1 This file lists recent changes to the GNU Fortran compiler. Copyright
2 (C) 1995, 1996 Free Software Foundation, Inc. You may copy,
3 distribute, and modify it freely as long as you preserve this copyright
4 notice and permission notice.
9 Changes made to recent versions of GNU Fortran are listed below,
10 with the most recent version first.
12 The changes are generally listed with code-generation bugs first,
13 followed by compiler crashes involving valid code, new features, fixes
14 to existing features, new diagnostics, internal improvements, and
15 miscellany. This order is not strict--for example, some items involve
16 a combination of these elements.
21 * Fix a code-generation bug introduced by 0.5.20 caused by loop
22 unrolling (by specifying `-funroll-loops' or similar). This bug
23 afflicted all code compiled by version 2.7.2.2.f.2 of `gcc' (C,
24 C++, Fortran, and so on).
26 * Fix a code-generation bug manifested when combining local
27 `EQUIVALENCE' with a `DATA' statement that follows the first
28 executable statement (or is treated as an executable-context
29 statement as a result of using the `-fpedantic' option).
31 * Fix a compiler crash that occured when an integer division by a
32 constant zero is detected. Instead, when the `-W' option is
33 specified, the `gcc' back end issues a warning about such a case.
34 This bug afflicted all code compiled by version 2.7.2.2.f.2 of
35 `gcc' (C, C++, Fortran, and so on).
37 * Fix a compiler crash that occurred in some cases of procedure
38 inlining. (Such cases became more frequent in 0.5.20.)
40 * Fix a compiler crash resulting from using `DATA' or similar to
41 initialize a `COMPLEX' variable or array to zero.
43 * Fix compiler crashes involving use of `AND', `OR', or `XOR'
46 * Fix compiler bug triggered when using a `COMMON' or `EQUIVALENCE'
47 variable as the target of an `ASSIGN' or assigned-`GOTO' statement.
49 * Fix compiler crashes due to using the name of a some non-standard
50 intrinsics (such as `FTELL' or `FPUTC') as such and as the name of
51 a procedure or common block. Such dual use of a name in a program
52 is allowed by the standard.
54 * Place automatic arrays on the stack, even if `SAVE' or the
55 `-fno-automatic' option is in effect. This avoids a compiler
58 * New option `-Wno-globals' disables warnings about "suspicious" use
59 of a name both as a global name and as the implicit name of an
60 intrinsic, and warnings about disagreements over the number or
61 natures of arguments passed to global procedures, or the natures
62 of the procedures themselves.
64 The default is to issue such warnings, which are new as of this
67 * New option `-fno-globals' disables diagnostics about potentially
68 fatal disagreements analysis problems, such as disagreements over
69 the number or natures of arguments passed to global procedures, or
70 the natures of those procedures themselves.
72 The default is to issue such diagnostics and flag the compilation
73 as unsuccessful. With this option, the diagnostics are issued as
74 warnings, or, if `-Wno-globals' is specified, are not issued at
77 This option also disables inlining of global procedures, to avoid
78 compiler crashes resulting from coding errors that these
79 diagnostics normally would identify.
81 * Diagnose cases where a reference to a procedure disagrees with the
82 type of that procedure, or where disagreements about the number or
83 nature of arguments exist. This avoids a compiler crash.
85 * Improve performance of the `gcc' back end so certain complicated
86 expressions involving `COMPLEX' arithmetic (especially
87 multiplication) don't appear to take forever to compile.
89 * Fix a couple of profiling-related bugs in `gcc' back end.
91 * Integrate GNU Ada's (GNAT's) changes to the back end, which
92 consist almost entirely of bug fixes.
94 * Include some other `gcc' fixes that seem useful in `g77''s version
95 of `gcc'. (See `gcc/ChangeLog' for details--compare it to that
96 file in the vanilla `gcc-2.7.2.2.tar.gz' distribution.)
98 * Fix `libU77' routines that accept file and other names to strip
99 trailing blanks from them, for consistency with other
100 implementations. Blanks may be forcibly appended to such names by
101 appending a single null character (`CHAR(0)') to the significant
104 * Fix `CHMOD' intrinsic to work with file names that have embedded
105 blanks, commas, and so on.
107 * Fix `SIGNAL' intrinsic so it accepts an optional third `Status'
110 * Fix `IDATE()' intrinsic subroutine (VXT form) so it accepts
111 arguments in the correct order. Documentation fixed accordingly,
112 and for `GMTIME()' and `LTIME()' as well.
114 * Make many changes to `libU77' intrinsics to support existing code
117 Such changes include allowing both subroutine and function forms
118 of many routines, changing `MCLOCK()' and `TIME()' to return
119 `INTEGER(KIND=1)' values, introducing `MCLOCK8()' and `TIME8()' to
120 return `INTEGER(KIND=2)' values, and placing functions that are
121 intended to perform side effects in a new intrinsic group,
124 * Improve `libU77' so it is more portable.
126 * Add options `-fbadu77-intrinsics-delete',
127 `-fbadu77-intrinsics-hide', and so on.
129 * Fix crashes involving diagnosed or invalid code.
131 * `g77' and `gcc' now do a somewhat better job detecting and
132 diagnosing arrays that are too large to handle before these cause
133 diagnostics during the assembler or linker phase, a compiler
134 crash, or generation of incorrect code.
