gbdk/sdcc/doc/SDCCBugsFixed.html
2015-01-10 16:25:06 +01:00

1555 lines
53 KiB
HTML

<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<meta name="GENERATOR" content="Mozilla/4.7 [en] (X11; I; Linux 2.2.12-20 i686) [Netscape]">
</head>
<body>
<b><font size=+2>Enhancements and Bug fixes.</font></b>
<br>&nbsp;
<blockquote>
<hr WIDTH="100%">
<br><b>Release 2.1.9G</b>
<ul>
<li>
<font color="#CC0000">Applied several patches provided by&nbsp; "Anton
Voloshin" &lt;vav@isv.ru> mainly to SDCCglue.c, SDCCmain.c &amp; sdcdb/cmd.c</font></li>
<li>
<font color="#CC0000">First parameter of struct member function pointer
not passed in register.</font></li>
<li>
<font color="#CC0000">Parameter passing problem if first parameter of a
routine passed as second parm of another routine, vice versa. This bug
would caused some of the library routines to fail.</font></li>
<li>
<font color="#CC0000">Address of structure member passed as parameter caused
problems.</font></li>
<li>
<font color="#CC0000">#pragma callee-saves did not always work.</font></li>
<li>
<font color="#CC0000">Furthur bug fix for live range computation within
loops.</font></li>
<li>
<font color="#CC0000">Register 'b' could be clobbered when there was '*',
'/' or '%' operation for generic pointers.</font></li>
<li>
<font color="#CC0000">initialisation for pointers to structure elements
in code space generated errors.</font></li>
<li>
<font color="#CC0000">Debugger info lost when multiple files were linked
.</font></li>
<li>
<font color="#CC0000">Fixed buf in printf_large.c(vsprintf called with wrong parms)
.</font></li>
<li>
<font color="#CC0000">"extern" variables when spilt did not allocate spill locations
. i.e. "sloc" undefined error.
.</font></li>
<li>
<font color="#CC0000">Added optimization for PUSH. "xdata" variables do not need to be
assigned to temps to be pushed.
.</font></li>
<li>
<font color="#CC0000">Fixed aliasing problem where some pointer sub expressions
were carried across basic blocks .</font></li>
<li>
<font color="#CC0000">Upgraded to version 0.2.36 of ucsim .</font></li>
<li>
<font color="#CC0000">Added new optimization within basic blocks pointer
set values will replace get pointer values within a basic block. (across
basic blocks is a little bit complex will have to think about it some more).</font></li>
<br><font color="#CC0000">which in plain english means.</font>
<br><font color="#CC0000">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; a->b
= 10;</font>
<br><font color="#CC0000">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; putchar(a->b);</font>
<br><font color="#CC0000">will come..</font>
<br><font color="#CC0000">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; a->b
= 10;</font>
<br><font color="#CC0000">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
putchar(10);</font></ul>
</blockquote>
<blockquote>
<hr WIDTH="100%">
<br><b>Release 2.1.9F</b>
<ul>
<li>
<font color="#CC0000">pointers passed to functions as parameters are removed
from cse list.</font></li>
<li>
<font color="#CC0000">live range computation for variables inside loops
improved.</font></li>
<li>
<font color="#CC0000">upgraded ucsim to latest release 0.2.31</font></li>
<li>
<font color="#CC0000">loop invariant bug fix for (global variables assigned
constant values inside a loop)</font></li>
<li>
<font color="#CC0000">Applied patch provided by&nbsp; "Anton Voloshin"
&lt;vav@isv.ru> rel files put into separate lines in the .lnk file.</font></li>
<li>
<font color="#CC0000">union of structures caused compiler to generate incorrect
code for member access.</font></li>
<li>
<font color="#CC0000">added patch to vprintf.c to null terminate output
string. BESSIERE_Jerome@stna.dgac.fr (BESSIERE Jerome)</font></li>
<li>
<font color="#CC0000">Debugger had problem with structures/unions on stack</font></li>
<li>
<font color="#CC0000">Debugger will now stop when simulator stops even
when break point not hit.</font></li>
<li>
<font color="#CC0000">Debugger will display simulators "startup message",
this makes it a clearer to the user that the debugger is actually piggy
backing on top of the simulator.</font></li>
<li>
<font color="#CC0000">Applied patch to assembler/linker some names were
getting truncated @ 8 characters . patch by&nbsp; "Anton Voloshin" &lt;vav@isv.ru></font></li>
<li>
<font color="#CC0000">_gptrput.c was leaving a byte on stack if unknown
storage class.</font></li>
<li>
<font color="#CC0000">parameter passing was incorrect for reentrant functions
when prototype declared without parameter name.</font></li>
</ul>
</blockquote>
<blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.9E</font></b>
<ul>
<li>
<font color="#CC0000">"remat " was not being cleared for global variables.</font></li>
<li>
<font color="#CC0000">stack offset of a variable computed incorrectly in
