Digital-Research-Source-Code/CPM OPERATING SYSTEMS/CPM 68K/1.0X SOURCES/v103/parser/log

.he "C68 Parser"Change Log"Page %"
.de bg
.sp
.in +5
..
.de eg
.sp
.ne 8
.in -5
..
1. 4/12/82 - Undefined symbols not being detected
.bg
This was caused by expr setting the defined flag for the symbol and
then setting its type to label.  Solution was to check for undefined
symbol in expr and output error message there, since the only place
an undefined symbol can occur in a statement is after a goto.
This results in all undefined symbols in statements having one error
message output on the line the symbol is used, and all undefined
labels error messages being output once at the end of the function.
.eg
2. 4/12/82 - Operand stack underflow problem
.bg
Error recovery not very effective, skipped over closing right bracket,
which cascaded errors.  Changed synerr to put back the syncronizing
token, no matter which it was.  Also changed stmt to not call expr
when the first token on the line was right bracket.  This resulted
in "operand stack underflow" errors when a label immediately preceded
a right bracket, e.g. at the end of a function or at the end of
a switch statement.  Also added logic in expr and popopd so that
the "invalid expression" error would be output rather than
"operand stack underflow".
.eg
3. 4/13/82 - Error message cleanup
.bg
Changed "expression too complex" messages and "invalid
structure proto" message.
.eg
4. 4/13/82 - Missing =% operator
.bg
Lex did not check for =% operator, added it.
.eg
5. 4/16/82 - Bad function declarations
.bg
The functions lookup, tnalloc and sbrk were not declared as returning
pointers.
.eg
6. 4/19/82 - Added static declarations
.bg
Declaring functions and external variables as static did not work.
Added fixes to gettype, doextdef, dodecl and expr to correctly handle
static variable and function declartions.
.eg
7. 4/23/82 - Typedef stuff again....
.bg
Adding declarator info to typedef declaration in cast did not work
correctly.  This required adding typedef info carefully into the
type info already obtained from gettype.  The routine addtdtype
was added to do all this jazz correctly.  Also, declarator was changed
so that it would return the super-type info in the correct (rather
than reverse) order.  Both expr and dodecl were changed to reflect the
the addtdtype routine.  Also, it turned out that expr was being
recursively called the declarator (via cexpr).  This required the
careful setting of the pointers (opp, opdp, exprp, opap) to the
expression stacks and workspace.  This resulted in changes to doreturn
and to copyargs, which require the expression area be set up.
.eg
8. 4/26/82 - structure pointer initialization
.bg
A pointer to a structure could not be initialized, problem in
initlist checking only for a structure being an aggregate type.
added (btype(type)==STRUCT) && notpointer(type) into aggregate
type test.
.eg
9. 4/26/82 - memory fault after invalid conversion error
.bg
cvopgen was not setting rtp pointer if invalid conversion, resulting
in invalid address in tree.  Changed cvopgen so it would return
tp upon invalid conversion problem.
.eg
10. 5/4/82 - external definition syntax loop
.bg
The error "external definition syntax" would not necessarily skip
the offending token.  Changed next(RCURBR) to gettok()!=RCURBR to
insure token was skipped.
.eg
11. 5/4/82 - Char pointer array string init error
.bg
The syntax: char *x[] {"xx"}; would not work because the type
given to oneinit was ARRAY of CHAR.  Added code in initlist to
give oneinit the "basic" type of the data structure.  Also added
routine revsp.
This routine
doesn't fix anything, but it's probably clearer what's going on...
.eg
12. 5/4/82 - Declarator syntax problem
.bg
The syntax: int (*x)(); would not work because the check for a
valid declarator in getdecl was to check sp, a local register,
rather than dsp, the declarator symbol pointer.  The parenthesis
caused a recursive call, and the local sp was set in the recursive
instance, rather than the one that did the check.
.eg
13. 5/5/82 - Missing } error for non-aggregate inits
.bg
Got "missing }" for: char *p {"xx"}; had mistakenly changed
next(RCURBR)==0 to next(RCURBR) in initlist when fixing bug 10.
.eg
14. 5/10/82 - Braces around string init for char array
.bg
Got "invalid initializer" for: char x[] {"xx"}; was not checking for
this case in initlist.  Added code in initlist to check for this.
