1 @c Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002
3 @c Free Software Foundation, Inc.
4 @c This is part of the G77 manual.
5 @c For copying conditions, see the file g77.texi.
7 @c The text of this file appears in the file NEWS
8 @c in the G77 distribution, as well as in the G77 manual.
10 @c Keep this the same as the dates above, since it's used
11 @c in the standalone derivations of this file (e.g. NEWS).
12 @set copyrights-news 1995,1996,1997,1998,1999,2000,2001,2002,2003,2004
14 @set last-update-news 2004-06-17
18 @c The immediately following lines apply to the NEWS file
19 @c which is derived from this file.
20 @emph{Note:} This file is automatically generated from the files
21 @file{news0.texi} and @file{news.texi}.
22 @file{NEWS} is @emph{not} a source file,
23 although it is normally included within source distributions.
25 This file lists news about the @value{which-g77} version
26 (and some other versions) of the GNU Fortran compiler.
27 Copyright (C) @value{copyrights-news} Free Software Foundation, Inc.
28 You may copy, distribute, and modify it freely as long as you preserve
29 this copyright notice and permission notice.
32 @chapter News About GNU Fortran
38 @chapter User-visible Changes
39 @cindex versions, recent
40 @cindex recent versions
41 @cindex changes, user-visible
42 @cindex user-visible changes
44 This chapter describes changes to @command{g77} that are visible
45 to the programmers who actually write and maintain Fortran
46 code they compile with @command{g77}.
47 Information on changes to installation procedures,
48 changes to the documentation, and bug fixes is
49 not provided here, unless it is likely to affect how
50 users use @command{g77}.
51 @xref{News,,News About GNU Fortran}, for information on
52 such changes to @command{g77}.
57 @chapter News About GNU Fortran
58 @cindex versions, recent
59 @cindex recent versions
65 @emph{@code{GCC} 3.4.x is the last edition of @code{GCC} to contain @command{g77} - from @code{GCC} 3.5 onwards, use @command{gfortran}}
67 Changes made to recent versions of GNU Fortran are listed
68 below, with the most recent version first.
70 The changes are generally listed in order:
74 Code-generation and run-time-library bug-fixes
77 Compiler and run-time-library crashes involving valid code
84 Fixes and enhancements to existing features
96 This order is not strict---for example, some items
97 involve a combination of these elements.
100 Note that two variants of @command{g77} are tracked below.
101 The @code{egcs} variant is described vis-a-vis
102 previous versions of @code{egcs} and/or
103 an official FSF version, as appropriate.
104 Note that all such variants are obsolete @emph{as of July 1999} -
105 the information is retained here only for its historical value.
107 Therefore, @code{egcs} versions sometimes have multiple listings
108 to help clarify how they differ from other versions,
109 though this can make getting a complete picture
110 of what a particular @code{egcs} version contains
111 somewhat more difficult.
114 For information on bugs in the @value{which-g77} version of @command{g77},
115 see @ref{Known Bugs,,Known Bugs In GNU Fortran}.
119 For information on bugs in the @value{which-g77} version of @command{g77},
120 see @file{@value{path-g77}/BUGS}.
124 @emph{Warning:} The information below is still under development,
125 and might not accurately reflect the @command{g77} code base
126 of which it is a part.
127 Efforts are made to keep it somewhat up-to-date,
128 but they are particularly concentrated
129 on any version of this information
130 that is distributed as part of a @emph{released} @command{g77}.
132 In particular, while this information is intended to apply to
133 the @value{which-g77} version of @command{g77},
134 only an official @emph{release} of that version
135 is expected to contain documentation that is
136 most consistent with the @command{g77} product in that version.
138 Nevertheless, information on @emph{previous} releases of @command{g77}, below,
139 is likely to be more up-to-date and accurate
140 than the equivalent information that accompanied
142 assuming the last-updated date of the information below
143 is later than the dates of those releases.
145 That's due to attempts to keep this development version
146 of news about previous @command{g77} versions up-to-date.
149 The following information was last updated on @value{last-update-news}:
151 @heading In @code{GCC} 3.4 versus @code{GCC} 3.3:
154 Problem Reports fixed (in chronological order of submission):
157 g77 doesn't accept INTEGER*8 constant in PARAMETER multiplication.
159 (libf2c) isatty does not call f_init.
161 Incorrect documentation for Fortran debugging features.
164 Roger Sayle (@email{roger@@eyesopen.com}) fixed the remaining
165 problems with regard to the support of INTEGER*8, INTEGER*2 and INTEGER*1
166 as a fallout of fixing PR 8485.