some situations.</font></li>
<li>
<font color="#CC0000">Applied patch for floating point routine provided
by<font size=+1> </font>Wojciech Zabolotny &lt;wzab@ise.pw.edu.pl></font></li>
<li>
<font color="#CC0000">Force allocation of local variable if address taken
(even if not initialised)</font></li>
<li>
<font color="#CC0000">Fixed loop reversal bug .</font></li>
<li>
<font color="#CC0000">Fixed another bug in register allocator , long variables
can be printfed now.</font></li>
<li>
<font color="#CC0000">DEBUGGER can now print values on stack and aggregates.</font></li>
<li>
<font color="#CC0000">Upgraded to 0.2.26 of ucsim.</font></li>
</ul>
</blockquote>
<blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.9D</font></b>
<ul>
<li>
<font color="#CC0000">Fixed loop optimization bug was treating variables
defined twice as loop induction.</font></li>
<li>
<font color="#CC0000">Fixed loop optimization constrained induction variable
detection.</font></li>
<li>
<font color="#CC0000">Fixed bug in some floating point routines (fix provided
by Wojciech Zabolotny &lt;wzab@ise.pw.edu.pl>)</font></li>
<li>
<font color="#CC0000">Fixed code generation bug for return literal floating
point.</font></li>
<li>
<font color="#CC0000">Fixed register packing was not moving return value
under certain conditions.</font></li>
<li>
<font color="#CC0000">Improved error recovery for undefined symbols.</font></li>
</ul>
</blockquote>
<blockquote>
<hr WIDTH="100%">
<br><b><font size=+2>Release 2.1.9C</font></b>
<ul>
<li>
<font color="#CC0000">Fixed bug with register allocation, effected mostly
large routines compiled using the --model-large option.</font></li>
<li>
<font color="#CC0000">Fixed bug in array of structures dereferncing.</font></li>
<li>
<font color="#CC0000">Simplified the intermediate code generattion , removed
some tree nodes. The parser will not generate nodes with structure derefs,
and the operators +=,-= ..etc will be changed at parse time, this creates
a simplififed parse tree.</font></li>
<li>
<font color="#CC0000">Upgraded to latest version of "gc", compile problems
on Redhat 6.0 should be fixed.</font></li>
<li>
<font color="#CC0000">Fixed passing simulator parameters from the debugger
command line.</font></li>
<li>
<font color="#CC0000">Fixed problem with the floating point routines.</font></li>
</ul>
</blockquote>
<blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.9B</font></b>
<ul>
<li>
<font color="#CC0000"><font size=+1>Fixed pointer access and array reference
bug introduced in the previous version.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>integrated ucsim.0.2.25</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug for code not generated for
global pointers assigned to address of array element.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed one more bug in live range computation.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in register allocator , it
was overwriting computed liverange values.</font></font></li>
</ul>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.9A</font></b>
<ul>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in common subexpression elimination,
problem was found when a local variable was assigned values more than once
in a basic block.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added warning when a "void" object
is cast to a non-void object.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added warning when a non-pointer type
is cast to a _generic pointer.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added a&nbsp; peephole optimization
rule to remove some moves.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem in register packing ,
registers assigned to parameters were not being saved on the stack.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Bug in loop inductions, the induction
variable detection was not correctly replacing all references.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Merged bug fixes provided by <i>Wojciech
Stryjewski &lt;wstryj1@tiger.lsu.edu to></i> SDCCgen51.c</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Local variable treated as invariant
even when address is taken.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed Parameter problem for functions
with empty bodies.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem with array dereferncing
when array of pointers.</font></font></li>
<br>
<hr WIDTH="100%"></ul>
</blockquote>
<blockquote><b><font size=+1>Release 2.1.9</font></b>
<ul>
<li>
<font color="#CC0000"><font size=+1>Integrated new simulator from Daniel
"ucsim-0.20.xx"</font></font></li>
<li>
<font color="#CC0000"><font size=+1>.cdb file will be created always, and
will contain global variables &amp; function entry/extis information always.
