1
0
mirror of https://github.com/fadden/6502bench.git synced 2024-11-12 21:09:29 +00:00
6502bench/SourceGen/RuntimeData/Help/editors.html
Andy McFadden 03a0fc13fd Expand max local variable width to 257
Code generated by one of the C compilers sets up the stack frame and
then maps the direct page on top of it.  If the value at the top of
the stack is 16 bits, it will be referenced via address $ff.  The
local variable editor was regarding this as illegal, because lvars are
currently only defined for direct page data, and the value doesn't
entirely fit there (unless you're doing an indirect JMP on an NMOS
6502, in which case it wraps around to $00... but let's ignore that).

The actual max width of a local variable is 257 because of the
possibility of a 16-bit access at $ff.

Older versions of SourceGen don't seem to have an issue when they
encounter this situation, as worrying about (start+width) is really
just an editor affectation.  The access itself is still a direct-page
operation.  You won't be able to edit the entry without reducing the
length, but otherwise everything works.  I don't think there's a need
to bump the file version.
2021-07-20 13:08:19 -07:00

415 lines
21 KiB
HTML

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<link href="main.css" rel="stylesheet" type="text/css" />
<title>Editors - 6502bench SourceGen</title>
</head>
<body>
<div id="content">
<h1>6502bench SourceGen: Editors</h1>
<p><a href="index.html">Back to index</a></p>
<h2><a name="address">Edit Address</a></h2>
<p>This adds a target address directive (".ORG") to the current offset.
If you leave the text field blank, the directive will be removed.</p>
<p>The text entry field is initialized to the address of the
first selected line. The "load address", i.e. the place where the
code or data will live when the file is first loaded into memory,
is shown for reference.</p>
<p>If multiple lines were selected, some additional information will be
shown, and an address directive will be added after the last selected
line. This directive will set the address to the "load address".
This is useful for "relocating" a block of code or data in the middle of
the file. You're not allowed to do this when the selected range of
lines spans another address directive.</p>
<p>Addresses are always interpreted as hexadecimal. You can prefix
it with a '$', but that's not required.
24-bit addresses may be written with a bank separator, e.g. "12/3456"
would resolve to address $123456.</p>
<p>There will always be an address directive at the start of the file.
Attempts to remove it will be ignored.</p>
<h2><a name="flags">Edit Status Flag Override</a></h2>
<p>The state of the processor status flags are tracked for every
instruction. Each individual flag is recorded as zero, one, or
"indeterminate", meaning it could hold either value at the start of
that instruction. You can override the value of individual flags.</p>
<p>The 65816 emulation bit, which is not part of the processor status
register, may also be set in the editor.</p>
<p>The M, X, and E flags will not be editable unless your CPU configuration
is set to 65816.</p>
<h2><a name="label">Edit Label</a></h2>
<p>Sets or clears a label at the selected offset. The label must have the
<a href="intro.html#about-symbols">proper form</a>, and not have the same
name as another symbol, unless it's specified to be non-unique. If you
edit an auto-generated label you will be required to change the name.</p>
<p>The label may be marked as non-unique local, unique local, global,
or global and exported. The default is global. If you start typing
a label with the non-unique label prefix character (usually '@',
configurable in
<a href="settings.html#appset-displayformat">application settings</a>),
the selection will automatically switch to non-unique local.</p>
<p>Local labels may be "promoted" to global if the assembler requires it.