169 @heading In @code{GCC} 3.3 versus @code{GCC} 3.2:
172 Problem Reports fixed (in chronological order of submission):
175 -list directed i/o overflow hangs, -fbounds-check doesn't detect
177 g77 generates code which is rejected by GAS if COFF debugging info is
180 Broken links on web pages
182 (libf2c) multiple repeat counts confuse namelist read into array
184 Logical operations error on logicals when using -fugly-logint
186 Generation of C++ Prototype for FORTRAN and extern "C"
188 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os on irix6.5
190 (libf2c) OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should assume a direct
193 g77 "bug"; the executable misbehave (use of options -O2 -fno-automatic
196 (libf2c) DATE_AND_TIME milliseconds field inactive on Windows
198 Incorrect output with 0-based array of characters
200 Double complex zero ** double precision number -> NaN instead of zero
202 -ffixed-line-length-none -x f77-cpp-input gives: Warning: unknown register name line-length-none
204 ICE caused by invalid PARAMETER in implied DO loop
206 Direct access files not unformatted by default
208 Documentation for function IDATE Intrinsic (UNIX) is wrong [fixed in 3.3.1].
211 Richard Henderson (@email{rth@@redhat.com}) analyzed and improved the handling
212 of (no-)aliasing information for dummy arguments and improved the optimization
213 of induction variables in unrolled loops.
216 @heading In @code{GCC} 3.2 versus @code{GCC} 3.1:
219 Problem Reports fixed (in chronological order of submission):
222 ICE in compensate_edge, at reg-stack.c:2591
224 gcc-3.x does not compile files with suffix .r (RATFOR) [Fixed in 3.2.1]
226 [3.2/3.3/3.4 regression] ICE in compensate_edge, at reg-stack.c:2589
230 @heading In @code{GCC} 3.1 (formerly known as g77-0.5.27) versus @code{GCC} 3.0:
233 Problem Reports fixed (in chronological order of submission):
236 Data statement initialization with subscript of kind INTEGER*2
238 Reference to intrinsic `ISHFT' invalid
240 Function BESJN(integer,double) problems
242 g77 -pipe -xf77-cpp-input sends output to stdout
244 g77 -h" gives bogus output
246 ICE on valid input using CALL EXIT(%VAL(...))
248 g77 -v -c -xf77-version /dev/null -xnone causes ice
250 BACKSPACE example that doesn't work as of gcc/g77-3.0.x
252 g77 rejects accepted use of INTEGER*2 as type of DATA statement loop index
254 ICE on compiling source with 540 000 000 REAL array
256 ICE on BESJN(integer*8,real)
258 bug in loop unrolling
260 sparc-sun-solaris2.7 gcc-3.1 extra g77 testsuite failures w/-m64
262 Incorrect acces of integer*1 variables on PA
264 Failure of LAPACK test dtest on irix6.5 with -mabi=64 -O2
268 @command{g77} now has its man page generated from the texinfo documentation,
269 to guarantee that it remains up to date.
272 @command{g77} used to reject the following program on 32-bit targets:
275 DIMENSION A(140 000 000)
280 prog.f: In program `prog':
282 DIMENSION A(140 000 000)
284 Array `a' at (^) is too large to handle
286 because 140 000 000 REALs is larger than the largest bit-extent that can be
287 expressed in 32 bits. However, bit-sizes never play a role after offsets
288 have been converted to byte addresses. Therefore this check has been removed,
289 and the limit is now 2 Gbyte of memory (around 530 000 000 REALs).
290 Note: On GNU/Linux systems one has to compile and link programs that occupy
291 more than 1 Gbyte statically, i.e.@: @code{g77 -static ...}.
294 Based on work done by Juergen Pfeifer (@email{juergen.pfeifer@@gmx.net})
295 libf2c is now a shared library. One can still link in all objects with
296 the program by specifying the @option{-static} option.
299 Robert Anderson (@email{rwa@@alumni.princeton.edu}) thought up a two
300 line change that enables g77 to compile such code as:
304 DIMENSION A(N(1),N(2))
308 Note the use of array elements in the bounds of the adjustable array A.
311 George Helffrich (@email{george@@geo.titech.ac.jp}) implemented a change
312 in substring index checking (when specifying @option{-fbounds-check})
313 that permits the use of zero length substrings of the form
317 Based on code developed by Pedro Vazquez (@email{vazquez@@penelope.iqm.unicamp.br}),
318 the @code{libf2c} library is now able to read and write files larger than
319 2 Gbyte on 32-bit target machines, if the operating system supports this.
322 @heading In 0.5.26, @code{GCC} 3.0 versus @code{GCC} 2.95:
325 When a REWIND was issued after a WRITE statement on an unformatted
326 file, the implicit truncation was performed by copying the truncated
327 file to /tmp and copying the result back. This has been fixed by using
328 the @code{ftruncate} OS function. Thanks go to the GAMESS developers
329 for bringing this to our attention.
332 Using options @option{-g}, @option{-ggdb} or @option{-gdwarf[-2]} (where
333 appropriate for your target) now also enables debugging information
334 for COMMON BLOCK and EQUIVALENCE items to be emitted.
335 Thanks go to Andrew Vaught (@email{andy@@xena.eas.asu.edu}) and
336 George Helffrich (@email{george@@geology.bristol.ac.uk}) for
337 fixing this longstanding problem.
340 It is not necessary anymore to use the option @option{-femulate-complex}
341 to compile Fortran code using COMPLEX arithmetic, even on 64-bit machines
342 (like the Alpha). This will improve code generation.
