![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to LOG CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Research Unix] / researchv10no / cmd / lcc|
Return to LOG CVS log | Up to [Research Unix] / researchv10no / cmd / lcc |
1.1 root 1: 10/25/88:
2: lcc is an ANSI C compiler available to friendly users on coma.
3:
4: Bit fields are parsed but ignored. Trigraph sequences and the extended
5: character set are not supported. The code for register variables needs
6: improvement. The only cases now known to yield incorrect code involve
7: C++ output and programs that run out of registers, but there must be
8: more. -g seems to work with pi, but it's very green.
9:
10: ANSI-style header files are in /usr/include/lcc, which lcc searches.
11: A few prototypes may differ from what is actually in libc.a; the
12: consequence is usually a type error. A header file for some V9
13: functions is on the way; more later on this.
14:
15: For a man page, run "man lcc" or "troff -man /usr/cwf/book/man/lcc.1.v9".
16: Mail bug reports to both cwf and princeton!drh (Dave Hanson).
17: Please include a short program that exposes the bug. Thanks.
18:
19: 11/8/88:
20: fixes several bugs with error handling, float and string constants,
21: register structs, and programs that exhaust register variables.
22: makes code for char and short int functions compatible with cc's code.
23:
24: 11/17/88:
25: fixes bugs in spills, in the declaration of longjump in lcc's
26: <setjmp.h>, and in the type system ("char" had compared equal to
27: "signed char", which ansi forbids). improves error recovery. plus
28: internal changes that are theoretically invisible.
29:
30: it also puts some formals and locals in registers; "man lcc" elaborates.
31: the increased demand for registers may expose compiler bugs that have
32: been in hiding. if you have regression tests, try them.
33:
34: 12/8/88:
35: fixes sizeof("x"), overzealous reuse of common subexprs, and a loop in
36: the spill emitter. improves error reporting and code for many ++/--
37: ops (but prefix versions on registers still need work). folds the
38: difference of two constant pointers. defines standard "offsetof"
39: macro.
40:
41: 12/13/88:
42: fixes compiler crashes on certain spills in nested conditional exprs.
43: fixes bugs in hex char constants, in exprs with chains of unsigned
44: mods and divisions, and in *x=*x when x is volatile. uses gcc's
45: preprocessor to offer more ANSI extensions. adds bit fields, tho
46: they're pretty green, the layout's unique (see the man page), and the
47: code needs optimization.
48:
49: 12/19/88:
50: changes some diagnostics. handles arbitrarily long lines. emits
51: slightly better code for switches and functions that return structs.
52: fixes an infinite loop when undeclared identifiers appeared in
53: initializers, an incorrectly formatted error message for illegal
54: characters, and bad code for x:goto x and for *f()=g() when g returns
55: a struct. ignores -O, because /lib/c2 can trash lcc's code. adds -n,
56: which compiles code to detect certain zero-pointer errors; the man
57: page elaborates.
58:
59: 1/5/89:
60: improves diagnostics. deletes warnings about unused return values,
61: except when they're structures. fixes minor errors in the standard
62: include files. stops losing the token following asm("..."). emits
63: slightly worse code for nested calls, in preparation for a mips
64: target. several internal changes that should have no effect.
65:
66: 1/9/89:
67: improves entry code for some functions. gets clr/inc instructions even
68: if the 0/1 is expressed oddly (eg, 0x1). corrects bad code for
69: (unsigned)c==0xff, for f(g().b) when b's offset >0, and for -u>0 for
70: unsigned u. returns to the old code for nested calls, so f(x?h():0),
71: f(x&&h()), and f(x||h()) work again. corrects minor flaws in string.h.
72:
73: 2/1/89:
74:
75: improves diagnostics. improves handling of long parameter lists, long
76: string constants, and integral constants between 2^31-1 and 2^32-1.
77: responds to constant overflow in a machine-independent fashion.
78: flattens nested calls (for the mips target). corrects bad code for
79: a[2*i]=f(a[2*i]), (double)(float)0.3, stab entries for initialized
80: statics, and some nested conditionals. assigns registers differently:
81: r0-5 are now for temporaries, and r6-11 for variables. fixes several
82: crashes and assertion failures: if you've had some, try them now;
83: otherwise ignore.
84:
85: 2/10/89:
86:
87: -A now warns about calls to functions without prototypes. diagnostics
88: now include only the line number and omit the character position.
89: empty declarations in parameter lists (eg, "f() struct s {int x;};")
90: are now errors, and other empty declarations (eg, "int;") now draw
91: warnings. no longer forbids const struct members (eg, "struct {const
92: int a,b;}"). several internal but theoretically invisible changes.
93:
94: 2/11/89:
95:
96: fixes an endless compiler loop and a bug in bit field initialization
97: that showed up only on the mips.
98:
99: 2/14/89:
100:
101: improves constant folding (eg, 0&&x, -1&x). fixes bugs: applying
102: sizeof to explicit array types sometimes made lcc loop or crash; some
103: dissimilar floating point constants (eg, -111111.0 and
104: -111111.00000000002) were treated as equivalent during constant
105: folding; some conditional exprs involving constants gave a compiler
106: error (eg, i = 5.6 && i ? 1 : 2;) or bad code (eg, *p = (int *)0 &&
107: *(int *)0 ? 4 : 5;). fixes a bug in the code that spills a temporary
108: register when they're all gone; numerical programs were the most likely
109: to have been effected.
110:
111: 2/17/89:
112:
113: fixes an assertion failure in the code that spills a register.
114:
115: 2/24/89:
116:
117: /usr/include/nlcc shipped. it's a directory of mostly new .h files for
118: lcc. in 10 days or so, these files will replace those now in
119: /usr/include/lcc, which lcc searches before /usr/include.
120:
121: fio.h, libc.h, and u.h are non-ansi local contributions from various
122: sources. the rest are intended to match the standard, which can be
123: irritating but is necessary for portability. the standard errno.h is
124: omitted because the local one suits ansi.
125:
126: users interested more in convenience than in portability are welcome
127: to edit fio.h, libc.h, and u.h and to add new .h files; do so on
128: bowell or your changes will get lost. don't edit the other headers --
129: they're maintained mechanically from another source. if you find a bug
130: in one of these, post cwf and princeton!drh. for portability, we want
131: to keep non-standard stuff out of these files, but we do want to hear
132: about bugs.
133:
134: these are just headers. lcc will continue to use cc's libraries
135: indefinitely.
136:
137: 3/1/89:
138:
139: in the absence of -g, removes immediate tail recursion from
140: non-variadic functions. old-style, variadic, tail-recursive functions
141: are optimized incorrectly, so give a prototype.
142:
143: -s now controls which switches become branch tables; the man page
144: elaborates. also, the man page now warns of bad code for switches over
145: 32kb; repairs soon.
146:
147: fixes bug that put formals in scope too early, which caused bad code
148: or core dumps for some programs like "int a;f(a)int a[sizeof a];{}".
149: no longer treats 1/0 and similar expressions as constant expressions.
150: corrects diagnostics and compiler crashes for certain erroneous
151: inputs.
152:
153: adds gauss to distribution list.