This will facilitate debuggers especially NoICE.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug when using typedefed types
in a structure.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed aliasing bug in the loop optimizer,
when a global array is used in a loop it should not be treated as an invariant
if there are function calls in the loop.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem with register allocation,
this was causing some of the "long" routines to fail.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem of clobbering value of
_bp.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem of clobbering acc when
a long variable is the first argument to a reentrant function.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed a peephole optimizer problem,
when it was changing the ljmp to a sjmp in a jump-table.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Enhanced the initialization routines
to support initialization with '&amp;' operations.</font></font></li>
</ul>
</blockquote>
<blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.8</font></b>
<ul>
<li>
<font color="#CC0000"><font size=+1>Void function returning value check
was not working properly.(fixed)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Assigning 'void' values to variables
will generate warning.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Improved register allocation to use
"acc" , this will improve code generation , but for this register color
is severely restricted, will gradually work on improving this. This reduces
the redundant register moved drastically (once totally implemented)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added some peephole rules eliminate
jump to jumps.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Bug in SDCCgen51.c when using a bit
variable (locally defined) in an if statement.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Literal values being cast to pointers
was not working.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>'volatile' arrtibute for structure/union
elements were bieng ignored. Now we can selectively declare structure elements
to be volatile.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Did some profiling to improve performance,
changed peephole optimizer a little with profiler feedback.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Was generating incorrect "cjne" code
under some situations.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Improved data flow computation.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in geniCodeCast for floating
point conversions.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in modulong routine found
by B&eacute;la T&ouml;r&ouml;k &lt;aseht@ms1.kssg.ch></font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug for long shift left found
by B&eacute;la T&ouml;r&ouml;k &lt;aseht@ms1.kssg.ch></font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug with register allocator ,
was marking variables as rematerializable too aggresively.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added --version option.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem of compiling source files
not in current directory.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed a bug in a bitvector routine
(bitVectResize)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added error message when "continue/break"
are out of context.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in error checking for uninitialised
local variables.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in code generation for not
(!) operation.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in live range computation
when intervenening unreachable block is present.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed register saving scheme, bank
will be saved only when called from a interrupt service routine.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug with >> and &lt;&lt; with
unknown shift count.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>The fix to the bit vector exposed some
problem with the register allocation, fixed now.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added an extra heuristic for register
spilling, now the spilling will a lot lower than before.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added some more optimizations for register
packing.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added a bunch of preprocessor options
see command line options in the documentation for details.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Some cosmetics like timestamp in .asm
in addition to Version Number.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>fixed another bug with live range computation.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Automatic static variables will not
be assigned to registers.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>The parameter name mangling a little
so that it gives the warnings/errors with the correct name, also required
for debugger.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Jean Louis fixed some bugs with shifting
"long" variables.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug of access of local bit variables
when compiling with --model-large</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed two bugs in Register allocator.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added fix to peephole optimizer rule
matching , fixed provided by Frieder.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>"Mark O'Farrell" &lt;ofarrell@haq.com.au>
added segment type to the segment definition.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>"Mark O'Farrell" &lt;ofarrell@haq.com.au>
added option to start stack after the data segment --stack-after-data (see
documentation)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed the computation for default
stack location, now if all four banks are used the stack will be placed
after the data segment (--stack-after-data), so as not to clobber the bit
segment (this should make Daniel Drotos happy:).</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Compiler will no longer crash when
an uninitialised local variable is detected.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug for bad code generation for
anl,xor &amp; or when operand is on stack .</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug for side effecting parameters
when calling a reentrant function.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Improved optimization for ++ &amp;
-- operation when operand is on stack.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Was STILL saving entire register bank
when calling function with different register bank from non ISR function.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed register saveing for ISR functions
not using register bank (see documentation [Interrupt Service Routines])</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Josef Wolf &lt;jw@raven.inka.de> provided
a set of serial routines which do not use the external ram (ser_ir.c),
included these as libraries.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>PSW was not being saved for ISRs with
no register bank specification.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Integrated Daniel's simulator with
the compiler source distribution.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added reduced "printf" routine it works
but is very resource intensive.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Certain complex pointer declarations
were not being handled correctly.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>switch statements with empty bodies
were causing the compiler to crash.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>sizeof union object computed incorrectly.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>union element when passed as parameter
to a function generated incorrect code.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Martijn van Balen &lt;balen@natlab.research.philips.com>