345 INTRINSIC arithmetic functions are now treated as routines that do not
346 depend on anything but their argument(s). This enables further instruction
347 scheduling, because it is known that they cannot read or modify arbitrary
352 Upgrade to @code{libf2c} as of 2000-12-05.
354 This fixes a bug where a namelist containing initialization of LOGICAL
355 items and a variable starting with T or F would be read incorrectly.
358 The @code{TtyNam} intrinsics now set @var{Name} to all spaces (at run time)
359 if the system has no @code{ttyname} implementation available.
362 Upgrade to @code{libf2c} as of 1999-06-28.
364 This fixes a bug whereby
365 input to a @code{NAMELIST} read involving a repeat count,
366 such as @samp{K(5)=10*3},
367 was not properly handled by @code{libf2c}.
368 The first item was written to @samp{K(5)},
369 but the remaining nine were written elsewhere (still within the array),
370 not necessarily starting at @samp{K(6)}.
374 @heading In 0.5.25, @code{GCC} 2.95 (@code{EGCS} 1.2) versus @code{EGCS} 1.1.2:
378 @command{g77} no longer generates bad code for assignments,
379 or other conversions,
380 of @code{REAL} or @code{COMPLEX} constant expressions
381 to type @code{INTEGER(KIND=2)}
382 (often referred to as @code{INTEGER*8}).
384 For example, @samp{INTEGER*8 J; J = 4E10} now works as documented.
389 @command{g77} no longer truncates @code{INTEGER(KIND=2)}
390 (usually @code{INTEGER*8})
391 subscript expressions when evaluating array references
392 on systems with pointers widers than @code{INTEGER(KIND=1)}
398 @command{g77} no longer generates bad code
399 for an assignment to a @code{COMPLEX} variable or array
400 that partially overlaps one or more of the sources
401 of the same assignment
402 (a very rare construction).
403 It now assigns through a temporary,
404 in cases where such partial overlap is deemed possible.
409 @code{libg2c} (@code{libf2c}) no longer loses track
410 of the file being worked on
411 during a @code{BACKSPACE} operation.
416 @code{libg2c} (@code{libf2c}) fixes a bug whereby
417 input to a @code{NAMELIST} read involving a repeat count,
418 such as @samp{K(5)=10*3},
419 was not properly handled by @code{libf2c}.
420 The first item was written to @samp{K(5)},
421 but the remaining nine were written elsewhere (still within the array),
422 not necessarily starting at @samp{K(6)}.
427 @c Tim Prince reported this, regarding the TEST_FPU benchmark.
428 Automatic arrays now seem to be working on HP-UX systems.
433 The @code{Date} intrinsic now returns the correct result
434 on big-endian systems.
439 Fix @command{g77} so it no longer crashes when compiling
440 I/O statements using keywords that define @code{INTEGER} values,
441 such as @samp{IOSTAT=@var{j}},
442 where @var{j} is other than default @code{INTEGER}
443 (such as @code{INTEGER*2}).
444 Instead, it issues a diagnostic.
449 Fix @command{g77} so it properly handles @samp{DATA A/@var{rpt}*@var{val}/},
450 where @var{rpt} is not default @code{INTEGER}, such as @code{INTEGER*2},
451 instead of producing a spurious diagnostic.
452 Also fix @samp{DATA (A(I),I=1,N)},
453 where @samp{N} is not default @code{INTEGER}
454 to work instead of crashing @command{g77}.
459 The @option{-ax} option is now obeyed when compiling Fortran programs.
460 (It is passed to the @file{f771} driver.)
464 The new @option{-fbounds-check} option
465 causes @command{g77} to compile run-time bounds checks
466 of array subscripts, as well as of substring start and end points.
469 @code{libg2c} now supports building as multilibbed library,
470 which provides better support for systems
471 that require options such as @option{-mieee}
475 Source file names with the suffixes @samp{.FOR} and @samp{.FPP}
476 now are recognized by @command{g77}
477 as if they ended in @samp{.for} and @samp{.fpp}, respectively.
480 The order of arguments to the @emph{subroutine} forms of the
481 @code{CTime}, @code{DTime}, @code{ETime}, and @code{TtyNam}
482 intrinsics has been swapped.
483 The argument serving as the returned value
484 for the corresponding function forms
485 now is the @emph{second} argument,
486 making these consistent with the other subroutine forms
487 of @code{libU77} intrinsics.
490 @command{g77} now warns about a reference to an intrinsic
491 that has an interface that is not Year 2000 (Y2K) compliant.
492 Also, @code{libg2c} has been changed to increase the likelihood
493 of catching references to the implementations of these intrinsics
494 using the @code{EXTERNAL} mechanism
495 (which would avoid the new warnings).
498 @xref{Year 2000 (Y2K) Problems}, for more information.
503 @command{g77} now warns about a reference to a function
504 when the corresponding @emph{subsequent} function program unit
505 disagrees with the reference concerning the type of the function.
509 @option{-fno-emulate-complex} is now the default option.
510 This should result in improved performance
511 of code that uses the @code{COMPLEX} data type.
514 @cindex double-precision performance
515 @cindex -malign-double
517 The @option{-malign-double} option
518 now reliably aligns @emph{all} double-precision variables and arrays
519 on Intel x86 targets.