154:
155: 3/10/89:
156:
157: implements 2c's structure extensions, except for in-line structure
158: exprs. replaces -s with -d to avoid conflict with ld's -s. adds -v,
159: which prints commands as they are executed; see the man page. uses the
160: two-operand instructions like addl2 a little more often. places
161: initialized const data in a readonly segment. bags casel
162: instructions, so switches over 32kb now work. no longer omits
163: unreferenced, anonymous types from -g symbol tables. improves
164: diagnostics.
165:
166: corrects a few bugs in /usr/include/nlcc, which replaces
167: /usr/include/lcc on 3/14. speak now or forever hold your peace.
168:
169: 3/11/89:
170:
171: corrects bad code for switches and for "y>(float)0.0?y:-(double)y".
172: avoids an ld bug with external initialized consts.
173: deletes -B's "using" message.
174:
175: 3/16/89:
176:
177: /usr/include/lcc shipped, a couple of days later than promised. it's
178: exactly what's been soaking in /usr/include/nlcc, in which fio.h,
179: libc.h, and u.h were the latest local contributions (at least on
180: coma), and the rest were conformed with the standard, except that the
181: standard errno.h was omitted because the local one suits ansi.
182:
183: the old headers are in /usr/include/olcc, so try -I/usr/include/olcc
184: if you get in a jam. /usr/include/nlcc unshipped.
185:
186: users interested more in convenience than in portability are welcome
187: to edit fio.h, libc.h, and u.h and to add new .h files; do so on
188: bowell or your changes will get lost. don't edit the other headers --
189: they're maintained mechanically from another source.
190:
191: these are just headers. lcc continues to use cc's libraries.
192:
193: 3/23/89:
194:
195: fixes a register fragmentation bug by eschewing odd double registers;
196: if you use a lot of doubles, retry any regression tests that you might
197: have. fixes bugs in a few diagnostics. assert.h may now be included
198: more than once without calamity. -P prints function prototypes for all
199: defined functions and globals in a form suitable for use in header
200: files; see the man page. plus several internal changes that should
201: have no effect.
202:
203: 3/29/89:
204:
205: fixes several spill bugs; code with large double exprs is most
206: likely to have been bogus. accepts wide-character literals and treats
207: them like ordinary character literals. treats `long int' and `long
208: double' as types distinct from int and double, but with the same sizes
209: and alignments; they had been treated as identical types, which is
210: wrong. deletes bogus diagnostics for:
211:
212: int *const p = (int *const)0;
213: int *x = (f(), 0);
214: typedef int I;main(){I I(I);}
215:
216: 3/29/89:
217:
218: fixes two bugs new in this morning's (3/29) lcc:
219:
220: calls to char and short functions gave a fatal compiler error.
221:
222: type `unsigned long int' was missing; variables so declared were taken
223: to be `long int', so, for example, right shift of an unsigned long
224: erroneously extended the sign bit. the repair causes non-trivial
225: expressions involving long int, long unsigned, and long double to have
226: types int, unsigned, and double, respectively. this behavior should
227: not affect the generated code because ints *are* long ints, etc., but
228: it might mask type errors.
229:
230: 4/16/89:
231:
232: bug fixed: casts to enum types weren't treated as ints.
233:
234: new feature: in asm("string") occurrences of `%name' are replaced by
235: the address or register for the identifier `name' if `name' is
236: visible. the man page elaborates.
237:
238: /usr/include/lcc now holds only the headers specified by the ANSI
239: standard. The local headers (fio.h, libc.h, u.h) and a local version
240: of stdio.h are now in /usr/include/libc. lcc now searches the latter
241: before the former, so the default will find the local stdio.h. -N
242: tells lcc to discard its usual search list and use only directories
243: named in explicit -I args. So use "lcc -N -I/usr/include/lcc" to
244: ensure conformance with the ANSI standard. Report bugs in the standard
245: headers to cwf. Report bugs in the local headers to rob. Don't edit
246: either set directly -- they're maintained mechanically from central
247: sources, so edits will get lost.
248:
249: 4/19/89:
250:
251: some function definitions inconsistent their prototypes went
252: undetected (eg, extern int f(int); void f(x)int x;{}).
253:
254: long ints were erroneously promoted to ints. this bug showed up in
255: prototype errors (eg, extern f(long); f(n) long n; {}).
256:
257: implements 2c's in-line struct exprs ((struct foo) {...}). the code's
258: poor, and run-time alignment errors can occur for incomplete exprs.
259: improvements soon.
260:
261: structs have always been aligned to suit their strictest member. to
262: fix some alignment crashes on the mips, they are now additionally
263: guaranteed int alignment (four bytes on all current targets). so
264: recompile all modules that use structures composed entirely of chars
265: and shorts. yacc and lex produce some.
266:
267: 4/28/89:
268:
269: "-(unsigned long)x" crashed the compiler. bit fields and functions
270: were erroneously permitted as arguments to sizeof. pointers to enums
271: were not treated as pointers to ints, which drew erroneous type errors
272: (eg, in "enum {...} *ap; int *x = ap;").
273:
274: local arrays are now guaranteed int alignment. the old, weaker
275: alignment sometimes made array initialization code crash on the mips.
276:
277: "#pragma ref id" now simulates a reference to id; it may be used to
278: suppress a warning about an otherwise unreferenced variable (eg, a
279: sccs id).
280:
281: -N now turns off the non-ansi local extensions in addition to restricting
282: the search for header files to only those directories named by -I.
283:
284: 5/1/89:
285:
286: fixes bug in some subscript computations for multi-dimensioned char
287: arrays of length 4 or less. this bug occurred only in the presence of
288: char arrays in which each element was initialized or initialized
289: structs with bit fields.
290:
291: 5/11/89:
292:
293: improves code for arithmetic and boolean ops on chars and shorts. if
294: you use these a lot, run any regression tests that you have.
295:
296: the algorithm used to induce register variables has been changed.
297: registers are assigned to locals and parameters if their weighted
298: reference count is at least 3. each reference counts adds X, where X
299: starts out at 1 and is multiplied by 10 in each loop, by 1/2 in each
300: then/else, and by 1/10 in each switch.
301:
302: 4/28's fix that made `pointer to enum' equivalent to `pointer to int'
303: was too permissive. warnings are now issued for assignments between
304: `pointer to int' and `pointer to enum'. all other combinations are
305: errors.
306:
307: structs are no longer guaranteed int alignment. they are still aligned
308: to suit their strictest member. recompile modules that use structures
309: composed entirely of chars and shorts. yacc and lex produce some.
310:
311: tail recursion is no longer removed. errors occurred when locals were
312: aliased.
313:
314: some valid combinations of qualified pointers in result arms of ?:
315: caused erroneous diagnostics, and incorrect uses of types pointer and
316: int in result arms of ?: went undetected.
317:
318: some subscript computations for multi-dimensioned char arrays of
319: length 4 or less were incorrect because their types were mangled.
320: this bug occurred only in the presence of char arrays in which each
321: element was initialized or initialized structs with bit fields.
322:
323: casts from integer types to long double were silently ignored. real
324: constants suffixed by "l" or "L" were erroneously type double; they're
325: now type long double.
326:
327: UINT_MAX was missing from <limits.h>. time_t, wchar_t, size_t and
328: clock_t are now protected from redefinition. the compiler erroneously
329: accepted and silently ignored preprocessor statements; except for
330: #pragma, they now draw an error.