136 * Improve alias analysis code to properly handle output registers
137 (such as the `%o' registers on the SPARC).
139 * Add support for `restrict' keyword in `gcc' front end.
141 * Modify `make' rules and related code so that generation of Info
142 documentation doesn't require compilation using `gcc'.
144 * Add `INT2' and `INT8' intrinsics.
146 * Add `CPU_TIME' intrinsic.
148 * Add `ALARM' intrinsic.
150 * `CTIME' intrinsic now accepts any `INTEGER' argument, not just
153 * Warn when explicit type declaration disagrees with the type of an
154 intrinsic invocation.
156 * Support `*f771' entry in `gcc' `specs' file.
158 * Fix typo in `make' rule `g77-cross', used only for cross-compiling.
160 * Fix `libf2c' build procedure to re-archive library if previous
161 attempt to archive was interrupted.
163 * Fix `gcc' to more easily support configuring on Pentium Pro (686)
166 * Change `gcc' to unroll loops only during the last invocation (of
167 as many as two invocations) of loop optimization.
169 * Improve handling of `-fno-f2c' so that code that attempts to pass
170 an intrinsic as an actual argument, such as `CALL FOO(ABS)', is
171 rejected due to the fact that the run-time-library routine is,
172 effectively, compiled with `-ff2c' in effect.
174 * Fix `g77' driver to recognize `-fsyntax-only' as an option that
175 inhibits linking, just like `-c' or `-S', and to recognize and
176 properly handle the `-nostdlib', `-M', `-MM', `-nodefaultlibs',
177 and `-Xlinker' options.
179 * Upgrade to `libf2c' as of 1997-08-06.
181 * Modify `libf2c' to consistently and clearly diagnose recursive I/O
184 * `g77' driver now prints version information (such as produced by
185 `g77 -v') to `stderr' instead of `stdout'.
187 * The `.r' suffix now designates a Ratfor source file, to be
188 preprocessed via the `ratfor' command, available separately.
190 * Fix some aspects of how `gcc' determines what kind of system is
191 being configured and what kinds are supported. For example, GNU
192 Linux/Alpha ELF systems now are directly supported.
194 * Improve diagnostics.
196 * Improve documentation and indexing.
198 * Include all pertinent files for `libf2c' that come from
199 `netlib.bell-labs.com'; give any such files that aren't quite
200 accurate in `g77''s version of `libf2c' the suffix `.netlib'.
202 * Reserve `INTEGER(KIND=0)' for future use.
207 * The `-fno-typeless-boz' option is now the default.
209 This option specifies that non-decimal-radix constants using the
210 prefixed-radix form (such as `Z'1234'') are to be interpreted as
211 `INTEGER' constants. Specify `-ftypeless-boz' to cause such
212 constants to be interpreted as typeless.
214 (Version 0.5.19 introduced `-fno-typeless-boz' and its inverse.)
216 * Options `-ff90-intrinsics-enable' and `-fvxt-intrinsics-enable'
217 now are the defaults.
219 Some programs might use names that clash with intrinsic names
220 defined (and now enabled) by these options or by the new `libU77'
221 intrinsics. Users of such programs might need to compile them
222 differently (using, for example, `-ff90-intrinsics-disable') or,
223 better yet, insert appropriate `EXTERNAL' statements specifying
224 that these names are not intended to be names of intrinsics.
226 * The `ALWAYS_FLUSH' macro is no longer defined when building
227 `libf2c', which should result in improved I/O performance,
230 *Note:* If you have code that depends on the behavior of `libf2c'
231 when built with `ALWAYS_FLUSH' defined, you will have to modify
232 `libf2c' accordingly before building it from this and future
235 * Dave Love's implementation of `libU77' has been added to the
236 version of `libf2c' distributed with and built as part of `g77'.
237 `g77' now knows about the routines in this library as intrinsics.
239 * New option `-fvxt' specifies that the source file is written in
240 VXT Fortran, instead of GNU Fortran.
242 * The `-fvxt-not-f90' option has been deleted, along with its
243 inverse, `-ff90-not-vxt'.
245 If you used one of these deleted options, you should re-read the
246 pertinent documentation to determine which options, if any, are
247 appropriate for compiling your code with this version of `g77'.
249 * The `-fugly' option now issues a warning, as it likely will be
250 removed in a future version.
252 (Enabling all the `-fugly-*' options is unlikely to be feasible,
253 or sensible, in the future, so users should learn to specify only
254 those `-fugly-*' options they really need for a particular source
257 * The `-fugly-assumed' option, introduced in version 0.5.19, has
258 been changed to better accommodate old and new code.
260 * Make a number of fixes to the `g77' front end and the `gcc' back
261 end to better support Alpha (AXP) machines. This includes
262 providing at least one bug-fix to the `gcc' back end for Alphas.
264 * Related to supporting Alpha (AXP) machines, the `LOC()' intrinsic
265 and `%LOC()' construct now return values of integer type that is
266 the same width (holds the same number of bits) as the pointer type
269 On most machines, this won't make a difference, whereas on Alphas,
270 the type these constructs return is `INTEGER*8' instead of the
271 more common `INTEGER*4'.