523 Even without the @option{-malign-double} option,
524 @command{g77} reliably aligns local double-precision variables
525 that are not in @code{EQUIVALENCE} areas
526 and not @code{SAVE}'d.
531 @command{g77} now open-codes (``inlines'') division of @code{COMPLEX} operands
532 instead of generating a run-time call to
533 the @code{libf2c} routines @code{c_div} or @code{z_div},
534 unless the @option{-Os} option is specified.
538 @command{g77} no longer generates code to maintain @code{errno},
539 a C-language concept,
540 when performing operations such as the @code{SqRt} intrinsic.
544 @command{g77} developers can temporarily use
545 the @option{-fflatten-arrays} option
546 to compare how the compiler handles code generation
547 using C-like constructs as compared to the
548 Fortran-like method constructs normally used.
553 A substantial portion of the @command{g77} front end's code-generation component
555 It now generates code using facilities more robustly supported
556 by the @command{gcc} back end.
557 One effect of this rewrite is that some codes no longer produce
558 a spurious ``label @var{lab} used before containing binding contour''
563 Support for the @option{-fugly} option has been removed.
567 Improve documentation and indexing,
568 including information on Year 2000 (Y2K) compliance,
569 and providing more information on internals of the front end.
574 Upgrade to @code{libf2c} as of 1999-05-10.
578 @heading In 0.5.24 versus 0.5.23:
580 There is no @command{g77} version 0.5.24 at this time,
582 0.5.24 is the version number designated for bug fixes and,
583 perhaps, some new features added,
585 Version 0.5.23 requires @command{gcc} 2.8.1,
586 as 0.5.24 was planned to require.
588 Due to @code{EGCS} becoming @code{GCC}
589 (which is now an acronym for ``GNU Compiler Collection''),
590 and @code{EGCS} 1.2 becoming officially designated @code{GCC} 2.95,
591 there seems to be no need for an actual 0.5.24 release.
593 To reduce the confusion already resulting from use of 0.5.24
594 to designate @command{g77} versions within @code{EGCS} versions 1.0 and 1.1,
595 as well as in versions of @command{g77} documentation and notices
597 ``mainline'' @command{g77} version numbering resumes
598 at 0.5.25 with @code{GCC} 2.95 (@code{EGCS} 1.2),
599 skipping over 0.5.24 as a placeholder version number.
601 To repeat, there is no @command{g77} 0.5.24, but there is now a 0.5.25.
602 Please remain calm and return to your keypunch units.
604 @c 1999-03-15: EGCS 1.1.2 released.
605 @heading In @code{EGCS} 1.1.2 versus @code{EGCS} 1.1.1:
609 Fix the @code{IDate} intrinsic (VXT) (in @code{libg2c})
610 so the returned year is in the documented, non-Y2K-compliant range
612 instead of being returned as 100 in the year 2000.
615 @xref{IDate Intrinsic (VXT)},
616 for more information.
620 Fix the @code{Date_and_Time} intrinsic (in @code{libg2c})
621 to return the milliseconds value properly
625 Fix the @code{LStat} intrinsic (in @code{libg2c})
626 to return device-ID information properly
630 Improve documentation.
634 @c 1998-12-04: EGCS 1.1.1 released.
635 @heading In @code{EGCS} 1.1.1 versus @code{EGCS} 1.1:
639 Fix @code{libg2c} so it performs an implicit @code{ENDFILE} operation
641 whenever a @code{REWIND} is done.
643 (This bug was introduced in 0.5.23 and @code{egcs} 1.1 in
644 @command{g77}'s version of @code{libf2c}.)
647 Fix @code{libg2c} so it no longer crashes with a spurious diagnostic
648 upon doing any I/O following a direct formatted write.
650 (This bug was introduced in 0.5.23 and @code{egcs} 1.1 in
651 @command{g77}'s version of @code{libf2c}.)
654 Fix @command{g77} so it no longer crashes compiling references
655 to the @code{Rand} intrinsic on some systems.
658 Fix @command{g77} portion of installation process so it works
659 better on some systems
660 (those with shells requiring @samp{else true} clauses
661 on @code{if} constructs
662 for the completion code to be set properly).
666 @c 1998-09-03: EGCS 1.1 released.
667 @heading In @code{EGCS} 1.1 versus @code{EGCS} 1.0.3:
671 Fix bugs in the @code{libU77} intrinsic @code{HostNm}
672 that wrote one byte beyond the end of its @code{CHARACTER}
674 and in the @code{libU77} intrinsics
675 @code{GMTime} and @code{LTime}
676 that overwrote their arguments.
681 Assumed arrays with negative bounds
682 (such as @samp{REAL A(-1:*)})
683 no longer elicit spurious diagnostics from @command{g77},
684 even on systems with pointers having
685 different sizes than integers.
687 This bug is not known to have existed in any
688 recent version of @command{gcc}.
689 It was introduced in an early release of @code{egcs}.
694 Valid combinations of @code{EXTERNAL},
695 passing that external as a dummy argument
696 without explicitly giving it a type,
697 and, in a subsequent program unit,
698 referencing that external as
699 an external function with a different type
700 no longer crash @command{g77}.
705 @code{CASE DEFAULT} no longer crashes @command{g77}.