.eg
15. 5/12/82 - Integer initialization in long or pointer data type
.bg
The case: char *x[] { 0 }; only resulted in an integer initialization.
Fixed oneinit to extend int to long and output as long constant.
.eg
16. 5/20/82 - Long -1 initialization
.bg
The syntax: long x -1; initialized x to 1, added check in cexpr to
add check after '-' for CINT as well as CLONG.
.eg
17. 5/27/82 - sizeof structure prototype
.bg
The syntax: sizeof( struct x ) where x is a structure prototype name
was not working (although this is not part of OFFICIAL C syntax anyway).
Added code in expr to do right thing with type STRUCT.  Also saw and
fixed some STRPROTO bugs in gettype and dodecl.
.eg
18. 6/1/82 - int to pointer conversion not getting error
.bg
Test in maketree for integer to pointer conversion was not correct.
Changed test in maketree to flag any INT or UNSN assignment conversion if
the right hand side was not an constant.
.eg
19. 6/2/82 - int to pointer conversion again
.bg
Test in maketree fixing bug 18 was not quite correct, did not allow
*p =+ i; added check in maketree for only assignment operator.
.eg
20. 6/2/82 - line numbers on error messages
.bg
Line numbers on error messages sometimes off by one.  This is because
expr or stmt can gobble up one or two tokens, thereby incrementing the
line number.
.eg
21. 6/17/82 - unsigned initialization
.bg
Couldn't do: unsigned x = 1; There were no entries in oneinit for a
base type of UNSIGNED...
.eg
22. 11/23/82 - nonunique structure names
.bg
Added the system III capability to handle structures which have
elements which have the same name, but different types and/or
structure offsets.
.eg
23. 12/06/82 - forward referenced structures
.bg
Forward referenced structures are now allowed, checking to be sure
that they are defined is left until end of file is reached.
.eg
24. 12/13/82 - structure assignments						[vlh] 4.0
.bg
One structure may be assigned to another provided that the one on
the right is smaller than or equal to the left one in size.
.eg
25. 12/16/82 - Don't generate .globl's for external func's	[vlh] 4.0
.bg
Don't automatically generate .globl's for external functions, avoids
automatically loading routines which may never be used (like stdio).
.eg
26. 1/5/83 - sizeof as array size specification				[vlh] 4.0
.bg
Recursive call to 'getdecl' caused the size specified using sizeof
to be lost.
.eg
27. 1/12/83 - procedure/function name as identifier			[vlh] 4.0
.bg
Generated "tst.l #_ident" for "if (ident) ;" where ident is a procedure
or function name.  Changed this to generate as true instead.
.eg
28. 1/25/83 - outgoto macro									[vlh] 4.0
.bg
Made the outgoto procedure a macro, requires it's single parameter to
be a constant.
.eg
29. 2/8/83 - typedef of forward referenced structures		[vlh] 4.0
.bg
The dimension table pointer of a typedefed forward referenced structure
was being initialized incorrectly.
.eg
30.  2/12/83 - Pointer to structure comparisons to integers	[vlh] 4.0
.bg
Previously the integer value was being multiplied by the size of the 
structure in words.
.eg
31.  2/19/83 - Pointers assigned to structure array element [vlh] 4.1
.bg
struct {char d[10];} *p; char *q; q = &p->d;
.eg
32.  2/28/83 - For loop construction						[vlh] 4.1
.bg
Altered the code that is generated so that only one branch is inside
the for loop.  Better code is now generated.
.eg
33.  3/1/83 - While loop construction						[vlh] 4.1
.bg
Altered the code that is generated so that only one branch is inside the
while loop.  Better code is now generated.
.eg
34.  3/1/83 - Char and char * casting						[vlh] 4.1
.bg
Added the docast unary operator to fix the char cast problem in the
parser, changed the char * cast to do a int to long conv in the case
of a char *.
.eg
35.  3/7/83 - Unsigned/int global initialization			[vlh] 4.1
.bg
Initialization of integer to -32768 generated an error, as did unsigned
initializations > 32767.
.eg
36.  3/29/83 - string defines								[vlh] 4.1
.bg
Initialization's of strings has been broken up into dc.b's of no greater
than 32 bytes at a time.  This is to avoid building one which is too
long for the assembler and the editor.