331:
332: 5/19/89:
333:
334: f(){int (*a)[];(*a)[5] = 0;} is now accepted.
335:
336: variables referenced mainly in deeply nested control structure were
337: erroneously reported as unreferenced.
338:
339: no more arbitrary limit on the number of scope levels.
340:
341: 5/27/89:
342:
343: Type casts were erroneously accepted as l-values. On the vax, bad code
344: was silently generated for signed div and mod ops when the destination
345: was a (signed or unsigned) char or short.
346:
347: 6/2/89:
348:
349: now accepts "int a = 3;extern int a;". some constant expressions with
350: casts were not accepted as constants (eg, ~((unsigned short) 0)).
351: fixes bad code for struct args whose size is not divisible by four.
352:
353: 6/8/89:
354:
355: improves diagnostics. some valid field references in extended
356: structures were erroneously flagged as errors. e&0 always returned 0
357: and never evaluated e, even if e had side effects. ignores
358: unrecognized preprocessor control lines.
359:
360: 8/3/89:
361:
362: fixes compiler crashes caused by long string constants with lots of
363: escaped characters.
364:
365: the va_start macro in stdarg.h did not handle char or short arguments
366: correctly.
367:
368: on the vax, bad code was generated for some assignments to unsigned
369: char (eg, "char a[10]; unsigned char b = a[i];").
370:
371: signed divisions by a constant power of two are no longer performed by
372: shifts; the results were inconsistent with the standard, which
373: requires that (a/b)*b + a%b == a.
374:
375: some illegal casts in constant expressions were erroneously accepted
376: (eg, "int x, y=(int)&x;").
377:
378: missing tags were erroneously permitted in some struct declarations
379: (eg, "struct *p;").
380:
381: now allows a terminal comma in enums, unless -N (strict ANSI) is
382: specified.
383:
384: warns about switch statements with no cases and about functions
385: declared static but never defined or referenced.
386:
387: with -A, warnings about missing prototypes are now issued only once
388: instead of once per call.
389:
390: doug has merged the man pages for cc and lcc.
391:
392: 8/8/89:
393:
394: fixes bad code for x=y<<1 when x is a short.
395:
396: 8/29/89:
397:
398: preprocessor generated lines of the form `# n "name"' and `# n' are
399: accepted.
400:
401: repeated initializations of an array type defined by a typedef
402: (typedef char str[]; str a = "hi", b = "there";) elicited incorrect
403: error messages.
404:
405: on the mips, some initialized structures that began with a char or
406: short were misaligned.
407:
408: shamash and bartok added to the shipping list.
409:
410: 9/16/89:
411:
412: better code for comparing bit fields with 0 and for setting them to all
413: 1s or 0s; manipulating 1-bit fields is as efficient as using flags with
414: explicit | and & operators.
415:
416: eliminates more dead code.
417:
418: eliminates multiple copies of some constants.
419:
420: fixes bad code for multiple assignment to bit fields in the same word
421: (eg, struct { unsigned x:1,y:1; } *p; p->x = p->y = 1;).
422:
423: fixes bad code for some logical operators with constant operands (eg,
424: x||1).
425:
426: fixes a compiler crash on the mips for some calls that pass a float and
427: return a struct.
428:
429: 9/29/89:
430:
431: takes care to emit only non-empty .s files on the mips, whose assembler
432: cares. corrects bad mips code in some cases when one of the first four
433: formals gets assigned to a register and is declared (old-style) to be a
434: (single-precision) float.
435:
436: 10/4/89:
437:
438: in unions with bit fields, the offsets for subsequent fields were
439: wrong. unions comprised entirely of bit fields were thought to be of
440: size 0, which drew a bad diagnostic. math.h's HUGE_VAL is now less
441: conservative. several numbers in limits.h were wrong.
442:
443: 10/21/89:
444:
445: hex escapes now properly include the longest possible run of hex digits
446: after the \x, even when the value is too big for a char. lcc truncates
447: and warns about octal and hex escapes with values that won't fit in 8
448: bits. it also warns about undefined escapes like \c.
449:
450: local declarations of static functions that were previously defined
451: drew erroneous mismatched declaration warnings.
452:
453: better code for e1?e2:0.
454:
455: fixes bad struct returns for functions that explicitly declare more
456: register locals than lcc allows for the target.
457:
458: different code for spills.
459:
460: on the 68020, fixes bad code for "return i%j".
461:
462: limits.m4 now casts INT_MIN to int and LONG_MIN to long. without the
463: casts, the 0x80000000 ended up unsigned.
464:
465: 10/24/89:
466:
467: for new-style functions, float actuals corresponding to int or long
468: formals went unconverted. eg, "f(int x) { float y; f(y); }" failed to
469: convert y.
470:
471: error messages for the redeclaration of enum ids misreported the
472: location of the previous declaration.
473:
474: on the vax, <unsigned constant> >> <anything> generated bad code.
475:
476: on the vax, casts that narrowed and then widened an indexing expr
477: generated bad code.
478:
479: limits.h substitutes exprs for some hard constants to reduce machine
480: dependencies.
481:
482: 11/3/89:
483:
484: a missing } at the end of file drew a garbled diagnostic.
485:
486: lcc crashed when a conditional expr fed a binary op that fed a
487: bit field assignment.
488:
489: lcc crashed when the value of x?y:0 was not used.
490:
491: new driver for lcc. no -M. preprocessor and compiler are now
492: connected with a pipe. flags must now precede file names and
493: libraries.
494:
495: 11/7/89:
496:
497: on the vax, fixed bad code for *s&&*s++.
498:
499: 11/17/89:
500:
501: casts of constants to illegal types caused an assertion failure. no
502: longer warns about switches with only the default case. now allows
503: more generated labels. the new driver hung on long programs with too
504: many errors.
505:
506: on v9 vaxes, now supplies -J to the assembler.
507:
508: on tempel, RAND_MAX was too small. on all mips machines, lcc exited
509: with status 0 even with errors. an omission from the mipsco docs led
510: to an incompatibility with code from cc for new-style functions with
511: single-precision floats as arguments 2, 3, or 4.
512:
513: 12/26/89:
514:
515: constants burn less time and space. large initializations compile much
516: faster.
517:
518: vertical tabs and form feeds were previously treated as newlines;
519: they're now treated as blanks.
520:
521: ~~x and -(-x) have been replaced with just x.
522:
523: specifying both -A and -n drew inappropriate missing prototype
524: messages.
525:
526: types defined in parameter lists were inaccessible within the
527: function.
528:
529: the compiler crashed when the value of "x,y" was used and y was a
530: relational or conditional expr.
531:
532: the compiler crashed when a void (eg, via a cast) was assigned to a bit
533: field.
534:
535: on the mips, an invalid instruction was generated for a register float
536: used as argument 2, 3, or 4. the assembler rejected it.
537:
538: on the mips, literals have been moved to the text segment.
539:
540: on the mips, -p now works. as with cc on (only) these machines, it
541: does not count calls. for that, use "pixie", which works with lcc's
542: a.outs.
543:
544: on the 68020, -g generated incomplete symbol table information for
545: structures that were used before being defined
546: (eg "struct A; g(struct A *x) {} struct A { int a,b; };")
547:
548: CLOCKS_PER_SECOND is now 1000000 everywhere. that's now consistent
549: with the clock function on the mips and 68020 machines, and it's
550: irrelevant on the vaxes, because their libraries don't include the
551: function.