710 The @option{-Wunused} option no longer issues a spurious
711 warning about the ``master'' procedure generated by
712 @command{g77} for procedures containing @code{ENTRY} statements.
716 Support @samp{FORMAT(I<@var{expr}>)} when @var{expr} is a
717 compile-time constant @code{INTEGER} expression.
720 Fix @command{g77} @option{-g} option so procedures that
721 use @code{ENTRY} can be stepped through, line by line,
725 Allow any @code{REAL} argument to intrinsics
726 @code{Second} and @code{CPU_Time}.
729 Use @code{tempnam}, if available, to open scratch files
730 (as in @samp{OPEN(STATUS='SCRATCH')})
731 so that the @code{TMPDIR} environment variable,
735 @command{g77}'s version of @code{libf2c} separates out
736 the setting of global state
737 (such as command-line arguments and signal handling)
738 from @file{main.o} into distinct, new library
741 This should make it easier to write portable applications
742 that have their own (non-Fortran) @code{main()} routine
743 properly set up the @code{libf2c} environment, even
744 when @code{libf2c} (now @code{libg2c}) is a shared library.
748 @command{g77} no longer installs the @file{f77} command
749 and @file{f77.1} man page
750 in the @file{/usr} or @file{/usr/local} hierarchy,
751 even if the @file{f77-install-ok} file exists
752 in the source or build directory.
753 See the installation documentation for more information.
758 @command{g77} no longer installs the @file{libf2c.a} library
759 and @file{f2c.h} include file
760 in the @file{/usr} or @file{/usr/local} hierarchy,
761 even if the @file{f2c-install-ok} or @file{f2c-exists-ok} files exist
762 in the source or build directory.
763 See the installation documentation for more information.
768 The @file{libf2c.a} library produced by @command{g77} has been
769 renamed to @file{libg2c.a}.
770 It is installed only in the @command{gcc} ``private''
771 directory hierarchy, @file{gcc-lib}.
772 This allows system administrators and users to choose which
773 version of the @code{libf2c} library from @code{netlib} they
774 wish to use on a case-by-case basis.
775 See the installation documentation for more information.
780 The @file{f2c.h} include (header) file produced by @command{g77}
781 has been renamed to @file{g2c.h}.
782 It is installed only in the @command{gcc} ``private''
783 directory hierarchy, @file{gcc-lib}.
784 This allows system administrators and users to choose which
785 version of the include file from @code{netlib} they
786 wish to use on a case-by-case basis.
787 See the installation documentation for more information.
791 The @command{g77} command now expects the run-time library
792 to be named @code{libg2c.a} instead of @code{libf2c.a},
793 to ensure that a version other than the one built and
794 installed as part of the same @command{g77} version is picked up.
798 During the configuration and build process,
799 @command{g77} creates subdirectories it needs only as it
801 Other cleaning up of the configuration and build process
802 has been performed as well.
807 @code{install-info} now used to update the directory of
808 Info documentation to contain an entry for @command{g77}
809 (during installation).
813 Some diagnostics have been changed from warnings to errors,
814 to prevent inadvertent use of the resulting, probably buggy,
816 These mostly include diagnostics about use of unsupported features
817 in the @code{OPEN}, @code{INQUIRE}, @code{READ}, and
818 @code{WRITE} statements,
819 and about truncations of various sorts of constants.
823 Improve compilation of @code{FORMAT} expressions so that
824 a null byte is appended to the last operand if it
826 This provides a cleaner run-time diagnostic as provided
827 by @code{libf2c} for statements like @samp{PRINT '(I1', 42}.
832 Improve documentation and indexing.
837 The upgrade to @code{libf2c} as of 1998-06-18
838 should fix a variety of problems, including
839 those involving some uses of the @code{T} format
840 specifier, and perhaps some build (porting) problems
845 @c 1998-09-03: EGCS 1.1 released.
846 @heading In @code{EGCS} 1.1 versus @command{g77} 0.5.23:
850 @cindex stack, 387 coprocessor
854 Fix a code-generation bug that afflicted
855 Intel x86 targets when @option{-O2} was specified
856 compiling, for example, an old version of
857 the @code{DNRM2} routine.
859 The x87 coprocessor stack was being
860 mismanaged in cases involving assigned @code{GOTO}
866 @command{g77} no longer produces incorrect code
868 for @code{EQUIVALENCE} and @code{COMMON}
869 aggregates that, due to ``unnatural'' ordering of members
870 vis-a-vis their types, require initial padding.
875 Fix @command{g77} crash compiling code
876 containing the construct @samp{CMPLX(0.)} or similar.
881 @command{g77} no longer crashes when compiling code
882 containing specification statements such as
883 @samp{INTEGER(KIND=7) PTR}.
888 @command{g77} no longer crashes when compiling code
889 such as @samp{J = SIGNAL(1, 2)}.
893 @command{g77} now treats @samp{%LOC(@var{expr})} and
894 @samp{LOC(@var{expr})} as ``ordinary'' expressions
895 when they are used as arguments in procedure calls.
