Most of SourceGen uses standard WPF controls, which get their default
style from the system theme. The main disassembly list uses a
custom style, and always looks like the Windows default theme.
Some people greatly prefer white text on a black background, so we
now provide a way to get that. This also requires muting the colors
used for Notes, since those were chosen to contrast with black text.
This does not affect anything other than the ListView used for
code, because everything else can be set through the Windows
"personalization" interface. We might want to change the way the
Notes window looks though, to avoid having glowing bookmarks on
the side.
The last two tabs in the Edit App Settings dialog have "quick set"
buttons configure all fields for a particular assembler, or reset
them to default values. The previous UI was a little annoying,
because you had to pick something from the combo box and then hit
"set" to push the change. It was also confusing, because if you
came back later the combo box was just set to the first entry, not
the thing you picked last.
Now, picking an entry from the combo box immediately updates all
fields. The combo box selection is set to reflect the actual
contents (so if you set everything just right, the combo box will
change to a specific assembler). If nothing matches, a special
entry labeled "Custom" is selected.
Also, rearranged the tutorial sections in the manual so the
address table formatting comes last, and appears in the local TOC.
If you link to the file without escaping the '#', the browser will
think it's an anchor inside the page. Easier on everyone to just
alter the filename.
Changed the sort order on EQU lines so that constants come before
address definitions. This caused trivial changes to three of the
regression tests.
Added the ability to jump directly to an EQU line when an opcode
is double-clicked on.
If you select a local variable, double-click on a reference entry,
and then hit "back", you aren't taken back to the correct place in
the local variable table. This is annoying if you're trying to
explore how a local variable is used.
The NavStack Location object now has a "line delta" that can be
applied to position the selection correctly. This isn't stable
across undo/redo, but it solves the common cases.
This makes LineListGen's "Top" class redundant, so uses of that have
been replaced with Location.
The Find box now has forward/backward radio buttons. Find Next
searches forward, and Find Previous searches backward, regardless
of the direction of the initial search.
The standard key sequence for "find previous" is Shift+F3. The WPF
ListView has some weird logic that does something like: if you hit
a key, and the selection changes, and the shift key was held down,
then you must have meant to select a range. So Shift+F3 often (but
not always) selects a range. I think this might be fixable if I can
figure out how ListView keeps track of the current keyboard
navigation position (which is not the same as the selection). For
now I'm working around the problem by using Ctrl+F3 to search.
Yay WPF.
Early data sheets listed BRK as one byte, but RTI after a BRK skips
the following byte, effectively making BRK a 2-byte instruction.
Sometimes, such as when diassembling Apple /// SOS code, it's handy
to treat it that way explicitly.
This change makes two-byte BRKs optional, controlled by a checkbox
in the project settings. In the system definitions it defaults to
true for Apple ///, false for all others.
ACME doesn't allow BRK to have an arg, and cc65 only allows it for
65816 code (?), so it's emitted as a hex blob for those assemblers.
Anyone wishing to target those assemblers should stick to 1-byte mode.
Extension scripts have to switch between formatting one byte of
inline data and formatting an instruction with a one-byte operand.
A helper function has been added to the plugin Util class.
To get some regression test coverage, 2022-extension-scripts has
been configured to use two-byte BRK.
Also, added/corrected some SOS constants.
See also issue #44.
The "add platform symbol file" and "add extension script" buttons
create a file dialog with the initial directory set to the
RuntimeData directory inside the SourceGen installation directory.
This is great if you're trying to add a file from the platform
definitions, but annoying if you're trying to add it from the
project directory.
It's really convenient to not have to hunt around though, so now
there are two buttons: one for platform, one for project. The
latter is disabled if the project is new and hasn't been saved yet.
We were changing the control template for lines with long comments
and notes, matching the default Win10 style. This got ugly when a
non-default theme was being used, particularly "dark" themes,
because the long-comment lines looked significantly different from
everything else.
We now fully specify the style for the ListView and ListViewItems,
which means everybody's main window now looks like the default Win10
style. Which is unfortunate, but significantly easier than creating
a full set of theme-specific styles.
We now specify black text for highlighted address/label fields,
because they otherwise become illegible when we apply our background
highlight color. In the Notes window, we set the background of
un-highlighted entries to white, so that we can always read it with
black text.
Addresses issue #50.
Also, updated LZ4FH sample, which needed to have explicit widths on
a couple of zero-page pointers.
Also, updated Zippy sample, which had a ton of unnecessary format
entries for a couple of pointers.
If it's a known function, apply basic numeric formatting to the
various fields. Primarily of value for the pathname and buffer
parameters, which are formatted as addresses.
Also, enable horizontal scrolling in the generic show-text dialog.
Also exercise various formatting options.
Also, fix a bug where the code that applies project/platform symbols
to numeric references was ignoring inline data items.
The current AddressMap is now passed into the plugin manager, which
wraps it in an AddressTranslate object and passes that to the
plugins at Prepare() time. This allows plugins to convert addresses
to offsets, making it possible to format complex structures.
This breaks existing plugins.
If we have a bug, or somebody edits the project file manually, we
can end up with a very wrong string, such as a null-terminated
string that isn't, or a DCI string that has a mix of high and low
ASCII from start to finish. We now check all incoming strings for
validity, and discard any that fail the test. The verification
code is shared with the extension script inline data formatter.
Also, added a comment to an F8-ROM symbol I stumbled over.
Extension scripts (a/k/a "plugins") can now apply any data format
supported by FormatDescriptor to inline data. In particular, it can
now handle variable-length inline strings. The code analyzer
verifies the string structure (e.g. null-terminated strings have
exactly one null byte, at the very end).
Added PluginException to carry an exception back to the plugin code,
for occasions when they're doing something so wrong that we just
want to smack them.
Added test 2022-extension-scripts to exercise the feature.