Most assemblers define local scope as starting clean after each global
label, but there are exceptions. If a label's name conflicts or is
incompatible with the assembler, it will be renamed.</p>
<p>Exported labels are added to a table that may
be imported by other projects (see
<a href="advanced.html#multi-bin">Working With Multiple Binaries</a>).</p>
<h2><a name="instruction-operand">Edit Instruction Operand</a></h2>
<p>Operands can be formatted explicitly, or you can let the disassembler
select the format for you. By default, immediate constants and
addresses with no matching symbol are formatted as hex. Symbols
defined as address labels, platform/project symbols, and local
variables will be identified and applied automatically.</p>
<h3><a name="explicit-format">Explicit Formats</a></h3>
<p>Operands can be displayed in a variety of numeric formats, or as a
symbol. The character formats are only available for operands
whose value falls into the proper range. The ASCII format handles
both plain and high ASCII; the correct encoding is chosen based on
the operand's value.</p>
<p>Symbols may be used in their entirety, or, when used as constants,
can be shifted and masked.
The low / high / bank selector determines which byte is used as the
low byte. For 16-bit operands, this acts as a shift rather than a byte
select. If the symbol is wider than the operand field, e.g. you're
referencing a 16-bit address in an 8-bit constant, a mask will be
applied automatically.</p>
<p>The editor will try to prevent you from using auto-generated
labels and local variables in the symbol field. These types of symbols
can be freely renamed by SourceGen, and thus cannot be reliably
referenced by name.
You can reference a non-unique local by writing it with the non-unique
label prefix character (default '@'). Ambiguous non-unique references
are not allowed, so if the symbol can't be found the label will
be discarded.</p>
<p>When you select a non-default format option, a "preview" of the
formatted operand will be shown.</p>
<p>The <code>MVN</code> and <code>MVP</code> instructions on the 65816
are a bit peculiar, because they have two operands rather than one.
SourceGen currently only allows you to set one format, which will be
applied to both operands. If you specify a symbol, the symbol will
be used twice, adjusted if necessary. (This limitation may be addressed
in a future release.)</p>
<p>The <code>BBR</code> and <code>BBS</code> instructions on the W65C02
also have two operands: a direct page address, and a relative branch.
In general the direct page address is ignored, so these are treated as
branch instructions.</p>
<p>The bottom part of the window has some shortcuts for working with
address references and local variables. These are primarily used to
change the way things work when "Default" is selected. The shortcuts
don't cause any changes to the recorded format of the instruction
being edited. All of the actions can be performed elsewhere, by
editing the label at the target address, editing the project symbol
set, or editing a local variable table.</p>
<h3><a name="shortcut-nar">Numeric Address References</a></h3>
<p>For operands that are 8-bit, 16-bit, or 24-bit addresses, you can
define a symbol for the address as a label or
<a href="intro.html#symbol-types">project symbol</a>.</p>
<p>If the operand is an address inside the project, you can set a
label at that address. If the address falls in the middle of an
instruction or multi-byte data item, its position will be adjusted to
the start. Labels may be created, modified, or (by erasing the label)
deleted.</p>
<p>The label finder does not do the optional search for "nearby" labels
that the main analyzer does, so there will be times when an instruction
that is shown with a symbol in the code list won't have a symbol
in the editor.</p>
<p>If the operand is an address outside the project, e.g. a ROM
address or I/O location, you can define a project symbol. If a
match was found in the configured platform definition files, it will be
shown; it can't be edited, but it can be overridden by a project symbol.
You can create or modify a project symbol by clicking on "Create Project
Symbol" or "Edit Project Symbol". You can't delete project symbols
from this editor (use Project Properties instead).</p>
<p>It's possible to have more than one project symbol for the same
address. For example, on the Apple II, reading from the memory-mapped
I/O address $C000 returns the last key pressed, but writing to it
changes the state of the 80-column display hardware, so it's useful to
have two different names for it. If more than one project symbol has the
same address, the first one found will be used, which may not be
what is desired. In such situations, you should create the project
symbol and then copy the symbol name into the operand. You can do this
in one step by clicking the "Copy to Operand" button.
(In most cases you don't want to do this, because if the project
symbol is deleted or renamed, you'll have operands that refer to a
nonexistent symbol. Unlike labels, project symbol renames do not
refactor the rest of the project.)