896 This change applies only to global (filewide) analysis,
897 making it consistent with
898 how @command{g77} actually generates code
901 Previously, @command{g77} treated these expressions
902 as denoting special ``pointer'' arguments
903 for the purposes of filewide analysis.
907 Fix @command{g77} crash
908 (or apparently infinite run-time)
909 when compiling certain complicated expressions
910 involving @code{COMPLEX} arithmetic
911 (especially multiplication).
915 @cindex double-precision performance
916 @cindex -malign-double
918 Align static double-precision variables and arrays
920 regardless of whether @option{-malign-double} is specified.
922 Generally, this affects only local variables and arrays
923 having the @code{SAVE} attribute
924 or given initial values via @code{DATA}.
927 The @command{g77} driver now ensures that @option{-lg2c}
928 is specified in the link phase prior to any
929 occurrence of @option{-lm}.
930 This prevents accidentally linking to a routine
931 in the SunOS4 @option{-lm} library
932 when the generated code wants to link to the one
933 in @code{libf2c} (@code{libg2c}).
936 @command{g77} emits more debugging information when
939 This new information allows, for example,
940 @kbd{which __g77_length_a} to be used in @command{gdb}
941 to determine the type of the phantom length argument
942 supplied with @code{CHARACTER} variables.
944 This information pertains to internally-generated
945 type, variable, and other information,
946 not to the longstanding deficiencies vis-a-vis
947 @code{COMMON} and @code{EQUIVALENCE}.
950 The F90 @code{Date_and_Time} intrinsic now is
954 The F90 @code{System_Clock} intrinsic allows
955 the optional arguments (except for the @code{Count}
956 argument) to be omitted.
960 Upgrade to @code{libf2c} as of 1998-06-18.
965 Improve documentation and indexing.
970 @heading In previous versions:
972 Information on previous versions is not provided
973 in this @file{@value{path-g77}/NEWS} file,
975 See @file{@value{path-g77}/news.texi},
976 or any of its other derivations
977 (Info, HTML, dvi forms)
978 for such information.
982 @c 1998-05-20: 0.5.23 released.
983 @heading In 0.5.23 versus 0.5.22:
986 This release contains several regressions against
987 version 0.5.22 of @command{g77}, due to using the
988 ``vanilla'' @command{gcc} back end instead of patching
989 it to fix a few bugs and improve performance in a
992 Features that have been dropped from this version
993 of @command{g77} due to their being implemented
994 via @command{g77}-specific patches to the @command{gcc}
995 back end in previous releases include:
999 Support for @code{__restrict__} keyword,
1000 the options @option{-fargument-alias}, @option{-fargument-noalias},
1001 and @option{-fargument-noalias-global},
1002 and the corresponding alias-analysis code.
1004 (@code{egcs} has the alias-analysis
1005 code, but not the @code{__restrict__} keyword.
1006 @code{egcs} @command{g77} users benefit from the alias-analysis
1007 code despite the lack of the @code{__restrict__} keyword,
1008 which is a C-language construct.)
1011 Support for the GNU compiler options
1012 @option{-fmove-all-movables},
1013 @option{-freduce-all-givs},
1014 and @option{-frerun-loop-opt}.
1016 (@code{egcs} supports these options.
1017 @command{g77} users of @code{egcs} benefit from them even if
1018 they are not explicitly specified,
1019 because the defaults are optimized for @command{g77} users.)
1022 Support for the @option{-W} option warning about
1023 integer division by zero.
1026 The Intel x86-specific option @option{-malign-double}
1027 applying to stack-allocated data
1028 as well as statically-allocate data.
1031 @ifclear USERVISONLY
1032 Note that the @file{gcc/f/gbe/} subdirectory has been removed
1033 from this distribution as a result of @command{g77} no longer
1034 including patches for the @command{gcc} back end.
1037 @ifclear USERVISONLY
1039 Fix bugs in the @code{libU77} intrinsic @code{HostNm}
1040 that wrote one byte beyond the end of its @code{CHARACTER}
1042 and in the @code{libU77} intrinsics
1043 @code{GMTime} and @code{LTime}
1044 that overwrote their arguments.
1048 Support @command{gcc} version 2.8,
1049 and remove support for prior versions of @command{gcc}.
1051 @cindex -@w{}-driver option
1052 @cindex @command{g77} options, -@w{}-driver
1053 @cindex options, -@w{}-driver
1055 Remove support for the @option{--driver} option,
1056 as @command{g77} now does all the driving,
1057 just like @command{gcc}.
1059 @ifclear USERVISONLY
1061 @code{CASE DEFAULT} no longer crashes @command{g77}.
1064 @ifclear USERVISONLY
1066 Valid combinations of @code{EXTERNAL},
1067 passing that external as a dummy argument
1068 without explicitly giving it a type,
1069 and, in a subsequent program unit,
1070 referencing that external as
1071 an external function with a different type
1072 no longer crash @command{g77}.
1075 @ifclear USERVISONLY
1077 @command{g77} no longer installs the @file{f77} command
1078 and @file{f77.1} man page
1079 in the @file{/usr} or @file{/usr/local} hierarchy,
1080 even if the @file{f77-install-ok} file exists
1081 in the source or build directory.