We were providing platform symbols to plugins through the PlatSym
list, which allowed them to find constants and well-known addresses.
We now pass all project symbols and user labels in as well. The
name "PlatSym" is no longer accurate, so the class has been renamed.
Also, added a bunch of things to the problem list viewer, and
added some more info to the Info panel.
Also, added a minor test to 2011-hinting that does not affect the
output (which is the point).
Handle situation where a symbol wraps around a bank. Updated
2021-external-symbols for that, and to test the behavior when file
data and an external symbol overlap.
The bank-wrap test turned up a bug in Merlin 32. A workaround has
been added.
Updated documentation to explain widths.
Implement multi-byte project/platform symbols by filling out a table
of addresses. Each symbol is "painted" into the table, replacing
an existing entry if the new entry has higher priority. This allows
us to handle overlapping entries, giving boosted priority to platform
symbols that are defined in .sym65 files loaded later.
The bounds on project/platform symbols are now rigidly defined. If
the "nearby" feature is enabled, references to SYM-1 will be picked
up, but we won't go hunting for SYM+1 unless the symbol is at least
two bytes wide.
The cost of adding a symbol to the symbol table is about the same,
but we don't have a quick way to remove a symbol.
Previously, if two platform symbols had the same value, the symbol
with the alphabetically lowest label would win. Now, the symbol
defined in the most-recently-loaded file wins. (If you define two
symbols with the same value in the same file, it's still resolved
alphabetically.) This allows the user to pick the winner by
arranging the load order of the platform symbol files.
Platform symbols now keep a reference to the file ident of the
symbol file that defined them, so we can show the symbols's source
in the Info panel.
These changes altered the behavior of test 2008-address-changes,
which includes some tests on external addresses that are close to
labeled internal addresses. The previous behavior essentially
treated user labels as being 3 bytes wide and extending outside the
file bounds, which was mildly convenient on occasion but felt a
little skanky. (We could do with a way to define external symbols
relative to internal symbols, for things like the source address of
code that gets relocated.)
Also, re-enabled some unit tests.
Also, added a bit of identifying stuff to CrashLog.txt.
If you open the Actions menu when nothing is selected, the "can I
create a local variable table here" method crashes with a bad index
reference.
Issue #48.
If you open the Actions menu when nothing is selected, the "can I
create a local variable table here" method crashes with a bad index
reference.
Issue #48.
Added a Width column to the list in the project symbol editor.
Changed the local variable table editor and the project symbol editor
to use DataGrid instead of ListView. This gets us easy sorting on
arbitrary columns. The previous code was reloading the display list
after every change; now we just add/edit/remove individual items,
which helps keep the list position and selection stable.
The ability to give explicit widths to local variables worked out
pretty well, so we're going to try adding the same thing to project
and platform symbols.
The first step is to allow widths to be specified in platform files,
and set with the project symbol editor. The DefSymbol editor is
also used for local variables, so a bit of dancing is required.
For platform/project symbols the width is optional, and is totally
ignored for constants. (For variables, constants are used for the
StackRel args, so the width is meaningful and required.)
We also now show the symbol's type (address or constant) and width
in the listing. This gets really distracting when overused, so we
only show it when the width is explicitly set. The default width
is 1, which most things will be, so users can make an aesthetic
choice there. (The place where widths make very little sense is when
the symbol represents a code entry point, rather than a data item.)
The maximum width of a local variable is now 256, but it's not
allowed to overlap with other variables or run of the end of the
direct page. The maximum width of a platform/project symbol is
65536, with bank-wrap behavior TBD.
The local variable table editor now refers to stack-relative
constants as such, rather than simply "constant", to make it clear
that it's not just defining an 8-bit constant.
Widths have been added to a handful of Apple II platform defs.
Change + save + undo + change was being treated as non-dirty.
Added link to "export" feature to documentation TOC.
Added keyboard shortcut for high part in data operand editor.
Corrected various things in the tutorial.
Added a blank line after local variable tables. Otherwise they
just sort of blend in with the stuff around them.
Put prefixes before the DOS 3.3 platform symbols.
Added a BAS_HBASH entry. We were getting BAS_HBASL and MON_GBASH
paired up, which looks weird.
Apply a very light tint to the preview section of the Edit Long
Comment dialog, to hint that the window is read-only.
Having underlined blue text everywhere was too noisy. This changes
the CSS style for internal links to be plain black text that gets
blue and underliney when you hover the mouse over it.
Also, added the current date and time to the set of template
substitutions.
HTML output should have had double quotes around internal anchors.
(Chrome and Edge didn't complain, but the w3c validator wasn't
happy.)
Made the text areas in the load-time problem report dialogs
scrollable.
Updated the manual.
The analyzer sometimes runs into things that don't seem right, like
hidden labels or references to non-existent symbols, but has no way
to report them. This adds a problem viewer.
I'm not quite ready to turn this into a real feature, so for now it's
a free-floating window accessed from the debug menu.
Also, updated some documentation.
In a recent survey, three out of four cross assemblers surveyed
recommended not using opcode mnemonics to their patients who use
labels. We now remap labels like "AND" and "jmp", using the label
map that's part of the label localizer.
We skip the step for Merlin 32, which is perfectly happy to assemble
"JMP JMP JMP".
Also, fixed a bug in MaskLeadingUnderscores that could hang the
source generator thread.
Most assemblers end local label scope when a global label is
encountered. cc65 takes this one step further by ending local label
scope when constants or variables are defined. So, if we have a
variable table with a nonzero number of entries, we want to create
a fake global label at that point to end the scope.
Merlin 32 won't let you write " LDA #',' ". For some reason the
comma causes an error. IGenerator now has a "tweak operand format"
interface that lets us fix that.
Split "edit local variable table" into "create" and "edit prior".