552:
553: numerous internal improvements, all theoretically invisible.
554:
555: 2/2/90:
556:
557: sizeof erroneously returned an int. it now returns an unsigned as
558: dictated by ANSI. WARNING: the pre-ANSI sizeof returned an int, so lcc
559: has been masking a potential problem with the conversion of pre-ANSI
560: programs.
561:
562: casts in integer constant exprs in array bounds, enum defs, and field
563: widths sometimes gave bad values.
564:
565: string constant arguments of asm directives were handled wrong, which
566: made the compiler crash or emit bad diagnostics.
567:
568: comparisons of ints and pointers like i==(int*)1 crashed the compiler.
569:
570: arrays with unspecified bounds are now noted as an `incomplete array'
571: in error messages.
572:
573: reuses of common subexprs in conditional exprs sometimes generated bad
574: code.
575:
576: on the vax, bad code was generated for unsigned right shift by a
577: non-constant when r0 was busy.
578:
579: on the suns, the register save mask at function entry and exit for
580: floating point registers was incorrect.
581:
582: lcc -E foo.i (foo.s) erroneously compiled (assembled) the given file.
583:
584: several theoretically transparent internal changes.
585:
586: 2/20/90:
587:
588: lcc no longer crashes when presented with local arrays of undefined
589: structs or with calls to functions that return const structs.
590:
591: u%2^n now generates the same code as u&(2^n-1) for unsigned u and
592: constant 2^n.
593:
594: on the suns, RAND_MAX is now 2^31-1, which contradicts the man page but
595: is consistent with what rand() seems to do.
596:
597: 2/21/90:
598:
599: fixes bad code introduced yesterday on all targets for u%2^n for
600: unsigned u and constant 2^n.
601:
602: 2/28/90:
603:
604: multiple assignments to bit fields in the first word of a struct (like
605: struct {unsigned a:1,b:1;} x; x.a=x.b=1;) no longer generate bad code.
606:
607: extra parentheses in declarations (like int *(x(int));) sometimes crashed
608: the compiler.
609:
610: the driver sometimes erroneously removed input .o files.
611:
612: -Wp<arg> now passes <arg> to the preprocessor. -Wa, -Wl, and -Wf do
613: likewise for the assembler, linker, and compiler proper. <arg> is
614: passed untouched; if the recipient expects a leading dash, give it
615: explicitly.
616:
617: -T is gone. use -Wp-T instead.
618:
619: 3/2/90:
620:
621: fixed incorrect optimization of i%1 on all machines.
622:
623: on the vax, -Wf-k now avoids some of the instructions that may fail
624: when used to access i/o space. it does not protect against all
625: dangers: don't touch i/o space via the program counter, floating point
626: instructions, struct copies (including struct args and return values),
627: or calls (eg, don't index a table of function pointers with an i/o
628: cell). i can't thoroughly test this option myself, so beware.
629:
630: 3/8/90:
631:
632: relaxes an overzealous assert() in the register spiller.
633:
634: many internal cleanups, all theoretically invisible outside.
635:
636: 3/17/90:
637:
638: the compiler was crashing on x+=(const int)1
639:
640: on the mips, fixes bad code for functions that take a double or float
641: as arg#1 and return a struct.
642:
643: on the vax, slightly different tradeoffs between movl $_x and moval _x.
644:
645: 3/26/90:
646:
647: better assignment of variables to registers. explicit register dcls
648: are still obeyed first, but any remaining regs are now assigned to
649: variables in order of estimated payoff.
650:
651: bad floating constants like 1e and 1e+ were erroneously accepted.
652:
653: diagnostics for illegal break and continue statements were garbled.
654:
655: inline struct expressions now accept struct-valued fields, eg
656: struct Point { int x, y; } p, q;
657: struct Rectangle { struct Point min, max;} r = (struct Rectangle){p, q};
658:
659: on v9 vaxes, lcc now emits some symbol table data even without -g. the
660: intent is to mimic cc, though lcc documents globals as well as locals
661: and formals.
662:
663: compiler flags are no longer required to precede filenames. like cc,
664: they may appear anywhere but are processed before any files.
665:
666: lcc, like cc, now removes a .o file iff exactly one source file (.c,
667: .s, .i) appears and no other file (source, object, library) or -l
668: option appears.
669:
670: 5/9/90:
671:
672: warnings are no longer issued for calls to struct functions in which
673: the returned struct is not used.
674:
675: lines containing only white space and one # are now ignored.
676:
677: -N now suppresses recognition of asm("...").
678:
679: ansi multibyte characters like 'abc' are now accepted, and lcc warns
680: about ignoring the excess characters.
681:
682: -P now prints char* even if a typedef for char* has been defined.
683:
684: sizeof(L'x') is now correctly the same as sizeof(wchar_t), which is 1.
685: sizeof('x') is still the same as sizeof(int), as it must be.
686:
687: for constant y, 0&&y (0||y) erroneously returned y instead of y!=0.
688:
689: initializing char a[N] with a string of exactly N chars caused the
690: trailing null byte to be omitted from other N-char strings appearing
691: before the offending initialization.
692:
693: the value of bit field assignments was erroneously taken to be the
694: right operand instead of the value of the left operand. eg, for
695: an unsigned 3-bit field x, f(b.x=15) passed 15. it now passes 7.
696:
697: casts and structs returned by functions were erroneously permitted as
698: l-values.
699:
700: exprs like &(((struct foo)0)->x[0]) and array-sizeof(array) were not
701: recognized as constant. they were thus erroneously forbidden in
702: initializers.
703:
704: casts in the operands of ?: sometimes caused an assertion failure.
705:
706: using the value from a postincrement or postdecrement of a bit field
707: caused an assertion failure.
708:
709: legal comparisons of qualified and unqualified compatible pointers were
710: erroneously diagnosed as type errors.
711:
712: for function f, &f was erroneously diagnosed as an error.
713:
714: types `function returning T' and `array of T' were not uniformly converted
715: to `pointer to function returning T' and `pointer to T', respectively.
716:
717: char *a; void *b; a?b:c was erroneously diagnosed as a type error.
718:
719: initialization of automatic structs with const fields was erroneously
720: diagnosed as an assignment error.
721:
722: types other than int, signed int, or unsigned int were erroneously
723: permitted for bit fields and const and volatile qualifiers were
724: erroneously forbidden.
725:
726: repeated block-level declarations for functions were erroneously
727: diagnosed as redeclaration errors.
728:
729: valid assignment of pointers to qualified or unqualified versions of
730: compatible types was erroneously diagnosed as a type error, e.g.,
731: const char *a;const char *volatile *b=&a;
732:
733: on the 68020, lcc gave bad code for x=f(x) where f returns a struct.
734:
735: on the 68020, lcc erroneously converted floats and doubles to ints by
736: rounding. it now correctly truncates any fraction.
737:
738: on the 68020 and mips, lcc gave bad code for calls in which the
739: function was computed with another call, like (*g(f(1)))(2).
740:
741: on the mips, -p caused prof to report incorrect file names.
742:
743: on the mips, fixes bad code for f(x<op>y) when x is a floating point
744: register and <op> is an augmented assignment.