1082 See the installation documentation for more information.
1085 @ifclear USERVISONLY
1087 @command{g77} no longer installs the @file{libf2c.a} library
1088 and @file{f2c.h} include file
1089 in the @file{/usr} or @file{/usr/local} hierarchy,
1090 even if the @file{f2c-install-ok} or @file{f2c-exists-ok} files exist
1091 in the source or build directory.
1092 See the installation documentation for more information.
1095 @ifclear USERVISONLY
1097 The @file{libf2c.a} library produced by @command{g77} has been
1098 renamed to @file{libg2c.a}.
1099 It is installed only in the @command{gcc} ``private''
1100 directory hierarchy, @file{gcc-lib}.
1101 This allows system administrators and users to choose which
1102 version of the @code{libf2c} library from @code{netlib} they
1103 wish to use on a case-by-case basis.
1104 See the installation documentation for more information.
1107 @ifclear USERVISONLY
1109 The @file{f2c.h} include (header) file produced by @command{g77}
1110 has been renamed to @file{g2c.h}.
1111 It is installed only in the @command{gcc} ``private''
1112 directory hierarchy, @file{gcc-lib}.
1113 This allows system administrators and users to choose which
1114 version of the include file from @code{netlib} they
1115 wish to use on a case-by-case basis.
1116 See the installation documentation for more information.
1120 The @command{g77} command now expects the run-time library
1121 to be named @code{libg2c.a} instead of @code{libf2c.a},
1122 to ensure that a version other than the one built and
1123 installed as part of the same @command{g77} version is picked up.
1125 @ifclear USERVISONLY
1127 The @option{-Wunused} option no longer issues a spurious
1128 warning about the ``master'' procedure generated by
1129 @command{g77} for procedures containing @code{ENTRY} statements.
1133 @command{g77}'s version of @code{libf2c} separates out
1134 the setting of global state
1135 (such as command-line arguments and signal handling)
1136 from @file{main.o} into distinct, new library
1139 This should make it easier to write portable applications
1140 that have their own (non-Fortran) @code{main()} routine
1141 properly set up the @code{libf2c} environment, even
1142 when @code{libf2c} (now @code{libg2c}) is a shared library.
1144 @ifclear USERVISONLY
1146 During the configuration and build process,
1147 @command{g77} creates subdirectories it needs only as it
1148 needs them, thus avoiding unnecessary creation of, for example,
1149 @file{stage1/f/runtime} when doing a non-bootstrap build.
1150 Other cleaning up of the configuration and build process
1151 has been performed as well.
1154 @ifclear USERVISONLY
1156 @code{install-info} now used to update the directory of
1157 Info documentation to contain an entry for @command{g77}
1158 (during installation).
1162 Some diagnostics have been changed from warnings to errors,
1163 to prevent inadvertent use of the resulting, probably buggy,
1165 These mostly include diagnostics about use of unsupported features
1166 in the @code{OPEN}, @code{INQUIRE}, @code{READ}, and
1167 @code{WRITE} statements,
1168 and about truncations of various sorts of constants.
1170 @ifclear USERVISONLY
1172 Improve documentation and indexing.
1175 @ifclear USERVISONLY
1177 Upgrade to @code{libf2c} as of 1998-04-20.
1179 This should fix a variety of problems, including
1180 those involving some uses of the @code{T} format
1181 specifier, and perhaps some build (porting) problems
1186 @c 1998-03-16: 0.5.22 released.
1187 @heading In 0.5.22 versus 0.5.21:
1189 @ifclear USERVISONLY
1191 Fix code generation for iterative @code{DO} loops that
1192 have one or more references to the iteration variable,
1193 or to aliases of it, in their control expressions.
1194 For example, @samp{DO 10 J=2,J} now is compiled correctly.
1197 @ifclear USERVISONLY
1199 @cindex stack, 387 coprocessor
1203 Fix a code-generation bug that afflicted
1204 Intel x86 targets when @option{-O2} was specified
1205 compiling, for example, an old version of
1206 the @code{DNRM2} routine.
1208 The x87 coprocessor stack was being
1209 mismanaged in cases involving assigned @code{GOTO}
1213 @ifclear USERVISONLY
1215 Fix @code{DTime} intrinsic so as not to truncate
1216 results to integer values (on some systems).
1220 Fix @code{Signal} intrinsic so it offers portable
1221 support for 64-bit systems (such as Digital Alphas
1224 @ifclear USERVISONLY
1226 Fix run-time crash involving @code{NAMELIST} on 64-bit
1227 machines such as Alphas.
1230 @ifclear USERVISONLY
1232 Fix @command{g77} version of @code{libf2c} so it no longer
1233 produces a spurious @samp{I/O recursion} diagnostic at run time
1234 when an I/O operation (such as @samp{READ *,I}) is interrupted
1235 in a manner that causes the program to be terminated
1236 via the @code{f_exit} routine (such as via @kbd{C-c}).
1239 @ifclear USERVISONLY
1241 Fix @command{g77} crash triggered by @code{CASE} statement with
1242 an omitted lower or upper bound.