provided a complete set of printf routines including sprintf.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug when union member is a pointer
and is used on the right side of an assignment.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Improved register allocation&nbsp;
for arrays, structure &amp; union members in data space will no longer
be spilt.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Some optimizations for pointer access
had got clobbered put them back.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug for right shifting signed
quantity by zero.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Constant propagation had problem when
long variables were used in division.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in register allocator on
rare situations it saved wrong live ranges.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed Bug with Backend (gen51.c) incorrect
code was generated when far space variable assigned o variable in far space.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem with loop invariant processing,
was not taking into consideration when address or pointer was passed as
parameter to a function.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added --help &amp; -h option.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added the peephole optimization rules
provided by Frieder&nbsp;&nbsp; &lt;fe@lake.iup.uni-heidelberg.de></font></font></li>
<li>
<font color="#CC0000"><font size=+1>Function prototypes without parameter
names were occasionally causing problems.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem with _spx inc/dec when
using xternal stack.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Register allocation bug</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Addition of constant to generic pointer
was causing problems sometimes.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Improved register allocation some more.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in live range analysis (for
some complex contro flow situations.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug of 'b' getting clobbered
in some left &amp; right shift operations.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added --callee-saves (documentation
updated)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in hTabDeleteItem() in SDCChasht.c
(exposed during debugger development)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in grammar .. to allow absolute
address for local "bit" variables.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added #pragma CALLEE-SAVES (same effect
as --calle-saves)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added #pragma EXCLUDE (see docs)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem xrl generating incorrect
code.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Yet another problem in register allocator
fixed.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed a bug with dataflow computation,
was computing pointer kills incorrectly.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added loop reversal.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in structure declaration</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed floating point routines</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed parsing problem for LONG hex
constants</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug to allow local variables
to be declared in 'code' space.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in code generation for floating
point .</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Parameters access was incorrect when
callee-saves &amp;&amp; reentrant.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in hashTable routine.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug when string initialisers
are used for char arrays with bounds.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Was generating incorrect code when
options --model-large &amp;&amp; --stack-auto were used together</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Replaced 32bit multiplication routine
with JeanLouis VERN's much better routine.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem with loop-reversing</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Introduced parameter passing using
registers.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Applied improvements/cleanups supplied
by Larry Dolittle.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Moved the variable substitution by
register equivalent to earlier phase , this generates much more optimized
code.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Compiler crashed when first parameter
was declared "const"</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Assembler generated bad opcode for
"djnz" when the "." operator was used.</font></font></li>
</ul>
</blockquote>
<blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.7</font></b>
<ul>
<li>
<font color="#CC0000"><font size=+1>Fixed a bug in the DEF-USE chaining
algorithm , was keeping some dead DEFs alive in the DEF bit vector of a
symbol.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug with "(null)" operand generated
in rare situations.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>&nbsp;Trefor &lt;Trefor@magera.freeserve.co.uk>
did some more clean up.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug that was introduced with
'&amp;' with a literal.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added alternate serial package provided
by "Wolfgang Esslinger "&lt;wolfgang@WiredMinds.com> in ./sdcc51lib/_ser.c
(documentation updated)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added Header file&nbsp; sab80515.h
provided by B&eacute;la T&ouml;r&ouml;k &lt;aseht@ms1.kssg.ch></font></font></li>
<li>
<font color="#CC0000"><font size=+1>Had made flow computation for global
variables too restrictive, now constant propagation will be allowed with
global variables.</font></font></li>
</ul>
</blockquote>
<blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.6</font></b>
<br>&nbsp;
<blockquote>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in the routines <b>_*2fs.c
</b>the
variable 'sign' should be declared as 'long'.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Included _char2fs in libfloat.lib</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed register allocator&nbsp; bug
for allocating pointers (found by "Dr. Ed Croson" &lt;soarebc@internetcds.com>).</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Furthur loop induction variable detection
improvement.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed loop induction detection so
that it does not treat global variables as induction any more, this saves
some code, since assignments to global variables cannot be easily eliminated.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed problem of __sdcc_program_startup
being out of range when large amount of initialization code is present.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug with function returning pointers.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed register allocation to aggressively
allocate local variables to registers, users will see a vast improvement
in the generated code, not only in terms&nbsp; of speed but also in terms
of data space utilization. Previously if there were multiple definitions
for a local variable flowing into a basic block the compiler would allocate
space for the variable, now it will allocate space only if it runs out
of registers.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Jean-Louis VERN (jlvern@&lt;writeme.com>)
has done some more enhancements to the code generator ..... his comments.</font></font></li>
<blockquote>
<li>
<font color="#CC0000"><font size=+1>I did some light improvement s:</font></font></li>
<br><font color="#CC0000"><font size=+1>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
- increment 16&amp;32bits</font></font>
<br><font color="#CC0000"><font size=+1>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
- decrement 16&amp;32bits</font></font>