745:
746: on the vax, fixes bad code for signed comparisons of chars with
747: constants not in [-0x80..0x7f] and of shorts with constants not in
748: [-0x8000..0x7fff]. lcc treats plain chars as signed chars, so they
749: were afflicted too.
750:
751: on the vax, fixes bad code for u>0 and u<=0, for any unsigned expr u.
752:
753: the dummy parameter names were removed from the ansi include files.
754: eg, f(int x) is now f(int) instead.
755:
756: limits.h was rewritten to use simpler exprs and to avoid casts.
757: #if misevaluated some of the old exprs, and it doesn't do casts.
758:
759: don't use the unsigned constant suffix (eg, 0u) in #if commands.
760: the preprocessor can't cope.
761:
762: numerous theoretically invisible internal improvements.
763:
764: 5/9/90:
765:
766: The lcc compiler now passes the conformance section of Version 1.09 of
767: the Plum Hall Validation Suite for ANSI C, subject to the restrictions
768: below.
769:
770: We do not have a conforming preprocessor or library.
771:
772: The lcc command must use -N and enable trigraphs with -Wp-T. It must
773: use -I to search only the ANSI standard headers. It may need to
774: undefine some of the system-specific pre-defined symbols; -v exposes
775: these.
776:
777: The only versions of lcc tested to date are the ones running on the SGI
778: 4D/240S under IRIX System V Release 3.2.1, on the VAX 8550 under
779: UNIX V10, and on the Sun-3 under SunOS Release 4.0.
780:
781: 5/11/90:
782:
783: for float constant x, -x was not recognized as constant.
784:
785: for volatile x, x++ drew unwarranted warnings.
786:
787: 7/6/90:
788:
789: stores enums in unsigned/signed chars, unsigned/signed shorts, or ints
790: depending on the range of the enum constants. this change may alter
791: struct layouts and the size of globals, so discard old object files
792: that use enums. it also causes diagnostics when new-style declarations
793: are mixed with old-style definitions for functions with enum arguments,
794: so enum foo {a,b,c}; int f(enum foo); ... f(x) enum foo x; {...}
795: is equivalent to int f(char); ... f(x) char x; {...}
796: which is an error because the prototype in the declaration doesn't
797: match the prototype inferred from the definition (which is int f(int)).
798:
799: permits up to 127 formals.
800:
801: limits identifiers and numeric constants to 64 characters;
802: string constants can have up to 4096 characters.
803:
804: better diagnostics; some warnings are now errors.
805:
806: eliminates more branch chains.
807:
808: on the mips, lcc now truncates constant shift counts to 5 bits. the
809: assembler did this before, but it squawked first.
810:
811: -t produces code to print trace messages at function entry & exit.
812:
813: the driver now ignores duplicate `.o' files, eg, `lcc x.c x.o' is
814: identical to `lcc x.c'.
815:
816: provides errno.h, which simply includes /usr/include/errno.h.
817:
818: numerous internal improvements, all theoretically invisible.
819:
820: errors fixed:
821:
822: erroneously permitted 0x as a hexadecimal constant.
823:
824: failed to treat (double)<unsigned constant> as constant.
825:
826: did not detect overflow in enumeration constants, eg, for B in
827: enum {A=INT_MAX, B}. it still fails to detect underflow/overflow in
828: constant expressions, which it really ought to do.
829:
830: erroneously permitted declarations and definitions for functions
831: returning unknown structures.
832:
833: erroneously permitted application of unary + to any scalar type.
834:
835: erroneously permitted comma expressions in constant expressions.
836:
837: erroneously permitted negative bit field widths.
838:
839: did not recognize simple function types in casts, eg, int().
840:
841: incorrectly used the type of x or y for c?x:y where c is a constant;
842: the correct type is the composite type of x and y; eg, the type of
843: 1?(char*)p:(const char*q) is const char* not char* .
844:
845: `lcc -g foo.i' emitted incorrect line number information if foo.i
846: didn't include `# n file' directives.
847:
848: -g did not emit line number information for initialized locals.
849:
850: 7/9/90:
851:
852: on the vax, lcc generated code that wouldn't even assemble for
853: char c;unsigned short *v;f(){c = v[c];}
854:
855: 8/27/90:
856:
857: -b produces code that writes an expression-level profile into
858: prof.out. bprint turns the data into an annotated listing, and lcc
859: -Wf-a uses it to improve register assignments. the man pages for lcc
860: and bprint elaborate.
861:
862: warns about implicit returns for functions that return other than int
863: or void; -A includes int functions.
864:
865: -E now preprocesses unsuffixed file arguments.
866:
867: sign extends integer right shift in constant expressions on all targets.
868:
869: many internal improvements, all theoretically invisible.
870:
871: errors fixed:
872:
873: -t caused an erroneous redeclaration message for printf.
874:
875: failed to convert x from double to float in f(x) const/volatile float x; {...}
876:
877: erroneously forbid const/volatile enum {...} x in switch (x) and in int y=x.
878:
879: erroneously forbid enum bit fields for enum types that are compatible
880: with int, signed int, or unsigned.
881:
882: variadic functions with structure arguments did not work; 1 and 2-byte
883: structs still fail on big endians (non-DEC MIPS, Motorola, SPARC), and
884: 8-byte structs and those holding doubles fail on all MIPSes.
885:
886: with -g on V10 VAXes, bad symbol table names for anonymous structs,
887: unions, and enums.
888:
889: with -g on the Sun (68020 & SPARC), bad symbol table entries for
890: mutually referential structs.
891:
892: with -g, misplaced or omitted symbol table entries for statics,
893: externs, typedefs, and types defined in nested blocks.
894:
895: assertion failure on illegal casts of constants, eg, (struct foo)0.
896:
897: assertion failure on switch(x.y) {...} where y is an undefined field.
898:
899: 8/28/90:
900:
901: on the vax, an unsigned div/mod common subexpr that was also an
902: argument caused lcc to loop. on the sparc, it happened for all
903: int/unsigned div/mod/mul operations.
904:
905: 9/5/90:
906:
907: now permits any number of parameters.
908:
909: volatile locals and parameters are no longer assigned to registers.
910:
911: with -g, parameters are no longer assigned to registers; some debuggers
912: couldn't cope.
913:
914: generated bad symbolic addresses in constant expressions involving
915: externals declared inside blocks.
916:
917: misdiagnosed as errors some multiple assignments of returned structs,
918: eg, struct p {...} f(); struct r {struct p c;} x,y; x.c = y.c = f();
919:
920: on sparcs, generated unaligned code after some initialized const data.
921:
922: 9/13/90:
923:
924: erroneously elided calls to struct functions in conditional expressions
925: where the resulting value went unused, eg, i?y:f()
926:
927: assertion failure on conditional expressions with calls to struct
928: functions, eg, x=i?y:f()
929:
930: improved diagnostics.
931:
932: theoretically invisible internal improvements.
933:
934: 9/14/90:
935:
936: lcc crashed on f(){const struct { int a; } *fp;fp++;}
937:
938: 9/21/90:
939:
940: dropped backslashes that appeared at positions evenly divided by 4096.
941:
942: on v10 vaxes, now emits stab entries for locals even without -g,
943: for consistency with cc.