The motivation is to allow the user to make changes to the most
recently defined table without having to go search for it. Having
table creation be an explicit action, rather than something that
just happens if you edit a table that isn't there, feels reasonable.
Show table offset in LV table edit dialog, so if you really want
to go find it there's a (clumsy) way to do so.
Increased the maximum width of a variable from 4 to 8. (This is
entirely arbitrary.)
Typing a long comment in the project symbol editor caused the
window to expand, which wasn't intended. Use the mono font in
the comment editor. Set the focus to the OK button after creating
or editing a project property. Show constant vs. address in the
info panel when an EQU directive is selected.
Changing an ASCII character operand back to default was going
through a path that tried to resolve low vs. high ASCII, which
isn't useful when you're removing the item. The root of the problem
was that the "default" button wasn't properly resetting the UI.
Also, updated keyboard shortcuts to be in sync with the instruction
operand editor.
It felt a little weird tying it to the asm generation setting,
so now it's just another checkbox in the export options.
Implemented the feature for plain text output. Did some rearranging
in the code. Fixed suppression of Notes in text and CSV.
The functions started by trying to pad a column out to a width,
then changed to pad things to a certain length. What they really
should be doing is padding the start of an entry to a specified
column. This is much more natural and avoids a trim operation.
The only change to the output is to ORG statements from the HTML
exporter, which are now formatted correctly.
We weren't escaping '<', '>', and '&', which caused browsers to get
very confused. Browsers seem to prefer <PRE> to <CODE> for long
blocks of text, so switch to that.
Also, added support for putting long labels on their own lines in
the HTML output.
Also, fixed some unescaped angle brackets in the manual.
Also, tweaked the edit instruction operand a bit more.
If you set things up just right, it's possible for flag status
changes to fail to get merged.
Added a regression test to 1003-flags-and-branches.
Also, tweaked the instruction operand editor to be a bit smoother
from the keyboard: added alt-key shortcuts, and put the focus on the
OK button after creating/editing a label so you can just hit the
return key twice.
I was using the plain names, but when you've got symbols like
READ and WAIT it's too easy to have a conflict and it's not plainly
obvious where something came from. Now all monitor symbols begin
with MON_, and Applesoft symbols begin with BAS_.
The Amper-fdraw example ended up with a few broken symbol refs,
because it was created before project/platform symbols followed the
"nearby" rules, and was explicitly naming LINNUM and AMPERV. I
switched the operands to default, and they now auto-format correctly.
I added a few more entries to Applesoft while I was at it.
If a line has a comment with a cycle count and nothing else, it was
getting an extra space or two on the end.
Also, added a few end-of-line comments to the 2020 test to show how
they interact with the cycle counts.
Cycle counting is CPU-specific. The 2020 test exercises the
65816, but there are things unique to 6502 and 65C02 that should
also be checked if we want to be thorough.
No changes to the test itself.
A ".dd2 <address>" item would get linked to an internal label, but
references to external addresses weren't doing the appropriate
search through the platform/project symbol list.
This change altered the output of the 2019-local-variables test.
The previous behavior was restored by disabling "nearby" symbol
matching in the project properties.
Updated the "lookup symbol by address" function to ignore local
variables.
Also, minor updates to Applesoft and F8-ROM symbol tables.
I ran into a non-split table of 16-bit addresses, each of which
was (address-1) for a code location. I wanted to create a label,
add a code hint, and set the operand for each one, but there's no
easy way to do that.
It turns out the split-address table formatter can be made to work
for non-split tables with just a few minor changes.
It's possible to define multiple project symbols with the same
address. The way to resolve the ambiguity is to explicitly
reference the desired symbol from the operand. This was the
default behavior of the "create project symbol" shortcut in the
previous version.
It's rarely necessary, and it can get ugly if you rename a project
symbol, because we don't refactor operands in that case.
If you play games with code hints you can create a data operand that
overlaps with code. This causes problems (see issue #45). We now
check for that situation and ignore overlapping data descriptors.
Added a regression test to 2011-hinting.
Also, removed "include symbol table" from export dialog. You can
exclude the table by removing it from the template, which right
now you'd need to do anyway to get rid of the H2 header and other
framing. To make this work correctly as an option we'd need to
parse the "div" in the template file and strip the whole section,
or split the template into multiple parts that get included as
needed. Not worth doing the work until we're sure it matters.
If a symbol is defined at <addr>, and we counter STA <addr>-1,Y,
we want to use the symbol in the operand. This worked for labels
but not project/platform symbols.
Also, fixed a crash that happened if you tried to delete an auto
label.
This feature "exports" the source code in the same format it appears
in on screen. The five columns on the left are optional, the four
on the right can be resized. Exported text can span the entire file
or just the current selection. Output can be plain text or CSV.
(I still haven't figured out a good use for CSV.)
The old copy-to-clipboard function is now implemented via the export
mechanism.
This adds a new clipboard mode that includes all columns. Potentially
useful for filing bug reports against SourceGen.
Ported the column width stuff from EditAppSettings, which it turns
out can be simplified slightly.
Moved the clipboard copy code out into its own class.
Disabled "File > Print", which has never done anything and isn't
likely to do anything in the near future.
Also, added a note to 2019-local-variables about a test case it
should probably have.
Updated the manual, and changed tutorial #2 to use local variables
for pointers.
If the symbol text box isn't empty, use the string as the initial
value for the Label when creating a new project property.
Fixed a crash when editing a project property.
Implemented local variable editing. Operands that have a local
variable reference, or are eligible to have one, can now be edited
directly from the instruction operand edit dialog.
Also, updated the code list double-click handler so that, if you
double-click on the opcode of an instruction that uses a local
variable reference, the selection and view will jump to the place
where that variable was defined.
Also, tweaked the way the References window refers to references
to an address that didn't use a symbol at that address. Updated
the explanation in the manual, which was a bit confusing.
Also, fixed some odds and ends in the manual.
