acme/docs/Errors.txt

507 lines
19 KiB
Plaintext
Raw Normal View History

ACME
...the ACME Crossassembler for Multiple Environments
--- error messages ---
Here's a sorted list of all error messages ACME can give, possible
reasons and what you can do to sort it out.
----------------------------------------------------------------------
Section: Errors on startup
----------------------------------------------------------------------
Cannot open toplevel file "FILENAME".
Maybe you mistyped its name?
Error in CLI arguments: ...
There are several of these errors, but they should be quite self-
explanatory.
----------------------------------------------------------------------
Section: Warnings during assembly
----------------------------------------------------------------------
!warn: ...
This is given when the pseudo opcode "!warn" is executed. The
actual message varies according to the pseudo opcode's arguments.
Assembling buggy JMP($xxff) instruction
The original 6502 processor has a bug: When executing an indirect
JMP instruction where the low byte of the argument equals $ff, it
fetches the high byte of the jump target address not from memory
location ARGUMENT + 1, but from ARGUMENT - 255. Therefore ACME
issues this warning if you are about to generate such an
instruction.
Note that this warning is only given for CPU types 6502 and 6510,
because 65c02 and 65816 have been fixed in this respect.
Assembling unstable ANE #NONZERO instruction
Assembling unstable LXA #NONZERO instruction
These warnings are only ever given for CPU type 6510. ANE and LXA
are undocumented ("illegal") opcodes of this CPU, and they only
work reliably if the argument is zero or the accumulator contains
0xff.
Therefore ACME issues these warnings if it is about to generate
these instructions with a non-zero argument.
Binary literal without any digits.
Hex literal without any digits.
Octal literal without any digits.
A special literal was started, but then no digits followed. Expect
this to become an error in future!
Bug in ACME, code follows
A situation has been encountered implying there is a bug in ACME.
See the last section in this file.
C-style "==" comparison detected.
To check for equality, use a single '=' character instead.
Converted to integer for binary logic operator.
Applying binary logic to float values does not make much sense,
therefore floats will be converted to integer in this case.
"EOR" is deprecated; use "XOR" instead.
This means the operator, not the mnemonic.
Found old "!for" syntax.
Please update your sources to use the new "!for" syntax. See
AllPOs.txt for details.
You can suppress this warning using the "-Wno-old-for" switch.
Found new "!for" syntax.
When using the "-Wno-old-for" switch to disable the warning about
the older syntax, the new syntax will trigger this warning.
Found SED instruction for CPU with known decimal SBC bug.
This warning is only ever given for CPU types 65ce02 and 4502,
because they are known to be buggy in decimal mode.
Pavel Zima and Eric Smith found an example where $41 minus $08
gave $39 instead of $33.
Label name not in leftmost column.
A label definition has blanks before the label name.
Imagine this source code:
lda #00
imx
rts
Obviously, there's a typo in the middle line (imx instead of inx),
but ACME does not recognize this: It looks just like a label
definition! Therefore releases 0.89 and higher warn you when a
label name does not start in column 1. Releases 0.94 and higher
support a command line option to switch off this warning
("-Wno-label-indent").
Label name starts with a shift-space character.
The name of a global label starts with a shift-space character. It
is highly likely that this is a typing error, therefore this
warning is issued.
Memory already initialised.
The "!initmem" command was given more than once (or in addition to
the "--initmem" command line option). Only use it once.
Output file already chosen.
The "!to" command was given more than once (or in addition to the
"--outfile" command line option). Only use it once.
Segment reached another one, overwriting it.
The program counter has just reached the start of another segment.
Because some people might want to assemble "onto" a binary file
that was loaded before, this warning can be inhibited using
modifier keywords when changing the program counter via "*=".
For extra safety you can also turn this warning into an error
using the "--strict-segments" CLI switch. In future versions of
ACME this might become the default.
Segment starts inside another one, overwriting it.
The given value in a "*=" command is located inside another
segment. Because some people might want to assemble "onto" a
binary file that was loaded before, this warning can be inhibited
using modifier keywords when changing the program counter via
"*=".
For extra safety you can also turn this warning into an error
using the "--strict-segments" CLI switch. In future versions of
ACME this might become the default.
Symbol list file name already chosen.
The "!sl" command was given more than once (or in addition to the
"--symbollist" command line option). Only use it once.
Used "!to" without file format indicator. Defaulting to "cbm".
Now that "!to" can be given a file format keyword (either "plain"
or "cbm"), using "cbm" as default seems inappropriate. It still
works though.
Using oversized addressing mode.
ACME just assembled a command using an addressing mode that was
larger than needed. This only happens if ACME could not work out
the argument's value in the first pass, therefore assuming a 16-
bit addressing mode. If, in a later pass, ACME finds out that the
argument is small enough to fit in 8 bits, then this warning is
shown. If you define all your zeropage symbols *before* they are
first used, this shouldn't happen. If you know that a specific
argument fits in 8 bits, you can force ACME to use 8 bits
addressing by postfixing the command with "+1". Example:
lda+1 label
ACME will then use an 8-bit addressing mode, regardless of whether
the label is known or not. If the label value happens to be too
large to fit in 8 bits, ACME will show an error of course (To
always truncate a value to 8 bits, use the '<' operator).