944:
945: on sparcs, fixes bad code for some argument lists that perform
946: a mul/div/mod after the third argument.
947:
948: 10/22/90:
949:
950: improvements/changes:
951: warns about static declarations that specify incomplete types.
952:
953: diagnoses declarations that specify objects with sizes >= 2^31.
954:
955: detects overflow in constant expressions and treats such expressions
956: as non-constant expressions; compiling old code may fail.
957:
958: better register allocation for leaf functions. on sparcs, better
959: register allocation for all functions.
960:
961: elides unreferenced locals.
962:
963: warns about bit-field initializers that are too big.
964:
965: with -N, duplicate global definitions in separately compiled
966: files cause loader errors; compiling old code may fail.
967:
968: improved diagnostics.
969:
970: errors fixed:
971: erroneously permitted static function declarations in blocks.
972:
973: erroneously diagnosed valid multiple declarations within blocks.
974:
975: computed incorrect composite types from multiple declarations.
976:
977: with -g, the Sun assembler choked on long symbol table entries;
978: continuations are now emitted.
979:
980: with -t, assertion failure on `return f(x)' where f returns a struct.
981:
982: bad values for (unsigned)d when 2147483647 < double d <= 4294967295.
983:
984: erroneously forbid carriage returns in strings.
985:
986: 10/23/90:
987:
988: fixes floating overflow in constant multiplications in which one
989: nonzero argument is between -1.0 and 1.0.
990:
991: 11/28/90:
992:
993: Now accepts #line. It has for a month, but I forgot to say so before.
994:
995: Now ignores multiple occurrences of just `.o' file arguments.
996: Formerly, it ignored multiple occurrences of ALL file arguments.
997:
998: With -A, now warns about declarations and casts with prototype-less
999: function types. Formerly, it warned only at calls.
1000:
1001: Warns about casts from pointers to integral types that are too small.
1002:
1003: Overflow in constant exprs has been demoted from an error to a warning.
1004:
1005: Conversions between pointers to objects and pointers to functions are
1006: now properly diagnosed.
1007:
1008: Void and struct operands to bitwise and, or and xor operators no longer
1009: cause an assertion failure.
1010:
1011: With -t, an illegal void return value or argument no longer causes an
1012: assertion failure.
1013:
1014: On MIPSes, functions with 3 or more float arguments no longer cause an
1015: assertion failure.
1016:
1017: On Suns and V10 Vaxes, now detects overflow in float and double
1018: constants. It is supposed to do so on all targets, but some of them
1019: have sub-standard strtod's.
1020:
1021: On Suns, lcc now accepts the ld options -Bstatic and -Bdynamic.
1022:
1023: On SPARCs, fixes bad code when passing structures followed by
1024: floats/doubles.
1025:
1026: On SPARCs, -p and -pg now work.
1027:
1028: On V10 VAXes, formals may now get assigned to registers. An error
1029: introduced several months ago had been forcing them into the stack.
1030:
1031: 12/6/90:
1032:
1033: The compiler crashed after diagnosing some syntax errors involving
1034: top-level functions (eg, "void *a();static b(int)c(){}").
1035:
1036: With -b, the execution count for e2 in "for(e1;e2;e3)S" was wrong for
1037: loops not guaranteed to execute at least once.
1038:
1039: On the MIPS, the compiler crashed when frequently referenced locals
1040: were used in deeply nested loops. The reference count estimate
1041: overflowed.
1042:
1043: On the SPARC, fixes (again!) bad code when passing structures followed
1044: by floats/doubles.
1045:
1046: On the SPARC, adds missing prototype to infinity() in math.h.
1047:
1048: 12/13/90:
1049:
1050: Since 10/26/90, -N has arranged loader diagnostics for duplicate global
1051: definitions in separately compiled modules. The arrangement, however,
1052: placed uninitialized globals in the data segment, which made some .o
1053: files huge. Such globals have now been shifted back to the bss
1054: segment, saving space without losing the loader diagnostics. On the
1055: MIPS, discard old .o files created with -N; they might be compatable
1056: with new ones, but only an ld expert could guarantee it.
1057:
1058: [Without -N, uninitialized globals are emitted as "common" blocks, so
1059: loaders -- contrary to ANSI -- may silently merge duplicate
1060: definitions. -N is overloaded: it also forbids extensions and searches
1061: for includes in only directories given by -I options.]
1062:
1063: lcc crashed when unknown enum types were used in contexts that required
1064: their sizes.
1065:
1066: lcc crashed or died with an assertion failure when parameter names were
1067: missing from new-style function definitions (eg, f(int){}).
1068:
1069: bprint crashed when prof.out had data for functions with >200 execution
1070: point counters.
1071:
1072: 12/21/90:
1073:
1074: lcc issued an erroneous ``lvalue required'' diagnostic for x[0]=f()
1075: when f returns a struct.
1076:
1077: 1/24/91:
1078:
1079: now reports as errors identifiers longer than 64 characters when they
1080: may cause parsing errors (because they span input buffers).
1081:
1082: erroneously forbid initialized local const arrays:
1083: f(){const char msg[]="hi";}
1084:
1085: erroneously warned about missing return values for:
1086: const void f(){}
1087: volatile void f(){}
1088:
1089: erroneously permitted function definitions with invalid uses of void
1090: arguments, some of which caused the compiler to crash:
1091: f(void,...){}
1092: g(void, int){}
1093: h(void, void){}
1094:
1095: aborted with an assertion failure on casts of a return value to void:
1096: void f(){return(void)23;}
1097:
1098: crashed when printing warnings about unreferenced const volatile statics.
1099:
1100: [I can't reach the Sun-3's or quetzal now. I'll update these machines
1101: later.]
1102:
1103: 2/14/91:
1104:
1105: Misidentified characters with ASCII codes greater than 127.
1106:
1107: switch(x) where x is a struct or union caused an assertion failure.
1108:
1109: On the vax, silently generated bad code for
1110: *<expr with ++ or --> = <expr with unsigned div/mod>
1111: (This was fixed last time, but I forgot to announce it.)
1112:
1113: On the vax, does a better job of using two-operand instructions (eg,
1114: addl2) and deleting gratuitous tst instructions.
1115:
1116: The warnings have been regularized somewhat. Use "-A -A" to flag
1117: missing prototypes and as many undefined behaviors as possible; use
1118: "-A" to flag little more than missing prototypes; use no -A options for
1119: relative quiet.
1120:
1121: with -A, warns about:
1122: missing prototypes in function types
1123: missing return value in a return from an int function
1124: *assignments between pointers to ints and pointers to enumerations
1125: *conversions from pointers to smaller integral types
1126:
1127: a second -A also warns about:
1128: non-ANSI language extensions
1129: non-ANSI source characters in character and string literals
1130: *unrecognized control lines
1131: *unreferenced variables and static functions
1132: *more than 32kb in an array, struct, or union
1133: more than 509 characters in a string literal
1134: more than 127 case labels in a switch statement
1135: more than 31 formal parameters
1136: more than 15 levels of nested statements
1137: more than 127 identifiers declared in a block
1138: more than 511 external identifiers
1139: more than 127 fields in a struct or union
1140: more than 127 enumeration constants in an enumeration
1141:
1142: * marks those warnings that were given *without* -A previously.