.eg
37.  4/21/83 - structure alignment							[vlh] 4.1
.bg
Fixed the structure alignment call to salign which caused character pointers
to not be aligned properly.
.eg
38.  4/22/83 - structure fields with same names				[vlh] 4.1
.bg
Structure fields with the same names were only being inserted into the
symbol table once.  Required a test to ensure that all declarations of
fields are inserted and just statement handling is allowed to short
circuit it.
.eg
39.  5/2/83 - added '\f' recognization						[vlh] 4.1
.bg
Added recognizing and proper treatment of the form feed escaped 
character '\f'.
.eg
40.  5/10/83 - multi-dimensional arrays						[vlh] 4.1
.bg
The indices into multi-dimensional arrays were being computed incorrectly.
The arrayref routine was setting the proper dimension table pointer, and
then resetting it to it's parent's value.
.eg
41. 5/11/83 - int to long constant conversion				[vlh] 4.1
.bg
Constants were not being converted until the code generation pass.  For
this reason certain constants were not being calculated correctly.
The long constant 0x8000 was left as an integer constant value of -32768
and passed to the code generator.  The codegenerator does not know what
the original value was.  Constant conversions from integers to longs needed 
to be done in the parser.
.eg
42. 6/6/83 - return with null expression					[vlh] 4.1
.bg
A return call which included parenthesis but a null expression caused the
parser to generate an error and bus error.  There was no test of expr's
return value in case of an error in which the expression tree would be
non-existent.  Need to check for parens without expression.
.eg
43. 6/8/83 - sizeof array as a structure field				[vlh] 4.1
.bg
The sizeof all types of values was being treated as a single node.  If the
item in question was an array element ensconsed in a structure the value
was always determined to be the size of a pointer.  This required the addition
of a new subroutine which in the case of an array inside a structure traversed
the tree and took into account the two nodes nearest to the tips of the tree.
.eg
44. 6/13/83 - added block declarations						[vlh] 4.2
.bg
Reworked original link and movem to be done at the bottom of the routine
at which time the number of variables register and stack will be known.
This requires an extra branch at the top and bottom of each function.
.eg
45. 6/14/83 - added auto initilization						[vlh] 4.2
.bg
Added the recognizing of the equal sign as a local variable initialization.
Faked out expression by handling the symbol internally.  Turned on commastop
to ensure only one variable is parsed at a time.
.eg
46. 6/15/83 - local/global same name						[vlh] 4.2
.bg
Added the facility to the parser to allow for local/global/block/nested block
variables with the same name.  This required the addition of a scope level
to the symbol table node.  Global scope is 0.  Function parameters and
local variables are at level 1, etcetera.  Each left curly brace signifies
and additional nesting level.  This requires freeing symbols at every
right curly brace if there are any at that level.  Added an array to keep
track of what levels have variables declared so that the symbol table did
not have to be searched for every curly brace.
.eg
47. 6/21/83 - link file										[vlh] 4.2
.bg
Added the link file.  Writes link information (link and movem instructions)
into this file for remerging by the codegenerator.  This was added in order
to produce cleaner code with the addition of block variables and
local initializations.
.eg
48. 6/22/83 - structures									[vlh] 4.2
.bg
Rewrote all the code to deal with structures and their elements and how
and where they are put into the symbol table.  Necessitated the addition
of two more symbol table pointers: a child link (pointer from structure
to it's first element), and a sibling link (linking together all the
elements in a structure).  Every unnamed structure is now being added to
the symbol table with a fake name.
.eg
49. 6/27/83 - asm											[vlh] 4.2
.bg
Added the asm instruction to the parser in order to generate assembly
language code directly into the C source.
.eg
50. 6/28/83 - bit fields									[vlh] 4.2
.bg
Bit fields where not being aligned on a word boundary when necessary.
Also the value of s_dp which was used to decided the bit offset of
the field was not being reset if an int bit field followed a character
bit field.
