acme/docs/Upgrade.txt

201 lines
8.4 KiB
Plaintext
Raw Normal View History

ACME
...the ACME Crossassembler for Multiple Environments
--- upgrading from earlier versions ---
If you haven't used ACME before, you don't need to read this text.
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.97
----------------------------------------------------------------------
a) Single quotes vs. double quotes:
Since "anything in double quotes" is now considered to be a string,
problems can arise when trying to do a calculation with a character
code. Here are some examples:
lda #' ' ; loads 32 like before (ASCII code of space)
lda #' ' + 1 ; loads 33 like before (32 plus one)
lda #" " ; loads 32 like before (ASCII code of space)
lda #" " + 1 ; used to load 33, now fails with error!
The third example still works, because 1-char-strings are treated
just like single characters when returned by the expression parser as
an argument for a mnemonic.
However, the fourth example now fails because the expression parser
tries to add a string to an integer, which is an undefined operation.
Some examples for a related problem:
a = ' ' ; a is 32 (ASCII code of space)
b = ' ' + 1 ; b is 33 (32 plus one)
c = "!" ; c used to be 33, now it's a 1-char string
d = "!" + "!" ; d used to be 66, now it's a 2-char string
If you do not get any errors when compiling your old sources, you do
not need to worry about this problem.
If you _do_ get errors, just use single quotes instead of double
quotes. If you had to use double quotes because the quoted character
itself is a single quote, write '\'' instead (backslash escaping, see
below).
b) Backslash escaping:
Backslashes in single or double quotes are now used as escape
characters. You need to replace any backslash in older sources with a
sequence of two backslashes, so "some\string" becomes "some\\string",
and a single character '\' becomes '\\'.
If you have used backslashes as directory separators in path names (in
Windows/DOS environments), these also need changing - but instead of
using a double backslash, just use a single forward slash ('/')
instead. This has the added benefit of making the sources platform-
independent (*and* it's compatible to older ACME releases as well).
c) Character values are now unsigned:
When parsing a character in single quotes, ACME returns its character
code, according to the chosen encoding (raw/petscii/screencode). If
this resulted in a byte with its most significant bit set, the actual
number was architecture-dependent. Here's an example:
!ct pet ; choose PetSCII encoding
x = 'A' ; PetSCII 'A' is 0xc1, so MSB is set
Now x was either -63 or +193, depending on the host cpu architecture.
Since release 0.97, this example will give 193 on all architectures.
In most cases, this is not a problem, because the actual bit pattern
of the lower eight bits is the same. But if you have written any code
where the numerical value of a PetSCII character is used for
computations _in_the_source_code_, please check those computations.
Use the "--dialect 0.94.12" CLI switch to get the old behavior
concerning a) double quotes and b) backslashes. There is no way to get
the old behavior concerning c) character values, because, as explained
above, the old behavior was architecture-dependent, which is a bad
idea(tm).
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.95.2
----------------------------------------------------------------------
In 6510 mode, ACME now outputs 0x0b instead of 0x2b when assembling
the undocumented ("illegal") ANC #imm8 instruction. Both opcodes do
the same thing, this was only changed because all other mnemonics use
the lowest-numbered possible opcode as well.
Forcing the old behavior via the "--dialect" switch is not supported.
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.94.12
----------------------------------------------------------------------
The pseudo opcode "!for" has a new syntax. The old syntax still works,
but gives a warning.
You can use the "--dialect 0.94.8" CLI switch to get the old behavior.
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.94.8
----------------------------------------------------------------------
The pseudo opcodes "!cbm", "!subzone" and "!realpc" no longer give
warnings, but have now been disabled.
You can use the "--dialect 0.94.6" CLI switch to get the old behavior.
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.94.6
----------------------------------------------------------------------
The "to-the-power-of" operator ('^') is now right-associative, so
b^c^d will now give b^(c^d) instead of (b^c)^d
If you have never used the operator in this way, you don't need to
worry about it.
You can use the "--dialect 0.86" CLI switch to get the old behavior.
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.89
----------------------------------------------------------------------
The "logical shift right" operator has been changed. Note: This is
about ACME's expression parser and has nothing to do with the 6502
mnemonic called "LSR".
Older versions were supposed to work like this:
a = b >> c ; alias "LSR", do a logical shift right
But what they actually did depended on the compiler that was used to
create the ACME binary: many binaries did an "arithmetic shift right"
instead. This has now been fixed and changed to:
a = b >> c ; alias "ASR", do an arithmetic shift right
a = b >>> c ; alias "LSR", do a logical shift right
If you have never applied the old ">>"/"LSR" operator to a negative
value, you do not need to worry about this. If you have, please check
what you expected to happen in those instances (arithmetic or logical
shift) and update your source codes accordingly (use either ">>"/"ASR"
or ">>>"/"LSR").
Forcing the old behavior via the "--dialect" switch is not possible,
because as explained above, the old behavior was compiler-dependent
anyway.
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.07
----------------------------------------------------------------------
Re-defining the program counter via "* = NEW_VALUE" no longer starts
offset assembly. Instead, ACME will change its pointer into the output
buffer to the given value, so you can write your code in distinct
segments. These segments can be given in any order. After assembly,
ACME stores everything from the lowest address used to the highest
address used. Have a look at "AllPOs.txt" for an example on how to use
this facility.
Offset assembly is now done using a new pseudo opcode called
"!pseudopc". Have a look at "AllPOs.txt" for further information on
its syntax and usage.
The pseudo opcode "!end" has been removed. Use "!eof" instead.
The mnemonic BIT can no longer be assembled without any argument. If
you want to insert the opcode only to mask the next instruction, use
!src <6502/std.a>
to get the definitions for these two macros:
+bit8 ; output $24 to mask following 1-byte instruction
+bit16 ; output $2c to mask following 2-byte instruction
When using the 65816 cpu, ACME now uses the correct argument order for
the MVN and MVP mnemonics, which is:
mnemonic source_bank, destination_bank
Using leading zeroes in hexadecimal or binary values makes ACME use
bigger addressing modes than needed. This has now been extended to
symbols as well:
label2 = $00fa
lda label2
will be assembled to:
ad fa 00 lda $00fa
Forcing the old behavior via the "--dialect" switch is not supported.
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.04 beta
----------------------------------------------------------------------
The pseudo opcode "!module" has been removed. Use "!zone" instead.
Forcing the old behavior via the "--dialect" switch is not supported.
----------------------------------------------------------------------
Upgrading from earlier releases to ACME release 0.03 beta
----------------------------------------------------------------------
It is no longer possible to have more than one label in a single line.
Forcing the old behavior via the "--dialect" switch is not supported.