1143:
1144: Formerly, -N cleared the #include search list, disabled language
1145: extensions, and arranged for duplicate global definitions in separately
1146: compiled files to cause loader errors. It now fools only with
1147: #includes; -A -A handles the others.
1148:
1149: [I can't reach the Sun-4's now. I'll update them later.]
1150:
1151: 5/4/91:
1152:
1153: -A -A now warns about using function designators in conditionals
1154: (eg, f(){int g();if(g);})
1155:
1156: Null characters in string and character literals are now accepted.
1157:
1158: Returning the address of a parameter or a local is now an error.
1159:
1160: The compiler proper no longer ignores \<newline> in string and
1161: character literals; `line splicing' must be done by preprocessors.
1162:
1163: Erroneously complained about some valid uses of function types,
1164: (eg, int f(int (int));)
1165:
1166: Using -b to append data to a prof.out compressed by bprint -c corrupted
1167: the execution counts.
1168:
1169: Erroneously complained about a typedef name used as a label at the
1170: first token in a compound statement (eg, typedef int T; f(){T:;})
1171:
1172: Erroneously permitted array fields of structs returned by functions to
1173: be passed as arguments (eg, struct foo {char s[63];} g(); f(g().s);)
1174:
1175: Subsequent definition of unknown enums caused an assertion failure,
1176: (eg, enum foo x; enum foo {a,b,c};)
1177:
1178: Failed to diagnose empty extern/static declarations of enums or
1179: structs (eg, extern enum foo {a};)
1180:
1181: Failed to ignore leading white space that spanned input buffers.
1182:
1183: Failed to promote case labels to the promoted type of the switch
1184: expression (eg, case (char)16.2:).
1185:
1186: In structs/unions without named bit fields, padding specified by
1187: unnamed bit fields was erroneously omitted in initializations. (eg,
1188: struct { char a; int:0; char b; } x = {'a','b'} misinitialized x.b.)
1189:
1190: Erroneously complained about extern declarations of sizeless arrays
1191: with block scope that referred to statics with file scope (eg, static
1192: int a[3]; f() { extern int a[]; ...})
1193:
1194: On the MIPS, better code for conversions from char/short to unsigned.
1195:
1196: On the MIPS, failed to detect overflow in (1<<31)[*/](-1) and crashed.
1197:
1198: On the SPARC, taking the address of an argument in leaf functions
1199: caused junk to be stored in the actual argument, eg, f(x){int *p = &x;}
1200:
1201: On the SPARC, stdlib.h erroneously defined toupper() and tolower() as
1202: macros; they're library functions.
1203:
1204: On the SPARC, RAND_MAX in stdlib.h was erroneously defined as 32767; it
1205: should be 2^31-1.
1206:
1207: On Suns, -target sun[34] is now recognized (and ignored).
1208:
1209: 8/2/91:
1210:
1211: Now permits -o to name the output file for -c and -S when
1212: invoked with exactly one source file.
1213:
1214: -t now prints char*'s as a address in hex and as a string.
1215:
1216: Constants cast to const/volatile were not treated as constant operands
1217: (eg, (const int)23 + 80 was not accepted as a constant expression).
1218:
1219: Some generated labels were not of the form Ln, which clogged symbol
1220: tables and confused profilers on some targets.
1221:
1222: The declaration `struct S;' failed to introduce a new type when a
1223: struct S was defined in an enclosing scope.
1224:
1225: When built with a compiler where `char' is unsigned, lcc failed to
1226: sign-extend `signed char' in constant expressions (eg, (signed char)-4
1227: became 252 instead of -4.)
1228:
1229: In structs/unions with bit fields, intra-word padding specified by
1230: unnamed bit fields was erroneously omitted in initializations (eg,
1231: struct {int :8;int a:3;}x={1} initialized x.a to 0 instead of 1).
1232:
1233: Crashed when case labels were given outside of switches.
1234:
1235: Sometimes computed incorrect values for relationals over two floating
1236: point constants.
1237:
1238: Erroneously permitted old-style parameter lists in function
1239: delarators. Some caused the compiler to crash, eg,
1240: int(*f)(a);g(){(*f)(0);}
1241:
1242: On the 68020, SPARC and VAX, -g emitted bad symbol table data for
1243: typedefs of const/volatile types.
1244:
1245: On the 68020, lcc suffered an assertion failure occasionally when it
1246: ran out of floating point registers.
1247:
1248: On the MIPS, now places constants in the read-only data segment instead
1249: of the text segment.
1250:
1251: On the SPARC, now recognizes as leaves functions that return floats or
1252: doubles but use no other floating-point registers, eg, float f(float
1253: x){return x;}
1254:
1255: On the SPARC, emitted bad code for leaf functions that took the address
1256: of a float/double argument.
1257:
1258: 8/13/91:
1259:
1260: Generated bad code for expressions with multiple references to fields
1261: of structs returned by functions (eg, f().a + g().b).
1262:
1263: Erroneously diagnosed as errors valid function definitions in which
1264: old-style parameters appeared in a declarator nested in parens (eg, int
1265: (*f(a,b))() int a,b; {}).
1266:
1267: 9/30/91:
1268:
1269: "struct S;" is now never an error, even if repeated.
1270:
1271: Fixed assertion failure when a comparison was used as a value in an
1272: expression that had a type error, like char*f(){int x;x=(x==1|f());}
1273:
1274: -b profiled "while (e) S" wrong when e is constant and under -g.
1275:
1276: -P failed to include "const" and "volatile" in its output declarations.
1277:
1278: On the 68020, the driver now supplies -Dmc68000O in addition to
1279: -Dmc68000.
1280:
1281: On the MIPS, -Wf-Gn specifies the maximum size for data placed in the
1282: "short data" segment ala cc(1); the default is 8, and 0 causes all data
1283: to be placed in the "data" segment. -Wa-Gn and -Wl-Gn pass similar
1284: flags to the assembler and linker, resp.
1285:
1286: On the SPARC, structures are now passed using Sun's conventions.
1287: Discard old .o files.
1288:
1289: If bprint sees that it's about to print a count INSIDE a number or
1290: identifier, it moves the count to just BEFORE the token. This
1291: heuristic applies only when macros or comments have confused bprint
1292: about the source code. (Thanks to Doug.)
1293:
1294: 10/7/91:
1295:
1296: On the SPARC, generated bad code for functions that are declared to
1297: return a struct but then fail to, like struct node f(){}.
1298:
1299: -P failed to include "const" and "volatile" in its output declarations.
1300:
1301: 1/4/92:
1302:
1303: lcc formerly stopped compiling after 20 errors; -Wf-e<n> now sets the
1304: error limit to <n>.
1305:
1306: The driver now checks for non-existent or unreadable .c or .i files and
1307: does not try to compile them.
1308:
1309: Crashed when an unknown typedef declared parameters in old-style
1310: function definitions, like "f(x) T x; {...}".
1311:
1312: Crashed when both -b and -n were specified.
1313:
1314: Failed to fold case labels that exceeded the limits of the type of the
1315: associated switch expression, like "int i; switch (i) {case0xeb010000:;}".
1316: This yielded bad duplicate case label diagnostics or incorrect code.
1317:
1318: Failed to recognize unsigned comparisons with zero as constant expressions.
1319: For example, "int i=5U>=0;" is now accepted.
