6502bench SourceGen: Editors

Back to index

Edit Address

This adds a target address directive (".ORG") to the current offset. If you leave the text field blank, the directive will be removed.

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.

There will always be an address directive at the start of the file. Attempts to remove it will be ignored.

Edit Status Flag Override

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.

The 65816 emulation bit, which is not part of the processor status register, may also be set in the editor.

The M, X, and E flags will not be editable unless your CPU configuration is set to 65816.

Edit Label

Sets or clears a label at the selected offset. The label must have the proper form, and not have the same name as another symbol. If you edit an auto-generated label you will be required to change the name.

The label may be marked as local, global, or global and exported. Local labels may be modified by the assembly code generator to have a more convenient form, such as a local loop identifier. Global labels are always output as-is. Exported labels are added to a table that may be imported by other projects (see Working With Multiple Binaries).

Edit Instruction Operand

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.

Explicit Formats

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.

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.

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.

When you select a non-default format option, a "preview" of the formatted operand will be shown.

The MVN and MVP 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.)

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.

Numeric Address References

For operands that are 8-bit, 16-bit, or 24-bit addresses, you can define a symbol for the address as a label or project symbol.

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.

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.

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).

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.)

Local Variable References

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.

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.

Edit Data Operand

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.

If the range is discontiguous, or crosses a visual boundary such as a change in address, a user-specified label, or a long comment or note, 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.

(End-of-line comments do not split a region, and will disappear if they end up inside a multi-byte data item.)

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.

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.

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.

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).

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.

Due to fundamental limitations of the character set, C64 screen code strings cannot be null terminated ($00 is '@').

To avoid burying a label in the middle of a data item, contiguous areas are split at labels. This can sometimes have unexpected effects. For example, this can be formatted as two 16-bit words or one 32-bit word:

         .DD1    $01
         .DD1    $ef
         .DD1    $01
         .DD1    $f0

With a label in the middle, it be formatted as two 16-bit words, but not as a 32-bit word:

         .DD1    $01
         .DD1    $ef
LABEL    .DD1    $01
         .DD1    $f0
CODE     LDA     LABEL

If this is inconvenient, you can add a label at a 32-bit boundary, and reference that instead:

LABEL    .DD1    $01
         .DD1    $ef
         .DD1    $01
         .DD1    $f0
CODE     LDA     LABEL+2

With the label out of the way, the data can be formatted as desired.

Edit Comment

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.

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.

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.

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.

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.

Edit Long Comment

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.

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.)

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.

Clear the text field to delete the comment.

You can use Ctrl+Enter as a keyboard shortcut for "OK".

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 > Edit Header Comment.

Edit Note

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.

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.

Clear the text field to delete the note.

You can use Ctrl+Enter as a keyboard shortcut for "OK".

Edit Project Symbol

This is used to edit the properties of a project symbol.

The label must meet the criteria for symbols (see All About Symbols), and must not have the same name as another project symbol. It can overlap with platform symbols and user labels.

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.

You can optionally provide a width. 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.

If you enter a comment, it will be placed at the end of the line of the .EQ directive.

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.

Create/Edit Local Variable Table

Local variables 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.

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.

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, with their width factored in. So the maximum address for a two-byte pointer is $fe.

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.

Back to index