273 * Emulate `COMPLEX' arithmetic in the `g77' front end, to avoid bugs
274 in `complex' support in the `gcc' back end. New option
275 `-fno-emulate-complex' causes `g77' to revert the 0.5.19 behavior.
277 * Fix bug whereby `REAL A(1)', for example, caused a compiler crash
278 if `-fugly-assumed' was in effect and A was a local (automatic)
279 array. That case is no longer affected by the new handling of
282 * Fix `g77' command driver so that `g77 -o foo.f' no longer deletes
283 `foo.f' before issuing other diagnostics, and so the `-x' option
286 * Enable inlining of subroutines and functions by the `gcc' back end.
287 This works as it does for `gcc' itself--program units may be
288 inlined for invocations that follow them in the same program unit,
289 as long as the appropriate compile-time options are specified.
291 * Dummy arguments are no longer assumed to potentially alias
292 (overlap) other dummy arguments or `COMMON' areas when any of
293 these are defined (assigned to) by Fortran code.
295 This can result in faster and/or smaller programs when compiling
296 with optimization enabled, though on some systems this effect is
297 observed only when `-fforce-addr' also is specified.
299 New options `-falias-check', `-fargument-alias',
300 `-fargument-noalias', and `-fno-argument-noalias-global' control
301 the way `g77' handles potential aliasing.
303 * The `CONJG()' and `DCONJG()' intrinsics now are compiled in-line.
305 * The bug-fix for 0.5.19.1 has been re-done. The `g77' compiler has
306 been changed back to assume `libf2c' has no aliasing problems in
307 its implementations of the `COMPLEX' (and `DOUBLE COMPLEX')
308 intrinsics. The `libf2c' has been changed to have no such
311 As a result, 0.5.20 is expected to offer improved performance over
312 0.5.19.1, perhaps as good as 0.5.19 in most or all cases, due to
315 *Note:* This change requires version 0.5.20 of `libf2c', at least,
316 when linking code produced by any versions of `g77' other than
317 0.5.19.1. Use `g77 -v' to determine the version numbers of the
318 `libF77', `libI77', and `libU77' components of the `libf2c'
319 library. (If these version numbers are not printed--in
320 particular, if the linker complains about unresolved references to
321 names like `g77__fvers__'--that strongly suggests your
322 installation has an obsolete version of `libf2c'.)
324 * New option `-fugly-assign' specifies that the same memory
325 locations are to be used to hold the values assigned by both
326 statements `I = 3' and `ASSIGN 10 TO I', for example. (Normally,
327 `g77' uses a separate memory location to hold assigned statement
330 * `FORMAT' and `ENTRY' statements now are allowed to precede
331 `IMPLICIT NONE' statements.
333 * Produce diagnostic for unsupported `SELECT CASE' on `CHARACTER'
334 type, instead of crashing, at compile time.
336 * Fix crashes involving diagnosed or invalid code.
338 * Change approach to building `libf2c' archive (`libf2c.a') so that
339 members are added to it only when truly necessary, so the user
340 that installs an already-built `g77' doesn't need to have write
341 access to the build tree (whereas the user doing the build might
342 not have access to install new software on the system).
344 * Support `gcc' version 2.7.2.2 (modified by `g77' into version
345 2.7.2.2.f.2), and remove support for prior versions of `gcc'.
347 * Upgrade to `libf2c' as of 1997-02-08, and fix up some of the build
350 * Improve general build procedures for `g77', fixing minor bugs
351 (such as deletion of any file named `f771' in the parent directory
354 * Enable full support of `INTEGER*8' available in `libf2c' and
355 `f2c.h' so that `f2c' users may make full use of its features via
356 the `g77' version of `f2c.h' and the `INTEGER*8' support routines
357 in the `g77' version of `libf2c'.
359 * Improve `g77' driver and `libf2c' so that `g77 -v' yields version
360 information on the library.
362 * The `SNGL' and `FLOAT' intrinsics now are specific intrinsics,
363 instead of synonyms for the generic intrinsic `REAL'.
365 * New intrinsics have been added. These are `REALPART', `IMAGPART',
366 `COMPLEX', `LONG', and `SHORT'.
368 * A new group of intrinsics, `gnu', has been added to contain the
369 new `REALPART', `IMAGPART', and `COMPLEX' intrinsics. An old
370 group, `dcp', has been removed.
372 * Complain about industry-wide ambiguous references `REAL(EXPR)' and
373 `AIMAG(EXPR)', where EXPR is `DOUBLE COMPLEX' (or any complex type
374 other than `COMPLEX'), unless `-ff90' option specifies Fortran 90
375 interpretation or new `-fugly-complex' option, in conjunction with
376 `-fnot-f90', specifies `f2c' interpretation.
378 * Make improvements to diagnostics.
380 * Speed up compiler a bit.
382 * Improvements to documentation and indexing, including a new
383 chapter containing information on one, later more, diagnostics
384 that users are directed to pull up automatically via a message in
385 the diagnostic itself.
387 (Hence the menu item `M' for the node `Diagnostics' in the
388 top-level menu of the Info documentation.)
393 * Code-generation bugs afflicting operations on complex data have
396 These bugs occurred when assigning the result of an operation to a
397 complex variable (or array element) that also served as an input
400 The operations affected by this bug were: `CONJG()', `DCONJG()',
401 `CCOS()', `CDCOS()', `CLOG()', `CDLOG()', `CSIN()', `CDSIN()',
402 `CSQRT()', `CDSQRT()', complex division, and raising a `DOUBLE
403 COMPLEX' operand to an `INTEGER' power. (The related generic and
404 `Z'-prefixed intrinsics, such as `ZSIN()', also were affected.)