<br><font color="#CC0000"><font size=+1>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
- assignment with literal</font></font>
<br><font color="#CC0000"><font size=+1>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
- return with literal</font></font>
<br><font color="#CC0000"><font size=+1>and some cosmetics before optimizing
mult and div...*</font></font></blockquote>
<li>
<font color="#CC0000"><font size=+1>&nbsp;"BESSIERE Jerome" &lt;BESSIERE_Jerome@stna.dgac.fr>
made a lot of improvements to the support routines, these include reducing
some memory leaks.. here are his comments.</font></font></li>
<blockquote>
<li>
<font color="#CC0000"><font size=+1>//* Incorrect call of werror(E_OUT_OF_MEM,...);</font></font></li>
<br><font color="#CC0000"><font size=+1>//* 1-a In file SDCCBBlock.c /
function iCodeBreakDown (...)</font></font>
<br><font color="#CC0000"><font size=+1>//* 1-b In file SDCChasht.c / function
newHashTable(...)</font></font>
<br><font color="#CC0000"><font size=+1>//* 1-c In file SDCCmain.c / function
parseCmdLine(...)</font></font>
<li>
<font color="#CC0000"><font size=+1>//* Sometime the compiler crash due
to compilation error</font></font></li>
<br><font color="#CC0000"><font size=+1>//* In file SDCCast.c / function
createFunction(...)</font></font>
<li>
<font color="#CC0000"><font size=+1>//* Sometime SDCC access not allocated
memory</font></font></li>
<br><font color="#CC0000"><font size=+1>//* In file SDCChasht.c / function
hTabSearch(...)</font></font>
<li>
<font color="#CC0000"><font size=+1>/* Many and many memory leak when not
using CYGWIN32</font></font></li>
<br><font color="#CC0000"><font size=+1>//* and&nbsp; garblage collector
(I'm using MS VC++ 6.0 and WIN32)</font></font>
<br><font color="#CC0000"><font size=+1>//* specially in peepHole function
consuming Mo and Mo</font></font>
<br><font color="#CC0000"><font size=+1>//* even on compiling small function</font></font>
<br><font color="#CC0000"><font size=+1>//* I created a new function in
file SDCChasht.c for deleting an hTab</font></font>
<br><font color="#CC0000"><font size=+1>//* It Reduces memory leaks</font></font></blockquote>
<li>
<font color="#CC0000"><font size=+1>"Salvador Eduardo Tropea \(SET\)" &lt;salvador@inti.gov.ar>
provided patches required to compile the system using DJGPP, the executables
produced can run on plain DOS. He has also created a <b>readme.dj</b> file
for those who want to compile using DJGPP.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added code to issue warnings if something
is compared to / assigned a literal value which is greater than what the
variable can hold.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed the data/space extended message
to <b>info </b>, since this not really a warning message.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added routine auto baud, for automatic
baud rate detection , adapted from Paul Stoffregen's&nbsp; &lt;paul@ece.orst.edu>
autobaud.asm</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Was not doing check for constants in
case for ++/--, i.e. a variable declared as const could be incremented
or decremented</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug when setting &amp; fetching
data from far space .</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added bunch of peephole optimization
rules.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Trefor@magera.freeserve.co.uk. Updated
the build script to give a choice to compile the libraries as large model
or small model.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Trefor also got rid of those annoying
warnings when compiling with egcs ( braces about ambiguous 'else')</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added some heuristics in constant folding
routine , to prevent pointer live ranges from becoming too long, this was
causing some avoidable spill code to be generated.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added support for ANSI routines <b>setjmp</b>
and <b>longjmp. </b>Documentation updated.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Was generating code for unreachable
basic blocks ... now will not generate code for unreachable code... will
issue a warning.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Was generating incorrect code when
adding values to generic pointers (some times).</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Improved bitwise '&amp;' operation
a little bit to improve code generation.</font></font></li>
</blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.5</font></b>
<p><font size=+1>Mostly a bug fix release</font>
<br>&nbsp;
<blockquote>
<li>
<font color="#CC0000"><font size=+1>Fixed post &amp; pre increment / Decrement
problem and inefficiency</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Structures/Unions. Three things were
done this structure declarations.</font></font></li>
<blockquote>
<li>
<font color="#CC0000"><font size=+1>The long standing union problem has
been fixed, a very rudimentary solution right now , will some up with a
better solution shortly.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Structures declared inside function
scopes were not being properly cleaned up, came across this problem while
generating debug information fixed now.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>pointers declared inside structures
were broken after the pointer fix in Version 2.1.3, now fixed.</font></font></li>
</blockquote>
<li>
<font color="#CC0000"><font size=+1>Fixed some library routines , for --model-large
.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added whole lot of stuff for debug
information generation (DO NOT use --debug option might cause unpredictable
results).</font></font></li>
</blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.4</font></b>
<br>&nbsp;
<ul>
<li>
<font color="#CC0000"><font size=+1>Problem with parameter overlay : when
two functions with overlaid parameters were nested, the parameters of the
outer function was clobbered. This has been fixed, now the inner function
call will be lifted out and evaluated before parameters for the outer call
are assigned values.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>New optimization added (suggested by
Jean Louis VERN &lt;jlvern@writeme.com> to prevent global variables in
near space from being assigned to registers.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Improved loop optimization, now will
keep induction variables alive after loops even if loops have more that
one exit. This makes me happy, it takes care of almost all the common loop
constructs. So unless the user is doing something really weird the induction
variables&nbsp; will be carried outside the loop. This also lays down the
ground work for value based optimizations, i.e. when the loop count is
known, then value of the induction variables will also be known at certain
exit points.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in genPlus, genAnd,genXor,genOr
routines , was generating incorrect code for Far Space variable with Literal.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bugs in SDCCglue.c, was generating
-ve size for variables of size 32768 or greater.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed register allocator to try &amp;
keep all local variables in 'registers' regardless of they are declared
with the register storage class or not, this will reduce the amount of
data space required by the compiler generated code.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in SDCCmain.c , was unlinking
file preOutName before reading it (okay with POSIX , not okay with DOS).</font></font></li>
<li>
<font color="#CC0000"><font size=+1>David Bonnel &lt;dbonnell@ozemail.com.au>.
fixed the following bugs</font></font></li>
<ul>
<li>
<font color="#CC0000"><font size=+1>In SDCCicode.c at line# 1314 should
be calling valMod not valDiv.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>In line# 2728 missing a break statement
after the division case .</font></font></li>
</ul>
<li>
<font color="#CC0000"><font size=+1>David Bonnell &lt;dbonnell@ozemail.com.au>.
