1. Offset/start attributes within a memory area are ignored after an overflow.
2. If a previous segment ends past an offset/start address, then that address is not used.
3. Short map files were generated for memory overflows; now, they are generated for bad offset/start addresses, too.
These pseudo variables will return the size of the accumulator/index
in bits.
For the 65816 instruction set .ASIZE/.ISIZE will return either 8 or 16,
depending on the current size of the operand in immediate addressing
mode.
For all other CPU instruction sets, .ASIZE/.ISIZE will always return 8.
For example:
; Reverse Subtract with Accumulator
; A = memory - A
.macro rsb param
.if .asize = 8
eor #$ff
.else
eor #$ffff
.endif
sec
adc param
.endmacro
Most POSIX function libraries hid that long-time bug by putting zeroes in their dynamic RAM; but, MinGW's library doesn't do it. Therefore, a command like
cl65 foo.c -l
would crash with a "Segmentation fault" -- it should give a nice error message about "-l"; and, quit neatly.
The type needs to change (to array); but, the address shouldn't be changed -- it already points to the first element.
Based on a bug analysis by Daniel Serpell.
Uz classifies his own fix as broken:
http://www.cc65.org/mailarchive/2015-01/11721.html
Although the original issue fixed seemed rather significant looking
back now the regression caused by the (broken) fix seems even more
significant :-(
(reverted from commit 55815ea10c)
======================================================================
Equality problem (Ullrich von Bassewitz)
Neil Stockbridge reported a problem with equality comparisons on
cc65.org's mailing list:
http://www.cc65.org/mailarchive/2014-10/11680.html
Uz provided a fix for it:
http://www.cc65.org/mailarchive/2014-10/11683.html
This pull request ask to add the fix to cc65 on github.
Before this commit, we could define segment ranges; but, the disassembler wouldn't do anything with those definitions. Now, da65 will put ".segment" directives into its output.
Fixed da65's document.
For example:
int f(int i, int j) { return (char) (i + 1) == j; }
f(0x1234, 0x35) returned 0.
This bug caused zlib/uncompress return Z_DATA_ERROR on correct input.
- Force usage of shell wrapper in order to allow to suppress potential message about git not found (thanks to Greg King).
- Do $(info GIT_SHA ...) only if there's something special - as done with the other $(info ...).
After all it seems much cleaner to explicitly control the behaviour from the .travis.yml file than to behave differently "under the hood" when detecting a Travis CI build.
I must admit that I don't understand why but obviously it is necessary to place .PHONY below the definition of variables it refers to - although those variables are recursively expanded ones! Not doing so made libsrc/Makefile build only three target libraries.
It seems more appropriate to trigger the inclusion of Makefile.inc via $TARGET (instead of $SRCDIR). This is btw. more consistent with extzp.s inclusion which is triggered via $TARGET too.
- Code specific to Windows was #ifdef'ed with _MSC_VER so it wasn't included with MinGW. So _MSC_VER is replaced with _WIN32.
- MinGW doesn't support _get_pgmptr() so it is necessary to directly call the Win32 function GetModuleFileName(). This implies including windows.h which in turn causes a name clash with the Win32 function SearchPath(). So the cc65 type SearchPath is renamed to SearchPaths.
added a missing addressing mode for jmp/jml : Absolute Indexed Long
that opcode can be written like jmp[$1234] or jml[$1234]
removed Absolute Inderect addressing mode for jml since it's not a long adressing mode
Actually there's no need to fiddle with any automatic temp file name generation as we can just modify the library path name to generate a perfect temporary library path name.
There's no need to intentionally break things for people trying to build cc65 with some obscure compiler. It's enough to make sure that no new warnings sneak into the code base by having Travis CI builds fail.
- tmpfile() tries to create files in the C: root dir on Windows where users usually don't have access.
- tmpnam() uses the curent dir which doesn't seem to be a good idea either.
- tempnam(NULL,NULL) is supposed to be cleverer.
The 'install' target primarily aims to support pacaking tools. Therefore...