1320:
1321: Failed to diagnose structure initializations with excess initializers,
1322: like "struct { int a,b; } x = { 1,2,3 };".
1323:
1324: Failed to count as references occurrences of identifiers in asm, so
1325: locals with no other references were erroneously deleted.
1326:
1327: Failed to make each parameter visible at the end of its declarator,
1328: which led to erroneous diagnostics for valid prototypes. For example,
1329: "f(int a, int b[sizeof a])" is now accepted.
1330:
1331: errno.h defined more error codes than are specified by the standard; it
1332: now defines only EDOM and ERANGE.
1333:
1334: assert.h, errno.h, locale.h, stdio.h, stdlib.h, string.h, and time.h
1335: included other headers and thus defined symbols beyond those permitted
1336: by the standard. They now define only the specified symbols, except for
1337: assert.h, which defines write, sprintf, and abort on some targets.
1338:
1339: On the MIPS and SPARC, better code for structure assignment and
1340: structure arguments.
1341:
1342: On the SPARC with -g, gdb and dbx printed incorrect values for parameters
1343: that were changed because of extraneous symbol table entries.
1344:
1345: On the MIPS with -g, gdb and dbx mislocated some source files when
1346: #include directions contained function and variable definitions.
1347:
1348: 2/11/92:
1349:
1350: Erroneously identified T* as the offending type when sizeof(T) is
1351: unknown or 0 (eg, "void *a,*b;a-b;"). Now correctly fingers T
1352: instead.
1353:
1354: Erroneously allowed arithmetic on `void *const' pointers (eg,
1355: "void *const p; int i; p+i;").
1356:
1357: Crashed when comparing prototypes with an argument that is a pointer to
1358: a function with a prototype, like
1359: int (*g)(char *,...); void f(int (*)(char *,...)) {} ... f(g);
1360:
1361: On Suns and BSD VAXes, ctype.h defined static functions _tolower and
1362: _toupper, which prompted `unreferenced' warnings if they weren't used.
1363:
1364: 4/13/92:
1365:
1366: Crashed when comparing prototypes that differ in only top-level
1367: qualifiers (eg, int f(int); int f(const int);)
1368:
1369: The driver no longer ignores extra characters after flags. As always,
1370: unrecognized flags are sent to the linker; eg, -cg was formerly treated
1371: like -c, but it's now sent to ld, which will complain.
1372:
1373: Now recognizes the identity u/1U == u for unsigned u.
1374:
1375: -A -A now correctly complains about switch statements with more than
1376: 257 cases instead of 127.
1377:
1378: On the SPARC, generated incorrect code for calls with a floating-point
1379: argument followed anywhere by an argument involving an integer or
1380: unsigned division, multiplication, or modulus (eg, int i,j; f() {
1381: f(1.0,i/j); })
1382:
1383: On the SPARC, generated invalid assembly code for assignments of
1384: structures over 4095 bytes.
1385:
1386: 12/16/92:
1387:
1388: changes/improvements:
1389: with -d0, warns about switch statements with huge jump tables, eg,
1390: switch (i) { case INT_MAX: ...; case INT_MIN: ... }
1391:
1392: now accepts identifiers up to 4096 characters long, although the ANSI
1393: standard requires that only the first 31 be examined.
1394:
1395: returning a pointer to a local variable now elicits a warning
1396: instead of an error.
1397:
1398: now warns about eliding references to incomplete types, eg,
1399: void *x; *x;
1400:
1401: with -A -A, now warns when declaring arrays incomplete types,
1402: eg, extern struct s x[];
1403:
1404: offsetof is less machine-specific but equivalent to the old one.
1405:
1406: On the MIPS, uses fewer of the assembler's pseudo-ops. The assembly
1407: code looks longer, but it should generate about the same machine code.
1408:
1409: On the MIPS, a few stack frames may be larger, to enforce stricter
1410: alignment for the R4000.
1411:
1412: errors fixed:
1413: with -n, failed to recognize expressions that dereferenced constant
1414: pointers as valid constant expressions, eg,
1415: struct foo { int a, b; };
1416: unsigned off = (char *)&((struct foo *)0)->b - (char *)0;
1417:
1418: generated bad code for some inputs for which the compiler needed
1419: numerous temporaries.
1420:
1421: failed to accept function definitions with multiple top-level
1422: prototypes, eg, int (*g(int c))(char) { c = 0; }
1423:
1424: failed to warn about overflow in enumeration values, eg,
1425: enum { X = ~0U };
1426:
1427: ignored trailing characters of long file names in some "#" lines, which
1428: caused erroneous diagnostics. now accepts names upto 512 characters, and
1429: warns about longer ones.
1430:
1431: erroneously made enumerations visible before their optional values, eg,
1432: enum{A=2}; f() { enum {A=A} i = A; } assigned 0 to i instead of 2.
1433:
1434: exited with a code of 0 instead of 1 when there was exactly one
1435: "found X expecting Y" diagnostic.
1436:
1437: failed to detect overflow in hex character constants, eg, '\x100000000'.
1438:
1439: failed to accept assignment of 0U to pointers.
1440:
1441: failed to associate tags defined in a parameter list with defining
1442: declarations inside a function, eg,
1443: f(struct S *s) { struct S { int a; }; s->a = 1; }
1444:
1445: failed to reject function definitions with storage class typedef.
1446:
1447: failed to accept a trailing `,' in initializers for scalars
1448: and unions, eg, int x = {1,};
1449:
1450: erroneously defined wchar_t as `char' instead of `unsigned char';
1451: L'\xff' must have the integer value 0xff.
1452:
1453: failed to reject type `array of void'.
1454:
1455: failed to reject comparing a pointer to an object type
1456: with a pointer to an incomplete type, eg,
1457: extern int (*a)[],(*b)[10]; ...; if (a == b) ...
1458:
1459: failed to warn about comparing a function pointer with a void*,
1460: eg, int (*fp)(); ... if (fp == (void *)0) ...
1461:
1462: failed to reject (..., 0) as a null pointer constant.
1463:
1464: failed to accept redeclaration of functions with differently
1465: qualified parameters, eg, f(int); f(const int);
1466:
1467: on the SPARC, under -b, exit() failed to close all open files.
1468:
1469: on the MIPS, generated bad code for calls to variadic functions
1470: with a leading floating point arg (eg, f(double x,...)) when
1471: the 2nd actual argument was a float or a double.
1472:
1473: 12/17/92:
1474:
1475: fix bug introduced yesterday into stddef.h's offsetof macro.
1476:
1477: 12/18/92:
1478:
1479: On the MIPS, fixes code generated since 12/16 that crashes most
1480: procedures with >32kb of locals.
1481:
1482: 2/18/93:
1483:
1484: On the MIPS, silently generated incorrect code for some programs that
1485: mix (single-precision) float register variables with int conditional
1486: exprs or int register variables with float conditional exprs.
1487:
1488: next:
1489:
1490: all enums are now ints; old .o files that use enums are incompatible
1491: with new ones.
1492:
1493: asm("...") is gone.
1494:
1495: Erroneously treated the constant -0.0 as 0.0, which IEEE forbids.
1496:
1497: Erroneously diagnosed tentative unsized array definitions, eg,
1498: a lone "int x[]" is treated as "int x[]={0}"
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.