More about the postfixing method can be found in "AddrModes.txt".
Wrong type - expected address.
Wrong type - expected integer.
Wrong type for loop's END value - must match type of START value.
These warnings are only given when type checking has been enabled
using the "-Wtype-mismatch" switch. Make sure the argument type
matches the instruction's addressing mode.
In "!for" loops, START and END must have the same type, which then
gets used for the loop counter.
Zeropage pointer wraps around from $ff to $00
A zeropage-indirect addressing mode uses $ff as the argument. The
6502 will then fetch the second pointer byte from $00 instead of
$0100, therefore this warning is issued.
...called from here.
If warnings and/or errors are output during a macro call, messages
with this text are added to display the call stack (because you
might need to fix the call instead of the macro itself).
----------------------------------------------------------------------
Section: Errors during assembly
----------------------------------------------------------------------
"ACME" environment variable not found.
This will be shown if the source code references any files from
the library, but the library location variable wasn't set. This
can only be given on systems using the said variable.
"!cbm" is obsolete; use "!ct pet" instead.
This is given when the now obsolete "!cbm" pseudo opcode is
encountered.
"!pseudopc/!realpc" is obsolete; use "!pseudopc {}" instead.
This is given when one of the now obsolete !pseudopc/!realpc
pseudo opcodes is encountered.
"!subzone {}" is obsolete; use "!zone {}" instead.
This is given when the now obsolete "!subzone" pseudo opcode is
encountered.
!error: ...
This is given when the pseudo opcode "!error" is executed. The
actual message varies according to the pseudo opcode's arguments.
Argument out of range.
You called arcsin/arccos with something not in the [-1, 1] range.
Cannot open input file.
ACME had problems opening an input file ("!bin", "!convtab" or
"!src"). Maybe you mistyped its name.
Conversion table incomplete.
The conversion table file is too small. It needs to be exactly 256
bytes in size.
Division by zero.
Guess what - you attempted to divide by zero.
Exponent is negative.
Using negative exponents only give sensible results when using
floating point maths.
Expression did not return a number.
An expression returned a string or a list but a number (integer or
float) was expected.
File name quotes not found ("" or <>).
File names have to be given in quotes. Either "" quoting for files
located in the current directory or <> quoting for library files.
Found '}' instead of end-of-file.
ACME encountered a '}' character when it expected the file to end
instead (because no blocks were open).
Garbage data at end of statement (unexpected 'CHAR').
There are still arguments when there should not be any more. The
given character is the one where end-of-line was expected.
Hex digits are not given in pairs.
The two digits of a hex byte are separated by another character,
or there is an odd number of digits.
Illegal combination of command and addressing mode.
The given command cannot be used with the given addressing mode on
the CPU you have chosen.
Illegal combination of command and postfix.
The given command cannot be used with the addressing mode
indicated by the given postfix.
Illegal postfix.
You used a postfix other than "+1", "+2" or "+3".
Index is undefined.
You attempted an indexing operation with some undefined symbol.
Index out of range.
The value for an indexing operation wasn't in the allowed range.
Macro already defined.
Macros can only be defined once. If you define a macro twice, ACME
will help you find the definitions by giving a warning for the
first definition and a serious error (stopping assembly) for the
second definition.
Macro not defined (or wrong signature).
You tried to call a macro that either wasn't defined yet (always
define macros before using them) or was called with an illegal
argument list. There must be a 1:1 match between the definition's
formal parameters and the call's actual arguments.
Macro parameter twice.
The same symbol name is used two (or more) times in the same macro
parameter list.
Negative size argument.
The size argument of "!bin" or "!skip" must be zero or positive,
but cannot be negative.
Negative value - cannot choose addressing mode.
Because the argument is a negative value, ACME does not know what
addressing mode (8 bits, 16 bits, on a 65816 even 24 bits) to use.
You can overcome this problem using the postfix method. Or correct
your program to use positive addresses instead.
No string given.
ACME expects a string but doesn't find it.
Number out of range.
A value is too high or too low.
This can also mean the desired addressing mode is not available,
as in "ldy $e000, x".
Operation not supported: Cannot apply "OP" to "TYPE".
Operation not supported: Cannot apply "OP" to "TYPE" and "TYPE".
You tried to use an operator on the wrong type(s) of argument(s),
like indexing a float or negating a string.
Program counter is unset.
You didn't set the program counter, so ACME didn't know where to
start.
Quotes still open at end of line.
You forgot the closing quotes.
Source file contains illegal character.
Your source code file contained a null byte.
Symbol already defined.
You defined a symbol that already had a different value. To change
a symbol's value, use the "!set" pseudo opcode.
Syntax error.
Guess what - there's a syntax error.
Target not in bank (0xTARGET).
You tried to branch to an address not in the 0x0000..0xffff range.
Relative addressing (branch commands or PER) cannot leave the
current code bank of 64 KiB.
Target out of range (N; M too far).