- It just presumes a "capable" install program to be present.
- There's intentionally no 'uninstall' target.
Before the fix, that feature couldn't recognize a standard op-code mnemonic, that wasn't replaced by a macro, if it was on a line without a label.
This patch was written by Jeremy Turner.
The code
void foo(void)
{
int i;
long l = 1L * i;
}
triggered an
Internal compiler error:
Code generation messed up: StackPtr is -4, should be -2
Greg King: "We are lucky that the bug is simple -- a missing "else". The result is that the compiler thinks that it does the opposite of what it actually does: It thinks that it pushes the non-constant expression onto the stack. It doesn't; so, cc65's stack pointer is wrong."
Before this fix, BSS-type and ZP-type segments never were counted. Now, they are counted if their memory areas are filled. (It must be done because their places in the output file are filled.)
The fix allows us to build programs for the CBM510 and CBM610 platforms. We won't see an "Internal error" diagnostic message about a bad file-offset.
If the variable 'prefix' is defined then the builtin search paths are set to
$(prefix)/lib/cc65/... allowing to build binaries intended for installation.
Note that the library build still works with these binaries as it generally
overrides the builtin search paths by setting the CC65_HOME env var.
The target 'atarixl' is to be used for Atari XL (and better) machines.
It will disable the OS ROM and enable the Shadow RAM available on
those machine.
Note: This commit is only the inital step towards for this goal that just
replicates the target 'atari' as a starting point!
The targets allow to run cc65 programs in the sim65 exection
einvironment. As there are no "real" i/o facilities there's no
need for header files. Paravirtualized entry points are mapped
to $FFF0 ff. There's a large cc65 progam area from $0200-$FFEF.
The binary format includes a one-byte header indicating the required
execution environment: The value 0 means 6502 and the value 1
means 65C02. The load adress for the binary is fixed to $0200.
Note: Running sim65C02 programs currently doesn't work bcause
sim65 doesn't actually implement 65C02 opcodes.
The sim65 source code has been a construction site for over a decade.
I was looking for a simple cc65 program execution environment for
regression tests. So I decided to re-purpose sim65 for that task by
removing about everything but the 6502 emulation.
There's no memory mapped i/o emulation whatsoever. Rather exit(),
open(), close(), read() and write() calls are supported by mapping
them through a thin paravirtualization layer to the corresponding
host os functions.
Note: The sim65 6502 emulation provides means to switch between
6502 and 65C02 emulation but currently there are no actual 65C02
opcodes implemented.
There are two reasons for removing this dependency:
- If someone does 'make avail' on the top level Makefile he ends up with
binaries but without libraries - not nice. Better do just "nothing" and
have hin understand that he needs to do 'make [all]' on the top level
Makfile first.
- If 'make avail' is done via 'sudo' it isn't desirable to do a large amount
of work as root.
BTW: I wasn't sure if this dependency is a good thing in the first place
but I saw it in many examples ('install' depending on 'all') so I did it too.
The [un]install make goals have a rather fixed meaning for *ix users. The simple
symlinks provided here don't match the expectations users have from [un]install.
Therefore it is appropriate to rename them to "something" not tied to specific
expectations.
While ca65, cc65 and ld65 have built-in paths cl65 doesn't.
That means that up to now cl65 depended on the binaries
being found in the path env var. However it makes sense
to presume that the binaries are located in the very same
directory cl65 is located in. So whatever pathname was
suitable to run cl65 should be suitable to run the other
binaries too. But if for some reason there's no valid
argv[0] or if it doesn't contain at least one directory
delimiter ('/' or '\') then fall back to relying on the
path env var.
In contrast to *IX it doesn't make much sense to add compile time defined
search paths to Windows binaries: There's no standard path like /usr/local/bin
(and there are no symbolic links to link from there to another location).
On the other hand it's (again in contrast to *IX) easy for Windows binaries
to determine their own paths. Therefore it's appropriate to make use of that
to add run time defined default search paths.