406 For example, `C = CSQRT(C)', `Z = Z/C', and `Z = Z**I' (where `C'
407 is `COMPLEX' and `Z' is `DOUBLE COMPLEX') have been fixed.
412 * Fix `FORMAT' statement parsing so negative values for specifiers
413 such as `P' (e.g. `FORMAT(-1PF8.1)') are correctly processed as
416 * Fix `SIGNAL' intrinsic so it once again accepts a procedure as its
419 * A temporary kludge option provides bare-bones information on
420 `COMMON' and `EQUIVALENCE' members at debug time.
422 * New `-fonetrip' option specifies FORTRAN-66-style one-trip `DO'
425 * New `-fno-silent' option causes names of program units to be
426 printed as they are compiled, in a fashion similar to UNIX `f77'
429 * New `-fugly-assumed' option specifies that arrays dimensioned via
430 `DIMENSION X(1)', for example, are to be treated as assumed-size.
432 * New `-fno-typeless-boz' option specifies that non-decimal-radix
433 constants using the prefixed-radix form (such as `Z'1234'') are to
434 be interpreted as `INTEGER' constants.
436 * New `-ff66' option is a "shorthand" option that specifies
437 behaviors considered appropriate for FORTRAN 66 programs.
439 * New `-ff77' option is a "shorthand" option that specifies
440 behaviors considered appropriate for UNIX `f77' programs.
442 * New `-fugly-comma' and `-fugly-logint' options provided to perform
443 some of what `-fugly' used to do. `-fugly' and `-fno-ugly' are
444 now "shorthand" options, in that they do nothing more than enable
445 (or disable) other `-fugly-*' options.
447 * Fix parsing of assignment statements involving targets that are
448 substrings of elements of `CHARACTER' arrays having names such as
449 `READ', `WRITE', `GOTO', and `REALFUNCTIONFOO'.
451 * Fix crashes involving diagnosed code.
453 * Fix handling of local `EQUIVALENCE' areas so certain cases of
454 valid Fortran programs are not misdiagnosed as improperly
455 extending the area backwards.
457 * Support `gcc' version 2.7.2.1.
459 * Upgrade to `libf2c' as of 1996-09-26, and fix up some of the build
462 * Change code generation for list-directed I/O so it allows for new
463 versions of `libf2c' that might return non-zero status codes for
464 some operations previously assumed to always return zero.
466 This change not only affects how `IOSTAT=' variables are set by
467 list-directed I/O, it also affects whether `END=' and `ERR='
468 labels are reached by these operations.
470 * Add intrinsic support for new `FTELL' and `FSEEK' procedures in
473 * Modify `fseek_()' in `libf2c' to be more portable (though, in
474 practice, there might be no systems where this matters) and to
475 catch invalid `whence' arguments.
477 * Some useless warnings from the `-Wunused' option have been
480 * Fix a problem building the `f771' executable on AIX systems by
481 linking with the `-bbigtoc' option.
483 * Abort configuration if `gcc' has not been patched using the patch
484 file provided in the `gcc/f/gbe/' subdirectory.
486 * Add options `--help' and `--version' to the `g77' command, to
487 conform to GNU coding guidelines. Also add printing of `g77'
488 version number when the `--verbose' (`-v') option is used.
490 * Change internally generated name for local `EQUIVALENCE' areas to
491 one based on the alphabetically sorted first name in the list of
492 names for entities placed at the beginning of the areas.
494 * Improvements to documentation and indexing.
499 * Add some rudimentary support for `INTEGER*1', `INTEGER*2',
500 `INTEGER*8', and their `LOGICAL' equivalents. (This support works
501 on most, maybe all, `gcc' targets.)
503 Thanks to Scott Snyder (<snyder@d0sgif.fnal.gov>) for providing
506 Among the missing elements from the support for these features are
507 full intrinsic support and constants.
509 * Add some rudimentary support for the `BYTE' and `WORD'
510 type-declaration statements. `BYTE' corresponds to `INTEGER*1',
511 while `WORD' corresponds to `INTEGER*2'.
513 Thanks to Scott Snyder (<snyder@d0sgif.fnal.gov>) for providing
516 * The compiler code handling intrinsics has been largely rewritten
517 to accommodate the new types. No new intrinsics or arguments for
518 existing intrinsics have been added, so there is, at this point,
519 no intrinsic to convert to `INTEGER*8', for example.
521 * Support automatic arrays in procedures.
523 * Reduce space/time requirements for handling large *sparsely*
524 initialized aggregate arrays. This improvement applies to only a
525 subset of the general problem to be addressed in 0.6.
527 * Treat initial values of zero as if they weren't specified (in DATA
528 and type-declaration statements). The initial values will be set
529 to zero anyway, but the amount of compile time processing them
530 will be reduced, in some cases significantly (though, again, this
531 is only a subset of the general problem to be addressed in 0.6).
533 A new option, `-fzeros', is introduced to enable the traditional
534 treatment of zeros as any other value.