also added a new option <b>--iram-size &lt;value> .&nbsp;&nbsp;&nbsp;</b>
His description of it .....&nbsp; (Dave speaking)</font></font></li>
<ul>
<li>
<font color="#CC0000"><font size=+1>I added the command '-a [iram-size]'
to the linker. iram-size defaults to&nbsp;&nbsp; 128 (0x80) if it is omitted
(which is the amount of internal RAM I have on&nbsp; my DS5000T chip :-)</font></font></li>
</ul>
<font color="#CC0000"><font size=+1>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
Anyway, if this command is given, the (new) function iramcheck() is called
after linking to check that the OSEG, ISEG, and DSEG's do not extend&nbsp;
past the end of internal RAM.&nbsp; (Internal RAM is from&nbsp; 0..iram_size-1).&nbsp;
It spits out a warning such as:</font></font>
<p><font color="#CC0000"><font size=+1>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
WARNING! Segment DSEG extends past the end of internal RAM.&nbsp; Check
map file.</font></font>
<p><font color="#CC0000"><font size=+1>&nbsp;&nbsp; when such a problem
is detected.</font></font>
<p><font color="#CC0000"><font size=+1>&nbsp;&nbsp; Then I added a new
command-line option to SDCC, '--iram-size iram-size'. When this command
is given the linker command shown above is added to the linker command
file so that it will do the checking for internal RAM overflow.</font></font>
<br>&nbsp;
<li>
<font color="#CC0000"><font size=+1>&nbsp;BESSIERE Jerome &lt;BESSIERE_Jerome@stna.dgac.fr>
fixed the following bugs.</font></font></li>
<ul>
<li>
<font color="#CC0000"><font size=+1>File sdccicod.c / function operandOperation(...)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>File sdccval.c / function valDiv(...)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>File sdccbitv.c / function bitVectBitValue(...)</font></font></li>
<li>
<font color="#CC0000"><font size=+1>File sdccbitv.c / function bitVectUnSetBit(...)</font></font></li>
</ul>
</ul>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.3</font></b>
<br>&nbsp;
<ul>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in assembler&nbsp; <b>i51mch.c</b>&nbsp;
; <b>dot</b> operator when used in branch instructions was generating incorrect
opcode.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Enhanced Dead Code elimination, (it
was missing some ).</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added new rules to SDCCpeeph.def&nbsp;
provided by Frieder &lt;fe@lake.iup.uni-heidelberg.de></font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed pointer return bug found by David
Bonnel &lt;dbonnell@ozemail.com.au>.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed the POINTER fiasco, I had misinterpreted
the grammar.</font></font></li>
<ul>
<li>
<font color="#CC0000"><font size=+1>data char *p;&nbsp; was interpreted
as [generic pointer to char, with p in data space]. now interpreted as
pointer to character in data space with the pointer residing in unknown
space.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Updated the documentation for this.</font></font></li>
</ul>
<li>
<font color="#CC0000"><font size=+1>Updated documentation to include comments
about Parameter Overlaying.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Improved type checking to generate
error message when a void type is casted to some other type</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Jean-Louis VERN (instead of sleeping
at night) has made some GREAT!! improvements to the back end for frequently
used expressions like</font></font></li>
<ul>
<li>
<font color="#CC0000"><font size=+1>if ( x &amp; literal) ...</font></font></li>
<li>
<font color="#CC0000"><font size=+1>if ( x ^ literal) ...</font></font></li>
<li>
<font color="#CC0000"><font size=+1>if ( x |&nbsp; literal) ...</font></font></li>
<li>
<font color="#CC0000"><font size=+1>if ( bit &amp; x) ...</font></font></li>
<li>
<font color="#CC0000"><font size=+1>if ( bit | x ) ...</font></font></li>
<li>
<font color="#CC0000"><font size=+1>if (bit ^ x) ....</font></font></li>
</ul>
<li>
<font color="#CC0000"><font size=+1>Jean-Louis VERN rewrote the _decdptr.c
routine to improve it.</font></font></li>
</ul>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.2</font></b>
<br>&nbsp;
<ul>
<li>
<font color="#CC0000"><font size=+1>Bug in genAnd (SDCCgen51.c) caused
the compiler to issue error message for some bit wise and operations.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>File _spx.c missing from source bundle.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Invalid code generated when interrupt
service routine used with external stack option.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added --no-peep option to disable peep
hole optimization.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added --peep-asm option to optionally
pass inline assembler code through the peephole optimizer.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Fixed bug in _strcpy.c David Bonnell
&lt;dbonnell@ozemail.com.au>.</font></font></li>
</ul>
<ul>
<li>
</li>
</ul>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.1.1</font></b>
<br>&nbsp;
<ul>
<li>
<font color="#CC0000"><font size=+1><b>bit </b>variables could not be declared
<b>extern</b>
, bug was introduces while improving type checking for bit fields in structures.</font></font></li>
</ul>
<hr WIDTH="100%">
<p><b><font size=+1>Release 2.1.0</font></b>
<p><font size=+1>A full source code release. Lots of stuff done in this
release.</font>
<br>&nbsp;
<ul>
<li>
<font color="#CC0000"><font size=+1>Floating point support completed (although
cannot be used in small model).</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Data flow computation enhanced to stop
flow computation of global variables when there is a call to another function.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>JeanLouis VERN added a whole lot of
optimizations to the back end.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Function return value assigned to a
bit variable was crashing the compiler.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed the way peep hole optimizer
rules were being included. Now it is much easier to add default rules.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added two new pointer types "* pdata"
and "* idata" these will facilitate the following</font></font></li>
<ul>
<li>
<font color="#CC0000"><font size=+1>Optimized access to near data</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Implementation of external stack (External
stack is required if the compiler is going to be used to implement an RTOS)</font></font></li>
</ul>
<li>
<font color="#CC0000"><font size=+1>Fixed problem of Compiler giving warning
when it did not find any reaching definition for a local static variable.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added code to backed for more optimized
access to arrays, structures &amp; unions placed in near space, now the
compiler will not use indirect addressing to fetch the data if the offset
is can be determined at compile time.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Changed the way the startup code works.
