" as the PlatformName environment setting.
* win64/Makefile.inc, Mkfiles/Makefile.flat, Mkfiles/Makefile.dj: Update
build files to look for objfmt_x64 module.
Suggested by: Brian Gladman <brg@gladman.plus.com>
Peter Johnson [Fri, 13 Jan 2006 06:41:50 +0000 (06:41 -0000)]
* elf-x86-x86.c (elf_x86_x86_write_reloc): Don't cast symbol index to
unsigned char. This was causing relocation corruption after 255 symbols,
in ELF32 only.
* elfmanysym.asm: Additional test that has exactly 257 symbols.
We actually should have caught this with the very large 32-bit elf-x86id.asm
test, but that object file (unlike the 64-bit GAS version of it) has only
been manually reviewed. This bug was only in ELF32, and wasn't in ELF64.
Fixes bug #62 submitted by: Jens von der Heydt <mailme@vdh-webservice.de>
Peter Johnson [Wed, 11 Jan 2006 05:51:32 +0000 (05:51 -0000)]
Build cleanups (mostly for warnings) for VC8.
* Mkfiles/vc8/config.h: Define _CRT_SECURE_NO_DEPRECATE to avoid most warnings about *_s functions.
Also use _stricmp directly instead of stricmp to avoid warnings.
* libyasm/strcasecmp.c: Support _stricmp usage.
* util.h: Likewise.
* tools/re2c/actions.c: Include ctype.h to define toupper/tolower.
* Mkfiles/vc8/modules/modules.vcproj: Don't include nasm-macros.c directly; this file is included
into another file and has no public variables.
* Mkfiles/vc8/yasm.sln: Allow x64 cross-builds from Win32 by building genmacro, genmodule, and
re2c for Win32 target.
Peter Johnson [Fri, 16 Dec 2005 06:06:25 +0000 (06:06 -0000)]
Rather than pulling the objfmt keyword back from the objfmt when setting
__YASM_OBJFMT__, use whatever name was specified on the command line. This
also matches the behavior of yasm preprocess-only mode (-e). Both the
mismatch between preproc/non-preproc and the internal name change (e.g. from
elf to elf32) were quite confusing.
projects / yasm / log