Branch commands use relative addressing, which only has a limited
range. You exceeded it. N is the attempted offset, M is the
difference to the limit - so if you succeed in optimizing M bytes
away, the code would assemble.
There's more than one character.
You used a text string in an arithmetic expression, but the string
contained more than a single character.
Too late for postfix.
You can only postfix symbols at the start, before they are used for
the first time.
Too many '('.
A formula ends before all parentheses were closed.
Unknown encoding.
You used the "!convtab" command with a keyword ACME does not know.
Unknown function.
You used a mathematical function ACME does not know.
Unknown operator.
You used an arithmetic/logical operator ACME does not know.
Unknown output format.
You used the "!to" command with a keyword ACME does not know.
Unknown processor.
You used the "!cpu" command with a keyword ACME does not know.
Unknown pseudo opcode.
You have mistyped a "!" command.
Unknown "* =" segment modifier.
You used a modifier keyword ACME does not know.
Unterminated index spec.
An index was started with '[' but did not end with ']'.
Unterminated list.
A list was started with '[' but did not end with ']'.
Value not defined (SYMBOL NAME).
A value could not be worked out. Maybe you mistyped a symbol name.
Whether this is given as a "normal" or as a serious error depends
on the currently parsed pseudo opcode.
----------------------------------------------------------------------
Section: Serious errors (stopping assembly)
----------------------------------------------------------------------
!serious: ...
This is given when the pseudo opcode "!serious" is executed. The
actual message varies according to the pseudo opcode's arguments.
Found end-of-file instead of '}'.
The file ended when ACME expected the block to be closed instead
(because there was at least one block left open).
Loop count is negative.
You used the "!for" command with a negative loop count (getting
this error is only possible when using the now deprecated syntax).
Macro already defined.
Macros can only be defined once. If you define a macro twice, ACME
will help you find both definitions by giving a warning for the
first definition and a serious error (stopping assembly) for the
second definition.
Missing '{'.
ACME didn't find the expected '{' character. Remember that '{'
characters must be given on the same line as the command they
belong to.
Out of memory.
When ACME runs out of memory, it stops assembly, giving this
error. Free some memory and try again. It's highly unlikely anyone
will ever see this error, though. ;)
Produced too much code.
The program counter reached address $10000, leaving the output
buffer. At the moment, ACME can only produce a maximum of 64 KB.
Syntax error.
This is only given as a _serious_ error if it's in a "!do" loop
condition.
Too deeply nested. Recursive macro calls?
The only reason for ACME to have a limit on macro call nesting
at all is to find infinite recursions.
The default limit is 64, this can be changed using the
"--maxdepth" CLI switch.
Too deeply nested. Recursive "!source"?
The only reason for ACME to still have a limit on "!source"
nesting at all is to find infinite recursions.
The default limit is 64, this can be changed using the
"--maxdepth" CLI switch.
Value not yet defined.
A value could not be worked out. Maybe you mistyped a symbol name.
Whether this is given as a "normal" or as a serious error depends
on the currently parsed pseudo opcode.
----------------------------------------------------------------------
Section: Errors on closedown
----------------------------------------------------------------------
Cannot open symbol list file "FILENAME".
Cannot open output file "FILENAME".
Make sure the name doesn't contain wildcard characters and you
have write access to the directory.
No output file specified (use the "-o" option or the "!to" pseudo opcode).
You didn't specify the output file, so ACME did not create one.
----------------------------------------------------------------------
Section: Bugs in ACME
----------------------------------------------------------------------
The warning "Bug in ACME, code follows" is always followed by a
serious error message, stopping assembly. The second message
actually gives a hint about the bug's location in the source code.
If you ever get this combination of warning and serious error,
please send me an e-mail and tell me about it. If possible,
include a piece of source code that triggers it.
Please don't get this wrong - there are no known bugs. I just left
some debugging code in place in case there is a bug I failed to
notice during testing. In practice, this warning is not expected
to be given at all. That's the reason why I want to be notified if
it *does* decide to show up.
The hint messages are of no real interest to the end user, but here
they are for completeness' sake.
ArgStackEmpty
There was no data for a monadic operator to work on.
ArgStackNotEmpty
The expression parser has finished though there are still
arguments left to process.
IllegalGroupIndex
The mnemonic tree contains a group that I didn't add.
IllegalBlockTerminator
A RAM block (macro or loop) was terminated incorrectly.
IllegalOperatorId
IllegalOperatorGroup
The expression parser found an operator that does not exist.
IllegalImmediateMode
The mnemonic tree contains invalid info about the size of immediate
arguments.
IllegalSymbolNameLength
A sanity check on string lengths failed.
NotEnoughArgs
There was not enough data for a dyadic operator to work on.
OperatorIsNotDyadic
OperatorIsNotMonadic
A function was passed the wrong type of operator.
OperatorStackNotEmpty
The expression parser has finished though there are still
operators left to process.
SecondArgIsNotAnInt
A sanity check failed: An argument should have been converted to
integer but wasn't.
StrangeInputMode
The input state machine has reached a state that does not exist.
StrangeIndexBracket
StrangeListBracket
StrangeParenthesis
The expression parser found a non-existing operator.