<h3><a name="shortcut-local-var">Local Variable References</a></h3>
<p>For zero-page address operands and (65816-only) stack-relative
constant operands, a local variable can be created or modified. This
requires that a local variable table has been defined at or before
the instruction being edited.</p>
<p>If an existing entry is found, you will be able to edit the name
and comment fields. If not, a new entry with a generic name and
pre-filled value field will be created in the nearest table.</p>
<h2><a name="data-operand">Edit Data Operand</a></h2>
<p>This dialog offers a variety of choices, and can be used to apply a
format to multiple lines. You must select all of the bytes you want
to format. For example, to format two bytes as a 16-bit word, you must
select both bytes in the editor. (If you click on the first item, then
Shift+double-click on the operand field of the last item, you can do
this very quickly.) The selection does not need to be contiguous: you
can use Control+click to select scattered items.</p>
<p>If the range is discontiguous, crosses a logical boundary
such as a change in address or a user-specified label, or crosses a
visual boundary like a long comment, note, or visualization, the selection
will be split into smaller regions. A message at the
top of the dialog indicates how many bytes have been selected, and how
many regions they have been divided into.</p>
<p>(End-of-line comments do <i>not</i> split a region, and will
disappear if they end up inside a multi-byte data item.)</p>
<p>The "Simple Data" items behave the same as their equivalents in the
Edit Operand dialog. However, because the width is not determined by
an instruction opcode, and multiple items can be selected, you will need
to specify how wide each item is and what its byte order is. For data
you also have the option of setting the format to "Address", which marks
the selected bytes as a numeric reference.</p>
<p>Consider a simple example: suppose you find a table of 16-bit
addresses in the code. Click on
the first byte, shift-click the last byte, then select the Edit Data menu
item. The number of bytes selected should be even. Select
"16-bit words, little-endian", then over to the right click on
"Address". When you click OK, the selected data will be formatted as a
series of 16-bit address values. If the addresses can be resolved inside
the data file, each address will be assigned a label.</p>
<p>The "Bulk Data" items can represent large chunks of data compactly.
The "fill" option is only available if all selected bytes have the
same value.
If a region of bytes is irrelevant, perhaps used only as padding, you
can mark it as "junk". If it appears to be adding bytes to reach a
power-of-two address boundary, you can designate it as an alignment
directive. If you have multiple regions selected, only options that
work for all regions will be shown.</p>
<p>The "String" items are enabled or disabled depending on whether the
data you have selected is in the appropriate format. For example,
"Null-terminated strings" is only enabled if the data regions are
composed entirely of characters followed by $00. Zero-length strings
are allowed, but only if some non-zero-length strings are present.
DCI (Dextral Character Inverted) strings have the high bit on the last
byte flipped; for PETSCII this will usually look like a series of
lower-case letters followed by a capital letter, but may look odd if the
last character is punctuation (e.g. '!' becomes $A1, which is a
rectangle character that SourceGen will only display as hex).</p>
<p>The character encoding can be selected, offering a choice between
plain ASCII, low + high ASCII, C64 PETSCII, and C64 screen codes. When
you change the encoding, your available options may change. The
low + high ASCII setting will accept both, configuring the appropriate
encoding based on the data values, but when identifying multiple strings
it requires that each individual string be entirely one or the other.</p>
<p>Due to fundamental limitations of the character set, C64 screen code
strings cannot be null terminated ($00 is '@').</p>
<p>As noted earlier, to avoid burying elements such as labels in the middle
of a data item, contiguous areas may be split into smaller regions. This
can sometimes have unexpected effects. For example, this can be formatted
as two 16-bit words or one 32-bit word:</p>
<pre>
.DD1 $01
.DD1 $ef
.DD1 $01
.DD1 $f0
</pre>
<p>With a label in the middle, it can be formatted as two 16-bit words, but
not as a 32-bit word:</p>
<pre>
.DD1 $01
.DD1 $ef
LABEL .DD1 $01
.DD1 $f0
CODE LDA LABEL
</pre>
<p>If this is undesirable, you can add a label at a 32-bit boundary, and
reference that instead:</p>
<pre>
LABEL .DD1 $01
.DD1 $ef
.DD1 $01
.DD1 $f0
CODE LDA LABEL+2
</pre>
<p>With the label out of the way, the data can be formatted as desired.</p>
<h2><a name="comment">Edit Comment</a></h2>
<p>Enter an end-of-line (EOL) comment, or leave the text field blank to
delete it. EOL comments may be placed on instruction and data lines, but
not on assembler directives.</p>
<p>It's wise to restrict comments to the ASCII character set, because
not all assemblers can accept UTF-8 input. Code generators for such
assemblers will convert non-ASCII characters to '?' or something similar.