Also, fixed a nasty infinite recursion bug (issue #47).
Rearrange the UI elements, and convert the code-behind to a more
XAML-style form. The basic stuff works, but the old "shortcut"
system is still in the process of being replaced.
The code that checked to see if a data target was inside a data
operand wasn't going all the way back to the start of the file.
It was also failing to stop when it should, wasting time.
The anattrib validation method has code that avoids a false-positive
on certain complex embedded instruction arrangements. This was also
preventing it from seeing a transition from a data area to the
middle of an instruction (caused by issue #45).
Unlike 64tass and Merlin, which allow you to redefine symbols, ACME
uses "zones" that provide scope for local variables. This means
that, at the point of a local variable table definition, we have to
start a new zone and output the full set of active symbols, not just
the newly-defined ones. (If you set the "clear previous" flag in
the LvTable there's no difference.)
We could do a bit better by only outputting the symbols that are
actually used within the zone, similar to what we do for global
project/platform symbols, but that's a bunch of work for questionable
benefit.
After thrashing around a bit, I had to choose between making the
uniquifier more complicated, or making de-duplication a separate
step. Since I don't really expect duplicates to be a thing, I went
with the latter.
Updated the regression test.
This hits most of the edge cases, but doesn't exercise the two
duplicate name situations (var name same as user label, var name
same as project/platform symbol).
Also, fixed a bug in the EditDefSymbol uniqueness check where it
was comparing a symbol to itself.
Variables are now handled properly end-to-end, except for label
uniquification. So cc65 and ACME can't yet handle a file that
redefines a local variable.
This required a bunch of plumbing, but I think it came out okay.
Still primarily ascending numeric order, but now we use the symbol
type as the secondary sort instead of the label.
Also, fix References crash on first line of empty var table.
We now generate FormatDescriptors with WeakSymbolRefs for direct
page references that match variable table entries.
LocalVariableTable got a rewrite. We need to be unique in both
name and address, but for the address we have to take the width into
account as well. We also want to sort the display by address
rather than name. (Some people might want it sorted by name, but
we can worry about that some other time.)
Updated the DefSymbol editor to require value uniqueness. Note
addresses and constants exist in separate namespaces.
The various symbols are added to the SymbolTable so that uniqueness
checks work correctly. This also allows the operand generation to
appear to work, but it doesn't yet handle redefinition of symbols.
Multi-line item, with one .eq line per variable definition. Add
one header line if "clear previous" is set.
Also, limit variable values to 0-255 in the editor. This is
somewhat arbitrary, but I think a focus on DP is useful.
This involved adding a list to the DisasmProject, creating a new
UndoableChange type, and writing the project file serialization
code. While doing the latter I realized that the new Width field
was redundant with the FormatDescriptor Length field, and removed it.
I added a placeholder line type, but we're not yet showing the
table in the display list. (To edit the tables you just have to
know where they are.)
The table editor is now editing the table, and the DefSymbol editor
now asks for the Width data when editing a local var.
This also moves EditDefSymbol closer to proper WPF style, with
bound properties for the input fields.
No changes yet to serialization or analysis.
Fixed a minor bug in GenerateLineList that would cause a blank line
to disappear under certain circumstances. Harmless, but odd.
Added a width property to DefSymbol.
Updated comments.
Previously, we used the default character encoding from the project
properties to determine how strings and character constants in the
entire source file should be encoded. Now we switch between
encodings as needed. The default character encoding is no longer
relevant.
High ASCII is now an actual encoding, rather than acting like ASCII
that sometimes doesn't work. Because we can do high ASCII character
operands with "| $80", we don't output a .enc to switch from ASCII
to high ASCII unless we need to generate a string. (If we're
already in high ASCII mode, the "| $80" isn't required but won't
hurt anything.)
We now do a scan up front to see if ASCII or high ASCII is needed,
and only output the .cdefs for the encodings that are actually used.
The only gap in the matrix is high ASCII DCI strings -- the ".shift"
pseudo-op rejects text if the string doesn't start with the high
bit clear.
I didn't think it made sense, but I found something that used it,
so apparently it's a thing. This updates the operand editor to
let you choose PETSCII+DCI, and updates the assemblers to handle
it correctly (really just 64tass, since the others either don't
have a DCI directive or don't deal with PETSCII at all).
Changed the char-encoding sample from "bad dcI" to "pet dcI", and
updated the documentation.
The documentation for 64tass says you're required to pass "--ascii"
when the source file is ASCII (as opposed to PETSCII). We were
ignoring this, but it turns out that everything works a bit better
if we don't.
So we now pass "--ascii" on the command line, and add a two-line
character encoding definition to every file that is generated with
ASCII as the default encoding. The sg_petscii and sg_screen
encodings go away, as PETSCII is now the default, and we can use the
built-in "screen" encoding.
String operands used to be simple -- each line had 62 characters
plus two hard-coded non-ASCII delimiters -- but now we're mixing
character and hex data, so we can't use simple math to tell where
the lines will break. We want to render them and keep the result
around until some dependency changes, e.g. different delimiters
or a change to the pseudo-op table.
Also, cleaned up LineListGen a little. It had some methods that
were declared static because they were expected to be shared, but
that never happened.
Also, fixed a bug in GatherEntityCounts where multi-line items were
being scanned multiple times.
The PseudoOpNames class is increasingly being used in situations
where mutability is undesirable. This change makes instances
immutable, eliminating the Copy() method and adding a constructor
that takes a Dictionary. The serialization code now operates on a
Dictionary instead of the class properties, but the JSON encoding is
identical, so this doesn't invalidate app settings file data.
Added an equality test to PseudoOpNames. In LineListGen, don't
reset the line list if the names haven't actually changed.
Use a table lookup for C64 character conversions. I figure that
should be faster than multiple conditionals on a modern x64 system.