536 * With `-ff90' in force, `g77' incorrectly interpreted `REAL(Z)' as
537 returning a `REAL' result, instead of as a `DOUBLE PRECISION'
538 result. (Here, `Z' is `DOUBLE COMPLEX'.)
540 With `-fno-f90' in force, the interpretation remains unchanged,
541 since this appears to be how at least some F77 code using the
542 `DOUBLE COMPLEX' extension expected it to work.
544 Essentially, `REAL(Z)' in F90 is the same as `DBLE(Z)', while in
545 extended F77, it appears to be the same as `REAL(REAL(Z))'.
547 * An expression involving exponentiation, where both operands were
548 type `INTEGER' and the right-hand operand was negative, was
549 erroneously evaluated.
551 * Fix bugs involving `DATA' implied-`DO' constructs (these involved
552 an errant diagnostic and a crash, both on good code, one involving
553 subsequent statement-function definition).
555 * Close `INCLUDE' files after processing them, so compiling source
556 files with lots of `INCLUDE' statements does not result in being
557 unable to open `INCLUDE' files after all the available file
558 descriptors are used up.
560 * Speed up compiling, especially of larger programs, and perhaps
561 slightly reduce memory utilization while compiling (this is *not*
562 the improvement planned for 0.6 involving large aggregate
563 areas)--these improvements result from simply turning off some
564 low-level code to do self-checking that hasn't been triggered in a
567 * Introduce three new options that implement optimizations in the
568 `gcc' back end (GBE). These options are `-fmove-all-movables',
569 `-freduce-all-givs', and `-frerun-loop-opt', which are enabled, by
570 default, for Fortran compilations. These optimizations are
571 intended to help toon Fortran programs.
573 * Patch the GBE to do a better job optimizing certain kinds of
574 references to array elements.
576 * Due to patches to the GBE, the version number of `gcc' also is
577 patched to make it easier to manage installations, especially
578 useful if it turns out a `g77' change to the GBE has a bug.
580 The `g77'-modified version number is the `gcc' version number with
581 the string `.f.N' appended, where `f' identifies the version as
582 enhanced for Fortran, and N is `1' for the first Fortran patch for
583 that version of `gcc', `2' for the second, and so on.
585 So, this introduces version 2.7.2.f.1 of `gcc'.
587 * Make several improvements and fixes to diagnostics, including the
588 removal of two that were inappropriate or inadequate.
590 * Warning about two successive arithmetic operators, produced by
591 `-Wsurprising', now produced *only* when both operators are,
592 indeed, arithmetic (not relational/boolean).
594 * `-Wsurprising' now warns about the remaining cases of using
595 non-integral variables for implied-`DO' loops, instead of these
596 being rejected unless `-fpedantic' or `-fugly' specified.
598 * Allow `SAVE' of a local variable or array, even after it has been
599 given an initial value via `DATA', for example.
601 * Introduce an Info version of `g77' documentation, which supercedes
602 `gcc/f/CREDITS', `gcc/f/DOC', and `gcc/f/PROJECTS'. These files
603 will be removed in a future release. The files `gcc/f/BUGS',
604 `gcc/f/INSTALL', and `gcc/f/NEWS' now are automatically built from
605 the texinfo source when distributions are made.
607 This effort was inspired by a first pass at translating
608 `g77-0.5.16/f/DOC' that was contributed to Craig by David Ronis
609 (<ronis@onsager.chem.mcgill.ca>).
611 * New `-fno-second-underscore' option to specify that, when
612 `-funderscoring' is in effect, a second underscore is not to be
613 appended to Fortran names already containing an underscore.
615 * Change the way iterative `DO' loops work to follow the F90
616 standard. In particular, calculation of the iteration count is
617 still done by converting the start, end, and increment parameters
618 to the type of the `DO' variable, but the result of the
619 calculation is always converted to the default `INTEGER' type.
621 (This should have no effect on existing code compiled by `g77',
622 but code written to assume that use of a *wider* type for the `DO'
623 variable will result in an iteration count being fully calculated
624 using that wider type (wider than default `INTEGER') must be
627 * Support `gcc' version 2.7.2.
629 * Upgrade to `libf2c' as of 1996-03-23, and fix up some of the build
632 Note that the email addresses related to `f2c' have changed--the
633 distribution site now is named `netlib.bell-labs.com', and the
634 maintainer's new address is <dmg@bell-labs.com>.
639 * *Fix serious bug* in `g77 -v' command that can cause removal of a
640 system's `/dev/null' special file if run by user `root'.
642 *All users* of version 0.5.16 should ensure that they have not
643 removed `/dev/null' or replaced it with an ordinary file (e.g. by
644 comparing the output of `ls -l /dev/null' with `ls -l /dev/zero'.
645 If the output isn't basically the same, contact your system
646 administrator about restoring `/dev/null' to its proper status).
648 This bug is particularly insidious because removing `/dev/null' as
649 a special file can go undetected for quite a while, aside from
650 various applications and programs exhibiting sudden, strange
653 I sincerely apologize for not realizing the implications of the
654 fact that when `g77 -v' runs the `ld' command with `-o /dev/null'
655 that `ld' tries to *remove* the executable it is supposed to build
656 (especially if it reports unresolved references, which it should
659 * Fix crash on `CHARACTER*(*) FOO' in a main or block data program
662 * Fix crash that can occur when diagnostics given outside of any
663 program unit (such as when input file contains `@foo').