If this isn't a concern, you can enter any characters you like.</p>
<p>There is no fixed limit on the number of characters, but you may
want to limit the overall length if you're hoping to create 80-column
output. Some retro assemblers may have hard line length limitations,
which could result in the comment being truncated in generated sources.</p>
<p>A semicolon (';') is placed at the start of the comment. If an assembler
has different conventions, a different delimiter character may be used. You
don't need to include a delimiter explicitly in the comment field.</p>
<p>Comments on platform symbols are read from the platform symbol file, and
cannot be edited from within SourceGen. Comments on project symbols are
stored in the project file, and can be edited with the project symbol
editor.</p>
<h2><a name="long-comment">Edit Long Comment</a></h2>
<p>Long comments can be arbitrarily long and span multiple lines. They
will be word-wrapped at a line width of your choosing. They're always
drawn with a fixed-width font, so you can create ASCII-art diagrams.
Comment delimiters are added automatically at the start of each line.</p>
<p>For a true retro look you can "box" the comment with asterisks. You
can create a full-width row of asterisks by putting a '*' on a line by
itself. (Assembly source generators are allowed to use a character
other than '*' for the output, e.g. they might use a full set of
box outline characters, though that's somewhat against the spirit of
the thing. Regardless, a solo '*' results in a line.)</p>
<p>The bottom window will update automatically as you type, showing what
the output is expected to look like. The actual assembler source output
will depend on features of the target assembler, such as comment
delimiter choices and maximum line length limitations. For example,
Merlin allows a leading '*' to indicate a comment, while cc65 does not,
so cc65 code uses ";*' instead. Because the length limitation affects
the length of the line, not just the comment text, an asterisk-boxed
comment will have one fewer character per line in cc65 output.</p>
<p>Clear the text field to delete the comment.</p>
<p>You can use Ctrl+Enter as a keyboard shortcut for "OK".</p>
<p>The long comment at the very top of the project is special, as it's
not associated with a file offset. If you delete it, you can get it
back by using Edit &gt; Edit Header Comment.</p>
<h2><a name="data-bank">Edit Data Bank (65816 only)</a></h2>
<p>Sets the Data Bank Register (DBR) value for 65816 code. This is used
when matching 16-bit address operands with labels. The new value is
in effect from the line where it's declared to the end of the file, even
across bank boundaries.
If you leave the text field blank, the directive will be removed.</p>
<p>A hexadecimal value from $00 to $ff can be entered directly. As
with other address inputs, a leading '$' is not required. Entering
"K" will set the DBR to the current address, and will automatically
update if you change the address to a different bank.</p>
<p>The pop-up menu has a list of all banks that hold code or data.