Fixed a 64tass generator issue where we tried to query project
properties in a call that might not have a project available
(specifically, getting FormatConfig values out of the generator for
use in the "quick set" buttons for Display Format).
Fixed a regression test harness issue where, if the assembler reported
success but didn't actually generate output, an exception would be
thrown that halted the tests.
Increased the width of text entry fields on the Pseudo-Op tab of app
settings. The previous 8-character limit wasn't wide enough to hold
ACME's "!pseudopc". Also, use TrimEnd() to remove trailing spaces
(leading spaces are still allowed).
In the last couple of months, Win10 started stalling for a fraction
of a second when executing assemblers. It doesn't do this every
time; mostly it happens if it has been a while since the assembler
was run. My guess is this has to do with changes to the built-in
malware scanner. Whatever the case, we now change the mouse pointer
to a wait cursor while updating the assembler version cache.
All tests use the same data file and nearly the same project file.
The only difference is the default text encoding property setting.
For "-a" it's ASCII, for "-p" it's PETSCII, for "-s" it's C64 screen
code. Right now this only affects the code generated for 64tass.
The test itself is a collection of strings and characters in the
supported character encodings. How these are handled varies
significantly between assemblers.
The 64tass generator now uses the "default text encoding" project
property to determine how readable text should be encoded. For
example, if the property is set to PETSCII, an ASCII-to-PETSCII
encoding table is generated at the top of the output file.
Both dialogs got a couple extra radio buttons for selection of
single character operands. The data operand editor got a combo box
that lets you specify how it scans for viable strings.
Various string scanning methods were made more generic. This got a
little strange with auto-detection of low/high ASCII, but that was
mostly a matter of keeping the previous code around as a special
case.
Made C64 Screen Code DCI strings a thing that works.
We weren't checking to see if character operands matched their
delimiters, so bad code like "LDA #'''" was being generated.
There wasn't a test for this in 2006-operand-formats, so the test
has been updated with single and double quotes in low and high ASCII.
A delimiter definition is four strings (prefix, open, close, suffix)
that are concatenated with the character or string data to form an
operand. A delimiter set is a collection of delimiter definitions,
with separate entries for each character encoding.
This is a convenient way to configure Formatter objects, import and
export data from the app settings file, and manage the UI needed to
allow the user to customize how things look.
The full set of options didn't fit on the first app settings tab, so
there's now a separate tab just for specifying character and string
delimiters. (This might be overkill, but there are various plausible
scenarios that make use of it.)
The delimiters for on-screen display of strings can now be
configured.
The previous functions just grabbed 62 characters and slapped quotes
on the ends, but that doesn't work if we want to show strings with
embedded control characters. This change replaces the simple
formatter with the one used to generate assembly source code. This
increases the cost of refreshing the display list, so a cache will
need to be added in a future change.
Converters for C64 PETSCII and C64 Screen Code have been defined.
The results of changing the auto-scan encoding can now be viewed.
The string operand formatter was using a single delimiter, but for
the on-screen version we want open-quote and close-quote, and might
want to identify some encodings with a prefix. The formatter now
takes a class that defines the various parts. (It might be worth
replacing the delimiter patterns recently added for single-character
operands with this, so we don't have two mechanisms for very nearly
the same thing.)
While working on this change I remembered why there were two kinds
of "reverse" in the old Merlin 32 string operand generator: what you
want for assembly code is different from what you want on screen.
The ReverseMode enum has been resurrected.
The code that searches for character strings in uncategorized data
now recognizes the C64 encodings when selected in the project
properties.
The new code avoids some redundant comparisons when runs of
printable characters are found. I suspect the new implementation
loses on overall performance because we're now calling through
delegates instead of testing characters directly, but I haven't
tested for that.
It's not quite the same as the character encoding -- sometimes we
want a mix of things -- so it gets its own enum. The value is
saved to the project file, but not actually used yet.
Also, moved some combo box strings into XAML resources.
Remove left/right arrow PNGs. Remove duplicate copies of icon.
Tweak some comments. Set application icon. Adjust padding on
group boxes in first tab of app settings.
The previous code output a character in single-quotes if it was
standard ASCII, double-quotes if high ASCII, or hex if it was neither
of those. If a flag was set, high ASCII would also be output as
hex.
The new system takes the character value and an encoding identifier.
The identifier selects the character converter and delimiter
pattern, and puts the two together to generate the operand.
While doing this I realized that I could trivially support high
ASCII character arguments in all assemblers by setting the delimiter
pattern to "'#' | $80".
In FormatDescriptor, I had previously renamed the "Ascii" sub-type
"LowAscii" so it wouldn't be confused, but I dislike filling the
project file with "LowAscii" when "Ascii" is more accurate and less
confusing. So I switched it back, and we now check the project
file version number when deciding what to do with an ASCII item.
The CharEncoding tests/converters were also renamed.
Moved the default delimiter patterns to the string table.
Widened the delimiter pattern input fields slightly. Added a read-
only TextBox with assorted non-typewriter quotes and things so
people have something to copy text from.
High ASCII and other encodings will be noted in the operand field,
not the opcode, so we no longer need these.
This removes the six input fields from the Pseudo-Op tab of app
settings. Values were stored as a serialized class in settings,
which generally works correctly as far as forward/backward
compatibility goes, so no worries there.
This also adds four "delimiter pattern" fields to the Code View tab,
allowing the user to customize how encoded strings are marked up
for the code list. The values aren't actually used yet.
Also, fixed an issue where changes to text fields on the Pseudo-Op
tab weren't raising the dirty flag.
The plugin objects are MarshalByRefObject stubs, which means they
don't actually implement the interfaces we're checking for. There's
some additional overhead to do the interface check. We can avoid
it by doing the interface queries during initialization, and just
checking some bit flags later on.
Also, in the extension script info window, show a list of
implemented interfaces.