665 * Fix crashes, infinite loops (hangs), and such involving diagnosed
668 * Fix `ASSIGN''ed variables so they can be `SAVE''d or dummy
669 arguments, and issue clearer error message in cases where target
670 of `ASSIGN' or `ASSIGN'ed `GOTO'/`FORMAT' is too small (which
671 should never happen).
673 * Make `libf2c' build procedures work on more systems again by
674 eliminating unnecessary invocations of `ld -r -x' and `mv'.
676 * Fix omission of `-funix-intrinsics-...' options in list of
677 permitted options to compiler.
679 * Fix failure to always diagnose missing type declaration for
682 * Fix compile-time performance problem (which could sometimes crash
683 the compiler, cause a hang, or whatever, due to a bug in the back
684 end) involving exponentiation with a large `INTEGER' constant for
685 the right-hand operator (e.g. `I**32767').
687 * Fix build procedures so cross-compiling `g77' (the `fini' utility
688 in particular) is properly built using the host compiler.
690 * Add new `-Wsurprising' option to warn about constructs that are
691 interpreted by the Fortran standard (and `g77') in ways that are
692 surprising to many programmers.
694 * Add `ERF()' and `ERFC()' as generic intrinsics mapping to existing
695 `ERF'/`DERF' and `ERFC'/`DERFC' specific intrinsics.
697 *Note:* You should specify `INTRINSIC ERF,ERFC' in any code where
698 you might use these as generic intrinsics, to improve likelihood
699 of diagnostics (instead of subtle run-time bugs) when using a
700 compiler that doesn't support these as intrinsics (e.g. `f2c').
702 * Remove from `-fno-pedantic' the diagnostic about `DO' with
703 non-`INTEGER' index variable; issue that under `-Wsurprising'
706 * Clarify some diagnostics that say things like "ignored" when that's
709 * Clarify diagnostic on use of `.EQ.'/`.NE.' on `LOGICAL' operands.
711 * Minor improvements to code generation for various operations on
714 * Minor improvement to code generation for some `DO' loops on some
717 * Support `gcc' version 2.7.1.
719 * Upgrade to `libf2c' as of 1995-11-15.
724 * Fix a code-generation bug involving complicated `EQUIVALENCE'
725 statements not involving `COMMON'.
727 * Fix code-generation bugs involving invoking "gratis" library
728 procedures in `libf2c' from code compiled with `-fno-f2c' by
729 making these procedures known to `g77' as intrinsics (not affected
730 by -fno-f2c). This is known to fix code invoking `ERF()',
731 `ERFC()', `DERF()', and `DERFC()'.
733 * Update `libf2c' to include netlib patches through 1995-08-16, and
734 `#define' `WANT_LEAD_0' to 1 to make `g77'-compiled code more
735 consistent with other Fortran implementations by outputting
736 leading zeros in formatted and list-directed output.
738 * Fix a code-generation bug involving adjustable dummy arrays with
739 high bounds whose primaries are changed during procedure
740 execution, and which might well improve code-generation
741 performance for such arrays compared to `f2c' plus `gcc' (but
742 apparently only when using `gcc-2.7.0' or later).
744 * Fix a code-generation bug involving invocation of `COMPLEX' and
745 `DOUBLE COMPLEX' `FUNCTION's and doing `COMPLEX' and `DOUBLE
746 COMPLEX' divides, when the result of the invocation or divide is
747 assigned directly to a variable that overlaps one or more of the
748 arguments to the invocation or divide.
750 * Fix crash by not generating new optimal code for `X**I' if `I' is
751 nonconstant and the expression is used to dimension a dummy array,
752 since the `gcc' back end does not support the necessary mechanics
753 (and the `gcc' front end rejects the equivalent construct, as it
756 * Fix crash on expressions like `COMPLEX**INTEGER'.
758 * Fix crash on expressions like `(1D0,2D0)**2', i.e. raising a
759 `DOUBLE COMPLEX' constant to an `INTEGER' constant power.
761 * Fix crashes and such involving diagnosed code.
763 * Diagnose, instead of crashing on, statement function definitions
764 having duplicate dummy argument names.
766 * Fix bug causing rejection of good code involving statement function
769 * Fix bug resulting in debugger not knowing size of local equivalence
770 area when any member of area has initial value (via `DATA', for
773 * Fix installation bug that prevented installation of `g77' driver.
774 Provide for easy selection of whether to install copy of `g77' as
775 `f77' to replace the broken code.
777 * Fix `gcc' driver (affects `g77' thereby) to not gratuitously
778 invoke the `f771' program (e.g. when `-E' is specified).
780 * Fix diagnostic to point to correct source line when it immediately
781 follows an `INCLUDE' statement.
783 * Support more compiler options in `gcc'/`g77' when compiling
784 Fortran files. These options include `-p', `-pg', `-aux-info',
785 `-P', correct setting of version-number macros for preprocessing,
786 full recognition of `-O0', and automatic insertion of
787 configuration-specific linker specs.