To make them easier to identify, each is shown with the label on the
first address in the bank, if any.</p>
<p>While you can override automatically-generated data bank change
directives, you can't remove them individually. You can disable
automatic generation by un-checking "smart PLB handling" in the project
properties.</p>
<p>Because the directive is frequently associated with <code>PLB</code>
instructions, double-clicking on a <code>PLB</code> opcode in the
code list will open the editor.</p>
<h2><a name="note">Edit Note</a></h2>
<p>Notes are similar to long comments, in that they can be arbitrarily
long and span multiple lines. However, because they're never included
in generated output, options like line width formatting and boxing
aren't relevant.</p>
<p>Instead, you can select a highlight color for the note to make it
stand out. You may want to assign certain colors to specific things,
e.g. blue for "I don't know what this is" or green for "this is a
bookmark for the really interesting stuff". The color will be applied
to the note in the code list and in the "Notes" window.</p>
<p>If you don't like the standard colors you can define your own.
You can do this with web RGB syntax, which uses a '#' followed by
two hex digits per channel. For example, bright red is
<code>#ff0000</code>, while teal is <code>#008080</code>. You can
also simply type a color name like "violet" so long as it appears in the
<a href="https://docs.microsoft.com/en-us/dotnet/media/art-color-table.png?view=netframework-4.8">list of Microsoft .NET colors</a>.</p>
<p>Clear the text field to delete the note.</p>
<p>You can use Ctrl+Enter as a keyboard shortcut for "OK".</p>
<h2><a name="project-symbol">Edit Project Symbol</a></h2>
<p>This is used to edit the properties of a project symbol.</p>
<p>Symbols marked as "address" will be applied automatically when an
operand references an address outside the scope of the data file. They
will not be applied to addresses inside the data file. Symbols
marked as "constant" are not applied automatically, and must be
explicitly specified as an operand.</p>
<p>The label must meet the criteria for symbols (see
<a href="intro.html#about-symbols">All About Symbols</a>), and must
not have the same name as another project symbol. It can overlap
with platform symbols and user labels.</p>
<p>The value may be entered in decimal, hexadecimal, or binary. The numeric
base you choose will be remembered, so that the value will be displayed
the same way when used in a .EQ directive.</p>
<p>You can optionally provide a width for address symbols. For example,
if the address is of a two-byte pointer or a 64-byte buffer, you would
set the width field to cause all references to any location in that range
to be set to the symbol. Widths may be entered in hex or decimal. If
the field is left blank, a width of 1 is assumed. Overlapping symbols
are allowed. The width is ignored for constants.</p>
<p>If you enter a comment, it will be placed at the end of the line of
the .EQ directive.</p>
<p>For address symbols that represent a memory-mapped I/O location, it
can be useful to have different symbols for reads and writes. Use
the Read/Write checkboxes to specify the desired behavior.</p>
<h2><a name="lvtable">Create/Edit Local Variable Table</a></h2>
<p><a href="intro.html#local-vars">Local variables</a> are arranged in
tables, which are created at a specific file offset. They must be
associated with a line of code, and are usually placed at the start of
a subroutine.
The "Create Local Variable Table" action creates a new table, and
opens the editor. The "Edit Prior Local Variable Table" searches
for the closest table that appears at or before the selected line,
and edits that.</p>
<p>The editor allows you to create, edit, and delete entries, as well
as move and delete entire tables (though these last two options are not
available when creating a new table). Empty tables are allowed. These
can be useful if the "clear previous" flag is set. If you want to
delete the table, click the "Delete Table" button.</p>
<p>Use the buttons to add, edit, or remove individual variables. Each
variable has a name, a value, a width, and an optional comment. The
standard naming rules for symbols apply. Variables are only used for
zero-page and stack-relative operands, so all values must fall in the
range 0-255. The width may extend one byte past the end (to address $0100)
to allow 16-bit accesses to $ff (particularly useful on 65816).</p>
<p>You can move a table to any offset that is the start of an instruction
and doesn't already have a local variable table present. Click the
"Move Table" button and enter the new offset in hex. You can also click
on the up/down buttons to move to the next valid offset.</p>
</div>
<div id="footer">
<p><a href="index.html">Back to index</a></p>
</div>
</body>
<!-- Copyright 2018 faddenSoft -->
</html>