.eg
51. 6/28/83 - structure initialization and alignment        [vlh] 4.2
.bg
Rewrote all code which deals with initializing structures.  It now
compares the sizes and types of the structure fields against the
objects which are used to initialize the structures.  It defines
space for padding properly.  Arrays inside of structures are now
being initialized properly.  They are truncated at the appropriate
space if required, otherwise they are terminated by a null byte
and padded to the appropriate size.  If they are truncated then
the parser generates a warning message.
.eg
52. 7/7/83 - scope problem                                  [vlh] 4.2
.bg
Variables which were declared on more than one scoping level were
not always being identified properly.  In particular, if a variable
is accessed on a nesting level different from it's declaration
it would not always choose the right declaration.
.eg
53. 7/13/83 - long vs. short vs. int						[vlh] 4.2
.bg
All references to int variables have been changed to shorts.  This is
for the sake of portability between regulus and the vax.
.eg
54. 7/20/83 - bit field initialization						[vlh] 4.2
.bg
Added a routine called bfield_init which outputs the word or character
once it is built.  Added code to oneinit to build the bit fields as
they were evaluated.
.eg
55. 7/20/83 - symbolic debugger flag						[vlh] 4.2
.bg
Added recognizing of the '-g' flag.  Output the label required before
the link instruction.
.eg
56. 7/25/83 - fixed structure assignments again             [vlh] 4.2
.bg
Move the structure assignment size test into the maketree routine.
Make a right side structure larger than a left only a warning.
Determine the size of the structures in the maketree routine before
the original tree nodes are altered beyond comprehension.
.eg
57. 7/25/83 - unsigned char/long							[vlh] 4.2
.bg
Initial changes required to allow unsigned char and unsigned long
types.
.eg
58. 7/29/83 - long constant conversion to int				[vlh] 4.2
.bg
Operations in which one operand is a long constant and the other is
an integer will cause the long constant to be converted at all times.
Caused an additional condition to be added to the conversion code in
the maketree routine.
.eg
59. 8/1/83 - frstruct pointer as struct element				[vlh] 4.2
.bg
A pointer to a forward referenced structure was being incorrectly
linked into the structure declaration.  Required several tests
and some additional code in the get_s_or_u and get_type routines.
.eg
60. 8/5/83 - function label test							[vlh] 4.2
.bg
Bad code was being generated for a statement of the form "if(func_name)".
The immediate value of the function label was being tested.  Fixed this
by adding a test in funcref() to not generate addresses if the operand is
a SYMBOL or NOT.
.eg
61. 8/8/83 - constant expression evaluation					[vlh] 4.2
.bg
Added a test to tree() to avoid adding conversion ops to an expression
which involves only long/short constants.  This ensures that they will
be evaluated by the parser in binopeval.
.eg
62. 8/9/83 - source file to cgen							[vlh] 4.2
.bg
Altered the macro which outputs the line number to the intermediate file
to also output the file name, for use in generating errors in the
code generator pass.
.eg
63. 8/10/83 - constant test for long						[vlh] 4.2
.bg
Changed the gettok() routines test of a constant as a long to check to
see if any of the high bits (0xffff0000) are set in the value.  As all
numbers are positive at the time of this evaluation it is the most sure
fire way to decide.
.eg
64. 8/12/83 - array dimension using sizeof					[vlh] 4.2
.bg
Fixed a rather obscure parser bug which involved specifying the dimensions 
of an array of structures using the sizeof operator on a structure.
Required a reset in the getdecl routine of dp if cdp had been changed
because of a sizeof operation.
.eg
65. 8/24/83 - 16 bit hex and octal constants				[vlh] 4.2
.bg
By definition of Kernighan and Ritchie all octal and hex constants which
fit as unsigned values into an integer size are defined to be integers.
.eg
66. 8/25/83 - 2 dimensional array initialization			[vlh] 4.2
.bg
Two dimensional arrays were not being initialized properly in 4.2...
Requires multiple calls to str_init.
.eg
67. 9/2/83 - labels and function names						[vlh] 4.2
.bg
Fixed the compiler to handle local-globals with local labels which have
the same name as functions.  Fixes in stmt.c to force a new symbol table
entry for the label.
.eg
68. 9/6/83 - floating point constant, w/o preceding digit   [vlh] 4.2
.bg
Added a gettok entry for PERIOD to check for a floating point constant
which does not contain a leading digit and thus process it as such.
.eg