789 * Add new intrinsics that interface to existing routines in `libf2c':
790 `ABORT', `DERF', `DERFC', `ERF', `ERFC', `EXIT', `FLUSH',
791 `GETARG', `GETENV', `IARGC', `SIGNAL', and `SYSTEM'. Note that
792 `ABORT', `EXIT', `FLUSH', `SIGNAL', and `SYSTEM' are intrinsic
793 subroutines, not functions (since they have side effects), so to
794 get the return values from `SIGNAL' and `SYSTEM', append a final
795 argument specifying an `INTEGER' variable or array element (e.g.
796 `CALL SYSTEM('rm foo',ISTAT)').
798 * Add new intrinsic group named `unix' to contain the new intrinsics,
799 and by default enable this new group.
801 * Move `LOC()' intrinsic out of the `vxt' group to the new `unix'
804 * Improve `g77' so that `g77 -v' by itself (or with certain other
805 options, including `-B', `-b', `-i', `-nostdlib', and `-V')
806 reports lots more useful version info, and so that long-form
807 options `gcc' accepts are understood by `g77' as well (even in
808 truncated, unambiguous forms).
810 * Add new `g77' option `--driver=name' to specify driver when
811 default, `gcc', isn't appropriate.
813 * Add support for `#' directives (as output by the preprocessor) in
814 the compiler, and enable generation of those directives by the
815 preprocessor (when compiling `.F' files) so diagnostics and
816 debugging info are more useful to users of the preprocessor.
818 * Produce better diagnostics, more like `gcc', with info such as `In
819 function `foo':' and `In file included from...:'.
821 * Support `gcc''s `-fident' and `-fno-ident' options.
823 * When `-Wunused' in effect, don't warn about local variables used as
824 statement-function dummy arguments or `DATA' implied-`DO' iteration
825 variables, even though, strictly speaking, these are not uses of
826 the variables themselves.
828 * When `-W -Wunused' in effect, don't warn about unused dummy
829 arguments at all, since there's no way to turn this off for
830 individual cases (`g77' might someday start warning about
831 these)--applies to `gcc' versions 2.7.0 and later, since earlier
832 versions didn't warn about unused dummy arguments.
834 * New option `-fno-underscoring' that inhibits transformation of
835 names (by appending one or two underscores) so users may experiment
836 with implications of such an environment.
838 * Minor improvement to `gcc/f/info' module to make it easier to build
839 `g77' using the native (non-`gcc') compiler on certain machines
840 (but definitely not all machines nor all non-`gcc' compilers).
841 Please do not report bugs showing problems compilers have with
842 macros defined in `gcc/f/target.h' and used in places like
845 * Add warning to be printed for each invocation of the compiler if
846 the target machine `INTEGER', `REAL', or `LOGICAL' size is not 32
847 bits, since `g77' is known to not work well for such cases (to be
848 fixed in Version 0.6--*note Actual Bugs We Haven't Fixed Yet:
851 * Lots of new documentation (though work is still needed to put it
852 into canonical GNU format).
854 * Build `libf2c' with `-g0', not `-g2', in effect (by default), to
855 produce smaller library without lots of debugging clutter.
860 * Fix bad code generation involving `X**I' and temporary, internal
861 variables generated by `g77' and the back end (such as for `DO'
864 * Fix crash given `CHARACTER A;DATA A/.TRUE./'.
866 * Replace crash with diagnostic given `CHARACTER A;DATA A/1.0/'.
868 * Fix crash or other erratic behavior when null character constant
869 (`''') is encountered.
871 * Fix crash or other erratic behavior involving diagnosed code.
873 * Fix code generation for external functions returning type `REAL'
874 when the `-ff2c' option is in force (which it is by default) so
875 that `f2c' compatibility is indeed provided.
877 * Disallow `COMMON I(10)' if `I' has previously been specified with
880 * New `-ffixed-line-length-N' option, where N is the maximum length
881 of a typical fixed-form line, defaulting to 72 columns, such that
882 characters beyond column N are ignored, or N is `none', meaning no
883 characters are ignored. does not affect lines with `&' in column
884 1, which are always processed as if `-ffixed-line-length-none' was
887 * No longer generate better code for some kinds of array references,
888 as `gcc' back end is to be fixed to do this even better, and it
889 turned out to slow down some code in some cases after all.
891 * In `COMMON' and `EQUIVALENCE' areas with any members given initial
892 values (e.g. via `DATA'), uninitialized members now always
893 initialized to binary zeros (though this is not required by the
894 standard, and might not be done in future versions of `g77').
895 Previously, in some `COMMON'/`EQUIVALENCE' areas (essentially
896 those with members of more than one type), the uninitialized
897 members were initialized to spaces, to cater to `CHARACTER' types,
898 but it seems no existing code expects that, while much existing
899 code expects binary zeros.
904 * Don't emit bad code when low bound of adjustable array is
905 nonconstant and thus might vary as an expression at run time.
907 * Emit correct code for calculation of number of trips in `DO' loops
908 for cases where the loop should not execute at all. (This bug
909 affected cases where the difference between the begin and end
910 values was less than the step count, though probably not for
911 floating-point cases.)
913 * Fix crash when extra parentheses surround item in `DATA'
916 * Fix crash over minor internal inconsistencies in handling
917 diagnostics, just substitute dummy strings where necessary.
919 * Fix crash on some systems when compiling call to `MVBITS()'
922 * Fix crash on array assignment `TYPEDDD(...)=...', where DDD is a
923 string of one or more digits.