Documentation updated see section Startup for more details.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Completed implementation of external
stack. Documentation updated see section External stack.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Jean-Louis VERN added a LOT of optimizations
to the back end code generator, following is only a small list.</font></font></li>
<ul>
<li>
<font color="#CC0000"><font size=+1>gencjne improved</font></font></li>
<li>
<font color="#CC0000"><font size=+1>gencmp improved</font></font></li>
<li>
<font color="#CC0000"><font size=+1>genand improved</font></font></li>
<li>
<font color="#CC0000"><font size=+1>genor improved.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>_divuint improved</font></font></li>
<li>
<font color="#CC0000"><font size=+1>_divuint improved.</font></font></li>
</ul>
<li>
<font color="#CC0000"><font size=+1>Variables with absolute addresses now
considered as VOLATILE.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Dead Code elimination routine updated
: will not eliminate global assignments to global variables when function
call present.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Added file <b>at89S8252.h </b>to distribution
file supplied by "Michael Schmitt" &lt;mschmitt@mainz-online.de></font></font></li>
<li>
<font color="#CC0000"><font size=+1>Martijn van Balen &lt;balen@natlab.research.philips.com>.
Fixed problems in assembler and linker. Files affected are<b> asout.c</b>
and <b>lklist.c</b></font></font></li>
<li>
<font color="#CC0000"><font size=+1>Initialized character array variables
with no size were being generated one byte less in length.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>Access to bit fields within structures
mucho mucho improved.</font></font></li>
<li>
<font color="#CC0000"><font size=+1>The compiler was allocating space on
the stack for local variables (for reentrant functions) even when not needed,
this has been fixed, the compiler will not allocate space on the stack
for local variables if they are assigned to registers.</font></font></li>
</ul>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.0.21</font></b>
<br>&nbsp;
<blockquote>
<li>
<font size=+1><font color="#CC0000">Enhancement made to unsigned long &amp;
integer division by </font><font color="#FF0000">Martijn van Balen &lt;balen@natlab.research.philips.com></font></font></li>
<li>
<font color="#FF0000"><font size=+1>&nbsp;_gptrget.c &amp; _gptrput.c made
register bank independent&nbsp; by Martijn van Balen &lt;balen@natlab.research.philips.com>.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>sdccbuil.sh script fixed by Subject:
Markus Mathes &lt;mathes@prp.physik.tu-darmstadt.de></font></font></li>
<li>
<font color="#FF0000"><font size=+1>Martijn van Balen &lt;balen@natlab.research.philips.com>
found some problems with post &amp; pre increment of pointer variables
, these are fixed.</font></font></li>
</blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.0.20</font></b>
<p><font size=+1>Mostly enhancements in this release. Some bug fixes by
Jean Louis.</font>
<blockquote>
<li>
<font color="#FF0000"><font size=+1>Enhanced loop optimization (induction
variables), will generate more compact code for simple loops</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Enhanced some aspects of basic block
processing.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Added a couple more MCU header files.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Jean Louis made some enhancements in
SDCCgen51.c for better code generation.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Jean Louis also fixed some bugs in
the back end .</font></font></li>
</blockquote>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.0.19</font></b>
<br>&nbsp;
<blockquote>
<li>
<font color="#FF0000"><font size=+1>Assignment of array variable to a pointer
sometimes did not generate code.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>File containing intialised variables
ONLY with NO functions caused the compiler to crash</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Initialising global volatile variables
caused the compiler to fail.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Generated incorrect code when bit variables
were modified and subsequently used in a conditional statement.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Added 3 more header files for Atmel
MCUs , provided by Bernd&nbsp; Bartmann (Bernd.Bartmann@picard.isdn.cs.tu-berlin.de)</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Added a lot of peephole optimization
rules, some suggested by Daniel Drotos &lt;drdani@mazsola.iit.uni-miskolc.hu></font></font></li>
<li>
<font color="#FF0000"><font size=+1>Added check to make sure functions
do not return 'bit'.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Jean Louis-VERN (jlvern@writeme.com)
fixed bug for divide with literal in his code.</font></font></li>
</blockquote>
</blockquote>
<ul>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.0.18</font></b>
<br>&nbsp;
<ul>
<li>
<font color="#FF0000"><font size=+1>Function prototype problem when parameter
