BR 1974170: VCVTPD2PS, VCVTPD2DQ, VCVTTPD2DQ with a memory operand are
ambiguous without a specific operand size, so force one to be added.
Split the instruction pattern due to our current clunky handling of
MMX/XMM/YMM registers together with sizes. Fix in the future, please!
The compiler is free to store enum pp_token_type into any size integer
small enough to contain all the values up to 2^n-1 for the smallest n
which contains all the values. Force it to size it to integer size,
since we use it to hold macro positional parameters.
%? - substitute the macro name as invoked
%?? - substitute the macro name as defined
In particular:
%idefine keyword $%?
... can be used to make a new keyword "disappear".
Canonicalize the order of the prehash entries, so we don't have to
worry about looking up both pairs of edges.
When we find a collision that we decide to ignore, there is no point
in adding the same edge into the array again; instead, just skip the
current edge.
If the hash target is the same value, we can permit collisions. This
isn't relevant for the current applications of the hash generator, but
is useful for cases where one have a number of sources for the same
target. It's easy to check, either way.
For the versions of VCMPxx which already embed their condition code,
we do not want an extra immediate argument.
Todo: fix bytecode compiler to complain more about these.
Guess what, SEH again, but in Win64 context, which is completely
different matter from Win32. At lowest level this one boils down to
putting so called imagerel references, or in practical terms
relocations of type ADDR32NB, 0x0003, into .pdata and .xdata
segments. Two possibilities. 1. implement say 'wrt ..imagerel' or 'wrt
..imagebase'. 2. silently enforce ADDR32NB relocations in .pdata and
.xdata segments.
This is basically not a bug report, but a feature request.
It's desired to be able to link .obj modules compiled with 'nasm -f
win32' with Microsoft 'link /safeseh'. As well as to register symbols
(commonly subroutine's entry points or even external symbols) as "safe
handlers." In order to achieve this, several points are required.
First of all, object module has to have absolute symbol named @feat.00
with value of 1. This can actually be achived by adding 'absolute 1'
and '@feat.00:' to source code, but it's desirable that it's
autogenerated for win32 modules.
Handler registration is essentially symbol's *index* in current
module's symbol table in .sxdata, segment with 0x200 segment flags, an
"info" segment. It's also essential that symbol has type 0x20 (see
below). All this is depicted in following framgents of 'objdump -xD'
output:
Sections:
Idx Name Size VMA LMA File off Algn
0 .sxdata 00000004 00000000 00000000 0000003c 2**2
CONTENTS, READONLY, DEBUGGING
SYMBOL TABLE:
...
[ 5](sec -1)(fl 0x00)(ty 0)(scl 3) (nx 0) 0x00000001 @feat.00
[ 6](sec 0)(fl 0x00)(ty 20)(scl 2) (nx 0) 0x00000000 _handler
...
<.sxdata>
0: 06 00 00 00
Note [6] and (ty 20) in _handle line in SYMBOL TABLE. "06 00 00 00" in
.sxdata is little-endian 6, _handler's index. This is what makes up
"registration." It's impossible to achieve this with current nasm
facilities and it's probably appropriate to introduce a directive for
it, 'safeseh _handler' is probably most natural choice.
Normally, contexts aren't used with a large number of macros, but in
case someone does, do use hash tables for those as well. This
simplifies the code somewhat, since *all* handling of macros is now
done via hash tables.
Future note: consider if it wouldn't be better to allow struct
hash_table to be allocated by the caller, instead of being allocated
by the hash table routine.
The disassembler code gets cleaner if we do *not* separate out the
conditional instructions; instead, rely on the fact that the
conditionals are always at the end and use FIRST_COND_OPCODE as a
barrier.
