target in the generated dependency file. The cl65 utility will use this option
to override the depdendency target, if actual object files are to be generated
from C input. So the generated dependency will not have the intermediate .s
file as target, but the final .o file, which allows to use the dependency
files without further processing.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4660 b7a2c559-68d2-44c3-8de9-860c34a00d81
options, --create-dep and --create-full-dep. The latter will include files
that are passed via debug info to the assembler.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4653 b7a2c559-68d2-44c3-8de9-860c34a00d81
* There are now two options, --create-dep and --create-full-dep. One
will add system includes, the other not.
* Both options require a file name. This is an incompatible change(!)
but has the advantage that the user is in control of extension and
path of the generated file.
* Output will always include a phony target for the input files. This
may not work with all make programs.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4652 b7a2c559-68d2-44c3-8de9-860c34a00d81
after changing code. Without this correction, the insn pointer may be wrong
after one of the subfunctions has changed code, which may cause a valid
sequence not to be detected. This change (when proven good) may also go into
2.13.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4649 b7a2c559-68d2-44c3-8de9-860c34a00d81
proper replacements. Some other rearrangements for slightly better code.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4640 b7a2c559-68d2-44c3-8de9-860c34a00d81
cross-bank calls is located at $FECB and the stack grows downwards from
there.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4637 b7a2c559-68d2-44c3-8de9-860c34a00d81
* A newline between the macro name and the argument list of a function like
macro was not accepted.
* An unterminated macro argument list was not always detected.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4621 b7a2c559-68d2-44c3-8de9-860c34a00d81
stack. Useless instructions (duplicate loads or transfers) within the sequence
are left intact and may cause problems because the interfere with the
replacement code. A run of OptRemoveUnusedLoads and friends should fix the
problem, bit this step may be disabled, so the routine has to check for this
condition and avoid it (by not doing the replacement).
git-svn-id: svn://svn.cc65.org/cc65/trunk@4551 b7a2c559-68d2-44c3-8de9-860c34a00d81
named EXEHDR.
* Renamed BASICHDR to EXEHDR for the PET-II machines.
* Moved the call to CHRCH in front of the code that saves the zero page, since
open files are sometimes remembered in the zero page, so we need to close
them before we grab a copy.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4507 b7a2c559-68d2-44c3-8de9-860c34a00d81
data, and at the end of compilation merged if possible. Literals for unused
functions are removed together with the function.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4501 b7a2c559-68d2-44c3-8de9-860c34a00d81
together with a function, if it is not used. Literal storage can now be
controlled by #pragma writable-strings on a per function basis.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4499 b7a2c559-68d2-44c3-8de9-860c34a00d81
I added the "MAKEOPTS=" to the top-level "gcc.mak" that he wants, and I
changed the tool source tree's top-level make-file so that it supports
parallel building. (I added a rule that builds the common library before
building anything else.)
git-svn-id: svn://svn.cc65.org/cc65/trunk@4497 b7a2c559-68d2-44c3-8de9-860c34a00d81
previous opcode. This saves one byte per character with the exception of
characters without a vector representation, where we loose one byte, because
we need to add a dummy vector just to apply our END bit.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4433 b7a2c559-68d2-44c3-8de9-860c34a00d81
collection that may remove something from the collection, we may not use an
upcounting loop.
git-svn-id: svn://svn.cc65.org/cc65/trunk@4409 b7a2c559-68d2-44c3-8de9-860c34a00d81