925 * Fix crash on `DCMPLX()' with a single `INTEGER' argument.
927 * Fix various crashes involving code with diagnosed errors.
929 * Support `-I' option for `INCLUDE' statement, plus `gcc''s
930 `header.gcc' facility for handling systems like MS-DOS.
932 * Allow `INCLUDE' statement to be continued across multiple lines,
933 even allow it to coexist with other statements on the same line.
935 * Incorporate Bellcore fixes to `libf2c' through 1995-03-15--this
936 fixes a bug involving infinite loops reading EOF with empty
937 list-directed I/O list.
939 * Remove all the `g77'-specific auto-configuration scripts, code,
940 and so on, except for temporary substitutes for bsearch() and
941 strtoul(), as too many configure/build problems were reported in
942 these areas. People will have to fix their systems' problems
943 themselves, or at least somewhere other than `g77', which expects
944 a working ANSI C environment (and, for now, a GNU C compiler to
945 compile `g77' itself).
947 * Complain if initialized common redeclared as larger in subsequent
950 * Warn if blank common initialized, since its size can vary and hence
951 related warnings that might be helpful won't be seen.
953 * New `-fbackslash' option, on by default, that causes `\' within
954 `CHARACTER' and Hollerith constants to be interpreted a la GNU C.
955 Note that this behavior is somewhat different from `f2c''s, which
956 supports only a limited subset of backslash (escape) sequences.
958 * Make `-fugly-args' the default.
960 * New `-fugly-init' option, on by default, that allows
961 typeless/Hollerith to be specified as initial values for variables
962 or named constants (`PARAMETER'), and also allows
963 character<->numeric conversion in those contexts--turn off via
966 * New `-finit-local-zero' option to initialize local variables to
967 binary zeros. This does not affect whether they are `SAVE'd, i.e.
968 made automatic or static.
970 * New `-Wimplicit' option to warn about implicitly typed variables,
971 arrays, and functions. (Basically causes all program units to
972 default to `IMPLICIT NONE'.)
974 * `-Wall' now implies `-Wuninitialized' as with `gcc' (i.e. unless
975 `-O' not specified, since `-Wuninitialized' requires `-O'), and
976 implies `-Wunused' as well.
978 * `-Wunused' no longer gives spurious messages for unused `EXTERNAL'
979 names (since they are assumed to refer to block data program
980 units, to make use of libraries more reliable).
982 * Support `%LOC()' and `LOC()' of character arguments.
984 * Support null (zero-length) character constants and expressions.
986 * Support `f2c''s `IMAG()' generic intrinsic.
988 * Support `ICHAR()', `IACHAR()', and `LEN()' of character
989 expressions that are valid in assignments but not normally as
992 * Support `f2c'-style `&' in column 1 to mean continuation line.
994 * Allow `NAMELIST', `EXTERNAL', `INTRINSIC', and `VOLATILE' in
995 `BLOCK DATA', even though these are not allowed by the standard.
997 * Allow `RETURN' in main program unit.
999 * Changes to Hollerith-constant support to obey Appendix C of the
1002 - Now padded on the right with zeros, not spaces.
1004 - Hollerith "format specifications" in the form of arrays of
1005 non-character allowed.
1007 - Warnings issued when non-space truncation occurs when
1008 converting to another type.
1010 - When specified as actual argument, now passed by reference to
1011 `INTEGER' (padded on right with spaces if constant too small,
1012 otherwise fully intact if constant wider the `INTEGER' type)
1013 instead of by value.
1015 *Warning:* `f2c' differs on the interpretation of `CALL FOO(1HX)',
1016 which it treats exactly the same as `CALL FOO('X')', but which the
1017 standard and `g77' treat as `CALL FOO(%REF('X '))' (padded with
1018 as many spaces as necessary to widen to `INTEGER'), essentially.
1020 * Changes and fixes to typeless-constant support:
1022 - Now treated as a typeless double-length `INTEGER' value.
1024 - Warnings issued when overflow occurs.
1026 - Padded on the left with zeros when converting to a larger
1029 - Should be properly aligned and ordered on the target machine
1030 for whatever type it is turned into.
1032 - When specified as actual argument, now passed as reference to
1033 a default `INTEGER' constant.
1035 * `%DESCR()' of a non-`CHARACTER' expression now passes a pointer to
1036 the expression plus a length for the expression just as if it were
1037 a `CHARACTER' expression. For example, `CALL FOO(%DESCR(D))',
1038 where `D' is `REAL*8', is the same as `CALL FOO(D,%VAL(8)))'.
1040 * Name of multi-entrypoint master function changed to incorporate
1041 the name of the primary entry point instead of a decimal value, so
1042 the name of the master function for `SUBROUTINE X' with alternate
1043 entry points is now `__g77_masterfun_x'.
1045 * Remove redundant message about zero-step-count `DO' loops.
1047 * Clean up diagnostic messages, shortening many of them.
1049 * Fix typo in `g77' man page.
1051 * Clarify implications of constant-handling bugs in `f/BUGS'.
1053 * Generate better code for `**' operator with a right-hand operand of
1056 * Generate better code for `SQRT()' and `DSQRT()', also when
1057 `-ffast-math' specified, enable better code generation for `SIN()'
1060 * Generate better code for some kinds of array references.
1062 * Speed up lexing somewhat (this makes the compilation phase