We've been treating ASCII strings and instruction/data operands as
ambiguous, resolving low vs. high when generating output for the
display or assembler. This change splits it into two separate
formats, simplifying output generation.
The UI will continue to treat low/high ASCII as as single thing,
selecting the format appropriately based on the data. There's no
reason to have two radio buttons that are never both enabled.
The data operand string functions need some additional work, but
that overlaps substantially with the upcoming PETSCII changes, so
for now all strings set by the data operand editor are low ASCII.
The file format has changed again, but since there hasn't been a
release since the previous change, I'm leaving the file format
at v2. Code has been added to resolve the ASCII mode when loading
a v1 project file.
This removes some complexity from the assembly code generators.
DCI is handled with the ".shift" pseudo-op. The .null, .ptext,
and .shift operators all work correctly with escaped characters,
so we no longer redo those.
This generalizes the string pseudo-operand formatter, moving it into
the Asm65 library. The assembly source generators have been updated
to use it. This makes the individual generators simpler, and by
virtue of avoiding "test runs" should make them slightly faster.
This also introduces byte-to-character converters, though we're
currently still only supporting low/high ASCII.
Regression test output is unchanged.
This was the result of the earlier change to eliminate "reverse DCI"
strings. On further examination, it doesn't seem like we can do
much better than a hex dump without more work than the situation
merits. So hex dump it is.
During a discussion with the cc65 developers, I became convinced that
generating "MVN $01,$02" is wrong, and "MVN #$01,#$02" is correct.
64tass, cc65, and Merlin 32 all accept this syntax; only ACME does
not. Operands without a leading '#' should be treated as 24-bit
values, and have the bank byte extracted.
This change updates the on-screen display and assembled output to
include the '#'. The ACME generator uses a Quirk to suppress the
hash mark. (It doesn't currently accept values larger than 8 bits,
so there's no ambiguity.)
We used to use type="String", with the sub-type indicating whether
the string was null-terminated, prefixed with a length, or whatever.
This didn't leave much room for specifying a character encoding,
which is orthogonal to the sub-type.
What we actually want is to have the type specify the string type,
and then have the sub-type determine the character encoding. These
sub-types can also be used with the Numeric type to specify the
encoding of character operands.
This change updates the enum definitions and the various bits of
code that use them, but does not add any code for working with
non-ASCII character encodings.
The project file version number was incremented to 2, since the new
FormatDescriptor serialization is mildly incompatible with the old.
(Won't explode, but it'll post a complaint and ignore the stuff
it doesn't recognize.)
While I was at it, I finished removing DciReverse. It's still part
of the 2005-string-types regression test, which currently fails
because the generated source doesn't match.
There's no easy way to make non-zero-bank 65816 code work, so I'm
punting and just generating a whole-file hex dump for those. This
renders tests 2007 and 2009 useless, so I'm hesitant to claim that
ACME support is fully functional.
WDM <arg> now works. MVN/MVP are still broken. Correct code is
generated for whichever version of the assembler is configured.
Regression tests updated for new version.
Also, fixed a UI bug where manual edits to the assembler path were
being ignored.
The 65816 definition makes it a two-byte instruction, like COP. On
the 6502 it acted like a two-byte instruction, but in practice very
few assemblers treat it that way. Very few humans, for that matter.
So it's now treated as a single byte instruction, with the following
byte encoded as a data value.
Instead of providing no-op CheckJsr/CheckJsl, plugins now declare
which calls they support by defining interfaces on the plugin class.
I added a CheckBrk call for code like Apple /// SOS calls, which
use BRK as an OS call mechanism. The formatting doesn't work quite
right yet because I've been treating BRK as a two-byte instruction.
Hardly anything else does, and I think it's time I stopped (but not
in this commit).
Note: THIS BREAKS ALL PLUGINS that use the inline JSR/JSL feature,
which is pretty much all of them.
- Updated the tutorial to track changes to WPF, and to clarify
existing content.
- Fixed Ctrl+H Ctrl+C, which was getting masked by the Copy command
handler.
- Fixed initial selection of address in Set Address.
- MakeDist now copies CommonWPF.dll.
- Spent a bunch of time tracking down a null-pointer deref that only
happened when you didn't start with a config file. Fixed.
- The NPE was causing the program to exit without any sort of useful
diagnostic, so I added an uncaught exception handler that writes
the crash to a text file in the current directory.
- Added a trace listener definition to App.config that writes log
messages to a file, but it can't generally be enabled at runtime
because you can't write files from inside the sandbox. So it's
there but commented out.
- Made the initial size of the main window a little wider.
The filtering uses the DataGrid View filtering mechanism. The
built-in sorting only operates on a single column, and we really
want a secondary sort on label when the type is used as the key,
so we provide a custom sort method.
SourceGen creates "auto" labels when it finds a reference to an
address that doesn't have a label associated with it. The label for
address $1234 would be "L1234". This change allows the project to
specify alternative label naming conventions, annotating them with
information from the cross-reference data. For example, a subroutine
entry point (i.e. the target of a JSR) would be "S_1234". (The
underscore was added to avoid confusion when an annotation letter
is the same as a hex digit.)
Also, tweaked the way the preferred clipboard line format is stored
in the settings file (was an integer, now an enumeration string).
In the cross-reference table we now indicate whether the reference
source is doing a read, write, read-modify-write, branch, subroutine
call, is just referencing the address, or is part of the data.
This worked, sort of. The problem is that SourceGen will revert to
hex output in certain situations, such as a broken symbolic
reference. There happens to be one in the ZIPPY example, and it's
on a relative branch.
The goal with the segment stuff is to allow cc65 to treat the
source as relocatable code. In that context, a relative branch to
an absolute address doesn't make any sense, so the assembler reports
a range error.
We don't currently have a mechanism that guarantees no references
are broken (and no affordance for finding them), so we can't make
this mode the default yet.