name in the prototype was the same as a local variables.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Added header files for Atmel AT89x51
&amp; AT89X52 provided by : Bernd&nbsp; Bartmann (Bernd.Bartmann@picard.isdn.cs.tu-berlin.de)</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Jean Louis-VERN (jlvern@writeme.com)
has made </font><font size=+2>substantial improvements</font><font size=+1>
to the Back-end code generation for </font><font size=+2>shift operations
</font><font size=+1>and
</font><font size=+2>multiplication.</font></font></li>
</ul>
</ul>
<dl>
<dl>
<hr WIDTH="100%">
<br><b><font size=+1>Release 2.0.17</font></b></dl>
</dl>
<dl>
<ul>
<ul>
<li>
<font color="#FF0000"><font size=+1>Empty if body with non-emtpy else body
cause compiler to give error message. Patch provided by Martijn van Balen
&lt;balen@natlab.research.philips.com></font></font></li>
<li>
<font color="#FF0000"><font size=+1>Loop optimizer missed induction variables
;</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Added option <b>--main-return</b> ,
will generate a '<b>ret</b>' instead of '<b>ljmp .</b>' when this option
is used.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Add support for user defined startup
code.</font></font></li>
</ul>
</ul>
<dl>
<hr WIDTH="100%">
<ul>
<li>
<font color="#FF0000"><font size=+1>generating incorrect assembler code
(new line missing), for '<i>long</i>' variables initialization when defined
in '<i>code'</i> space.</font></font></li>
<li>
<font color="#FF0000"><font size=+1><i>aslink</i> occasionally core dumped,
after it finished processing. (patch provided by&nbsp; Eric Norum &lt;eric@skatter.USask.Ca>
)</font></font></li>
<li>
<font color="#FF0000"><font size=+1>comparing <i>pointers with literals
</i>,
e.g. ptr == NULL gave error message, (NULL had to be explicitly cast to
the pointer type)</font></font></li>
<li>
<font color="#FF0000"><font size=+1>assigning <b>signed char</b> values
to <b>sfr</b>s generated duplicate code.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>definitions for CHAR_MIN, INT_MAX &amp;
LONG_MAX incorrect in limits.h</font></font></li>
<li>
<font color="#FF0000"><font size=+1>some more patches from Eric Norum &lt;eric@skatter.USask.Ca>
his description of the patches.</font></font></li>
<ul>
<li>
<font color="#FF0000"><font size=+1>1) asxxxx/lkmain.c:181</font></font></li>
<br><font color="#FF0000"><font size=+1>&nbsp;&nbsp; startp->f_type is
an integer variable and so should not be</font></font>
<br><font color="#FF0000"><font size=+1>&nbsp;&nbsp; compared with NULL.</font></font>
<p><font color="#FF0000"><font size=+1>2) asxxxx/lkmain.c:195</font></font>
<br><font color="#FF0000"><font size=+1>&nbsp;&nbsp; *ip is a character
variable and so should not be</font></font>
<br><font color="#FF0000"><font size=+1>&nbsp;&nbsp; compared with NULL.</font></font>
<p><font color="#FF0000"><font size=+1>3) asxxxx/lkmain.c:717</font></font>
<br><font color="#FF0000"><font size=+1>&nbsp;&nbsp; ctype[c] is a character
variable so the expression (ctype[c] != ILL)</font></font>
<br><font color="#FF0000"><font size=+1>&nbsp;&nbsp; is always true on
machines where char variables are signed.</font></font>
<p><font color="#FF0000"><font size=+1>4) SDCCmain.c:780</font></font>
<br><font color="#FF0000"><font size=+1>&nbsp;&nbsp; The preprocessor output
file was left in the temporary directory.</font></font>
<br>&nbsp;</ul>
<li>
<font color="#FF0000"><font size=+1>&nbsp;volatile variables when used
as loop induction variables caused the compiler to crash.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>compiler generated '<b>char * idata
icp</b>'&nbsp; as a generic pointer instead of 1 byte pointer.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>link file problem, when linking multiple
files together the compiler sometimes generated a bad file name.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>library function _<b>atol.c </b>had
the function <b>main</b> in it.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Incorrect code generated left shift
of <b>one byte</b> variables <b>'rl'</b> should be <b>'rlc' </b>. Patch
provided by : Jean-Louis VERN &lt;jlvern@writeme.com></font></font></li>
<li>
<font color="#FF0000"><font size=+1>Bug in <b>./cpp/cpplib.c </b>. Patch
provided by :Martijn van Balen &lt;balen@natlab.research.philips.com></font></font></li>
<li>
<font color="#FF0000"><font size=+1>Consecutive <b>'if'</b> statements
sometimes caused the compiler to crash. This bug was a fairly serious.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Incorrect code generated right shift
of <b>long variables</b> variables <b>'rlc'</b> should be <b>'rrc' </b>.
Patch provided by : Jean-Louis VERN &lt;jlvern@writeme.com></font></font></li>
<li>
<font color="#FF0000"><font size=+1>Xor, And and Or operation with bit
variable &amp; non-bit variable generated incomplete code.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Expression folding was too aggresive,
was corrupting parameters in certain situations.</font></font></li>
<li>
<font color="#FF0000"><font size=+1>Function prototypes declared with parameter
names , occasionally caused problems.</font></font></li>
</ul>
</dl>
<dl>
<ul>
<li>
</li>
</ul>
</dl>
</dl>
</body>
</html>