Instead, we continue to use the generic config, but generate the
correct set of lines as comments.
(issue #39)
The system configuration you get with "-t none" works for smaller
files but fails for larger ones. This updates the generator to
produce a source file and linker script pair. (I kinda saw this
one coming -- it's why the gen/asm dialog has a combo box for the
file preview -- so it didn't require that much work.)
This currently generates a fixed script for a generic system with
64KiB of RAM, using .ORGs to set the addresses as before.
With this change, assembling a file with 65536 NOPs succeeds.
(issue #39)
In the data operand edit section, walk through selecting a single
byte vs. multiple bytes when you want to set a multi-byte format.
(inspired by issue #41)
If you double-click on the opcode of "JSR label", the code view
selection jumps to the label. This now works for partial operands,
e.g. "LDA #<label".
Some changes to the find-label-offset code affected the cc65 "is it
a forward reference to a direct-page label" logic. The regression
test now correctly identifies an instruction that refers to itself
as not being a forward reference.
The cc65 assembler runs in a single pass, which means forward
address references default to 16 bits. For zero-page references
we have to add an explicit width disambiguator. (This is an
unusual situation that only occurs if you have a zero-page .ORG
in the file after code that references it.)
With this change, 2014-label-dp passes, and no other regression
tests were affected.
(issue #40)
The 2014-label-dp test now passes. Prior regression tests are
unaffected.
Also, renamed an IGenerator interface to more accurately reflect
its role.
(issue #37)
This is primarily to exercise a Merlin 32 failure (issue #37).
However, it also exercises a problem with cc65 (issue #40).
Currently, only 64tass can assemble this project.
Change 80da6c replaced \u23e9 (black right-pointing double triangle)
with a downward-pointing triangle, because it didn't render under
Win7 or Linux. It was also being used in the "info" window. This
change replaces that occurrence with a right-pointing triangle.
To avoid confusing the assembler, expressions with a leading
parenthesis like "(foo & $ffff) + 1" are prefixed with a "0+". This
is not necessary if the operand begins with a '#'.
(issue #16)
This adds a null check on the dfd argument in FormatDataOp() to see
if we can prevent a crash. The opcode/operand are presented as
"!FAILED!" to make it obvious to the user that something has gone
wrong. Hopefully this will allow capture of a project that exhibits
the problem.
Before you could choose between Merlin-style and generic. Now
there's a combo box that lets you choose Merlin, cc65, or
"common", the latter being used for 64tass.
We now insert parenthesis as needed. This can cause problems in
some situations, so we always prefix parenthetical expressions with
"0+", which looks goofy and is unnecessary for immediate operands.
But it does generate working source code.
Renamed the "simple" expression mode to "common", as it's not
particularly simple but is what you'd expect most assemblers to do.
(OTOH, life has been full of surprises.)
(issue #16)
Gave cc65 its own expression generator, as the precedence table seems
atypical if not unique. Configured 64tass to use the "simple"
expression mode.
Added some operations on a 32-bit constant to 2007-labels-and-symbols
to exercise the current worst-case expression (shift + AND + add).
Tweaked the Merlin expression generator to handle it.
(issue #16)
Correctly handle pre-existing underscores and avoidance of
"reserved" labels.
Also, add more underscores to 2012-label-localizer to exercise
the code.
(issue #16)
Most tests pass, but 2007-labels-and-symbols fails because the
expressions recognized by 64tass don't match up with either of the
other assemblers.
This is currently using a workaround for the local label syntax.
64tass uses '_' as the prefix, which is unfortunate since SourceGen
explicitly allowed underscores in labels. (So does 64tass for that
matter, but it treats labels specially when the '_' comes first.)
We will need to rename any non-local user labels that start with '_'.
(issue #16)
We were using \u23e9, BLACK RIGHT-POINTING DOUBLE TRIANGLE, but
neither Win7 SP1 nor Linux was able to display the glyph. It also
gets all puffy in web browsers. We now use \u25bc, BLACK
DOWN-POINTING TRIANGLE, which seems to work everywhere. It also
feels more appropriate, because it appears next to the "containing"
opcode, with the embedded instruction appearing on the following
line.
Changed the "quick config" buttons for the asm config and pseudo-op
tabs into a drop-list and "set" button. The default values for
each assembler are now defined in the Asm*.cs file, rather than in
the settings code.
Rather than have each assembler get its own app config string for
the cross-assembler executable, we now have a collection of per-
assembler config items, of which the executable path name is one
member. The Asm Config tab has an auto-generated pop-up to select
the assembler.
The per-assembler settings block is serialized with the rather
unpleasant JSON-in-JSON approach, but nobody should have to look
at it.
This also adds assembler-specific column widths to the settings
dialog, though they aren't actually used yet.
Each supported assembler has an IGenerator interface and an
IAssembler interface. They're still two separate classes, but now
both are implemented in the same source file. (They'll probably
stay separate classes, since the two have little interaction.)
I'm keeping the "Asm*" filename. Seems the more natural fit.
Also, changed AssemblerInfo to try to get all assembler-specific
stuff into a single table.
Most of the decorative items associated with a file offset are
placed before the item in the display list, and given a span of
zero. This yields the correct behavior in a binary search: an
exact match finds the decorative item (e.g. a blank line), while a
match partway into the instruction or multi-byte data item causes
the binary search to move on to the next line, where it's resolved.
The problem is that we were adding a blank line *after* instructions
in the no-continue case. If the binary search found the blank line
before it found the instruction, it would guess "too high" rather
than "too low", and miss the actual instruction line.
We now set a flag and add the blank line as part of the following
item. We do a little dance at the start to ensure that the blank
line doesn't disappear during a partial update.
There's a common convention for using one of the thumb buttons on
the side of the mouse to mean "back", notably in web browsers. I
keep hitting it without thinking about it, so let's just make it
official.
When you edit the operand of an instruction that targets an in-file
address, you're given the opportunity to specify a shortcut that
applies the symbol to the instruction's target address in addition
to or instead of defining a weak symbol reference on the instruction
being edited.
This didn't work right for operands with adjustments, e.g. the store
instructions in self-modifying code. It put the label at the
unadjusted offset, which does nothing useful.
We now correctly back up to the start of the instruction or multi-
byte data area.
Change acf19870 fixed one bug but introduced another: the call to
update the highlight happened before the selection was restored, so
it could potentially refer to a line that no longer existed.
Worked around two crashes in Mono 5.16's WinForms implementation.
(See mono/mono#11070 for the details.)
Still very unstable, but it no longer crashes on startup.
Also, tweaked the "about" box title.
There are some useful interactions between C/N and maybe Z. Added
a quick test to 1003-flags-and-branches.
Also, updated the 2008-address-changes tests. Change b37d3dba
extended the nearby-target range of out-of-file symbols by one, so
one line that didn't get an operand label now does.
Once upon a time, symbol files and extension scripts could only be
defined in the RuntimeData directory, so having the documentation
there made sense. Since both of these things can now be defined in
project directories, the documentation belongs in the manual.
(issue #27)
If PTR is defined as an external symbol, we were automatically
symbol-ifying PTR+1. Now we also symbolify PTR+2. This helps with
24-bit pointers on the 65816, and 16-bit "jump vectors", where the
address is preceded by a JMP opcode.
Removed the "AMPERV_" symbol I added to make the tutorial look
right.
The instruction operand editor and data operand editor are very
different, but there's no need to impose that distinction on the
user. They want to edit the operand either way. We now provide a
single "edit operand" menu item, and open the appropriate dialog
based on what they have selected.
This uses Ctrl+O as the keyboard shortcut, stealing it from
File > Open.
(issue #11)
Two changes:
(1) Code and data hints are now only applied to the first byte on
each selected line. This allows you to slap a code hint on a
string without lighting up the whole string. Inline-data hints
and hint removal work as before.
(2) Added a menu item (with Ctrl+D as shortcut) to toggle the state
of the uncategorized data analyzer. This makes it easy to turn
off the feature that put the code into a string in the first
place.
Every once in a while, SourceGen will become unresponsive when it
tries to show a MessageBox. In the debugger you can see the GC
running frantically, but the stack trace is just sitting on the
MessageBox show call. Apparently, if you don't specify a parent
window argument, the MessageBox will occasionally end up behind
everything else, and you can get stuck.
I'm not sure what the GC frenzy is about, or whether this will fix
what I'm seeing, but it's easy to do and might solve the problem.
cf. https://stackoverflow.com/q/3467403/294248
Don't enable OK unless at least one address is valid.
Don't apply code hints unless asked.
Rename a couple of things for clarity.
Add documentation to manual.
(issue #10)
Allows specification of table data in various ways, for 16-bit and
24-bit addresses. Shows a preview so you can see if the addresses
look about right. Adds permanent labels at target offsets if none
are present. Optionally sets code hints.
Works beautifully on the A2-Amper-fdraw example, but needs some
additional testing, documentation, etc. Dialog is more complicated
that I would have liked, mostly because of 65816 support, but I
think it'll do.
(issue #10)
If we set the length word to assemble at address zero, the rest of
the code will try to use it as a zero-page label, so don't do that.
Instead, we use the start address, creating an overlapping region.
Easy enough to edit if that's undesirable.
(issue #23)
First is always at zero, second is at the address. This puts an
ORG directive right at the start of the code, and avoids potentially
assembler-specific wrap-around behavior when the desired load
address is $0000 or $0001.
(issue #23)
If set, the first word of the file is used to set the load address.
The initial code entry hint is placed at offset +000002 instead of
the start of the file.
Set it to true for the C64 system definition.
(issue #23)
These *almost* match what cc65 has, and are accepted as primary or
aliases by 64tass.
This combines the LAX and LXA operations. LXA is the immediate
form of LAX, and behaves somewhat differently (and is unstable).
I was treating them as two separate operations with independent
mnemonics, but that doesn't seem to be the preferred way to
handle it.
The cc65 generator wasn't generating LAX before; now it does. This
required nudging the width disambiguator, as LAX is a second
example of an instruction with both DP,Y and ABS,Y operands.
(issue #20)
It's possible to have format descriptors on instructions that are
left over from when the bytes were treated as data. Single-byte
formats were being allowed on single-byte instructions, which
confused things later when the code tried to apply the format to
an instruction with no operand.
The Symbols window showed Type-Name-Value, which feels like the
natural order. However, the Value field has a narrow max width,
while the Name field can get pretty long. It makes more sense to
let Name fill out to the right edge, allowing the user to scroll
horizontally to view longer-than-usual names.
Also, noticed that the column sort preference wasn't being
restored. Fixed that.
(Issue #12)
We were generating code for > 2.17, with various bug fixes, but
since that's not shipping yet it won't be usable by anybody who
doesn't have a tip-of-tree cc65 installation.
If the main window was maximized before, maximize it when we
restart.
Changed the preferences to record the width of the right panels,
rather than the splitter distance, which is actually the width of
the middle of the screen.
Seems to work correctly on my non-uniform multi-monitor setup. I
added a check to confirm that the middle of the title bar falls in
the working area of the screen that WinForms thinks we're in, so it
shouldn't be possible to "lose" a window off the edge by dragging
it or by changing screen resolutions.
Updated the RuntimeData directory finder to work, and made the
stuff that crashed when the directory wasn't found crash in less
obvious ways. Under Mono+Linux it still falls over with some
complaints about ListViews. This will need some work.
The code that tries to make the last column of the Symbols and
References windows exactly fit the window goes into an infinite
loop for everybody except me. I'm removing the feature for now.