1245 @ifclear USERVISONLY
1247 Fix @command{g77} crash compiling references to @code{CPU_Time}
1251 @ifclear USERVISONLY
1253 Fix @command{g77} crash
1254 (or apparently infinite run-time)
1255 when compiling certain complicated expressions
1256 involving @code{COMPLEX} arithmetic
1257 (especially multiplication).
1260 @ifclear USERVISONLY
1262 Fix @command{g77} crash on statements such as
1263 @samp{PRINT *, (REAL(Z(I)),I=1,2)}, where
1264 @samp{Z} is @code{DOUBLE COMPLEX}.
1267 @ifclear USERVISONLY
1269 Fix a @command{g++} crash.
1273 Support @samp{FORMAT(I<@var{expr}>)} when @var{expr} is a
1274 compile-time constant @code{INTEGER} expression.
1277 Fix @command{g77} @option{-g} option so procedures that
1278 use @code{ENTRY} can be stepped through, line by line,
1281 @ifclear USERVISONLY
1283 Fix a profiling-related bug in @command{gcc} back end for
1284 Intel x86 architecture.
1288 Allow any @code{REAL} argument to intrinsics
1289 @code{Second} and @code{CPU_Time}.
1292 Allow any numeric argument to intrinsics
1293 @code{Int2} and @code{Int8}.
1296 Use @code{tempnam}, if available, to open scratch files
1297 (as in @samp{OPEN(STATUS='SCRATCH')})
1298 so that the @code{TMPDIR} environment variable,
1299 if present, is used.
1302 Rename the @command{gcc} keyword @code{restrict} to
1303 @code{__restrict__}, to avoid rejecting valid, existing,
1305 Support for @code{restrict} is now more like support
1308 @ifclear USERVISONLY
1310 Fix @option{-fpedantic} to not reject procedure invocations
1311 such as @samp{I=J()} and @samp{CALL FOO()}.
1315 Fix @option{-fugly-comma} to affect invocations of
1316 only external procedures.
1317 Restore rejection of gratuitous trailing omitted
1318 arguments to intrinsics, as in @samp{I=MAX(3,4,,)}.
1321 Fix compiler so it accepts @option{-fgnu-intrinsics-*} and
1322 @option{-fbadu77-intrinsics-*} options.
1324 @ifclear USERVISONLY
1326 Improve diagnostic messages from @code{libf2c}
1327 so it is more likely that the printing of the
1328 active format string is limited to the string,
1329 with no trailing garbage being printed.
1331 (Unlike @command{f2c}, @command{g77} did not append
1332 a null byte to its compiled form of every
1333 format string specified via a @code{FORMAT} statement.
1334 However, @command{f2c} would exhibit the problem
1335 anyway for a statement like @samp{PRINT '(I)garbage', 1}
1336 by printing @samp{(I)garbage} as the format string.)
1339 @ifclear USERVISONLY
1341 Improve compilation of @code{FORMAT} expressions so that
1342 a null byte is appended to the last operand if it
1344 This provides a cleaner run-time diagnostic as provided
1345 by @code{libf2c} for statements like @samp{PRINT '(I1', 42}.
1348 @ifclear USERVISONLY
1350 Fix various crashes involving code with diagnosed errors.
1353 @ifclear USERVISONLY
1355 Fix cross-compilation bug when configuring @code{libf2c}.
1358 @ifclear USERVISONLY
1360 Improve diagnostics.
1363 @ifclear USERVISONLY
1365 Improve documentation and indexing.
1368 @ifclear USERVISONLY
1370 Upgrade to @code{libf2c} as of 1997-09-23.
1371 This fixes a formatted-I/O bug that afflicted
1372 64-bit systems with 32-bit integers
1373 (such as Digital Alpha running GNU/Linux).
1377 @c 1998-03-18: EGCS 1.0.2 released.
1378 @heading In @code{EGCS} 1.0.2 versus @code{EGCS} 1.0.1:
1380 @ifclear USERVISONLY
1382 Fix @command{g77} crash triggered by @code{CASE} statement with
1383 an omitted lower or upper bound.
1386 @ifclear USERVISONLY
1388 Fix @command{g77} crash on statements such as
1389 @samp{PRINT *, (REAL(Z(I)),I=1,2)}, where
1390 @samp{Z} is @code{DOUBLE COMPLEX}.
1393 @ifclear USERVISONLY
1395 @cindex support, ELF
1396 @cindex -fPIC option
1397 @cindex options, -fPIC
1399 Fix @option{-fPIC} (such as compiling for ELF targets)
1400 on the Intel x86 architecture target
1401 so invalid assembler code is no longer produced.
1404 @ifclear USERVISONLY
1406 Fix @option{-fpedantic} to not reject procedure invocations
1407 such as @samp{I=J()} and @samp{CALL FOO()}.
1410 @ifclear USERVISONLY
1412 Fix @option{-fugly-comma} to affect invocations of
1413 only external procedures.
1414 Restore rejection of gratuitous trailing omitted
1415 arguments to intrinsics, as in @samp{I=MAX(3,4,,)}.
1419 Fix compiler so it accepts @option{-fgnu-intrinsics-*} and
1420 @option{-fbadu77-intrinsics-*} options.
1423 @c 1998-01-06: EGCS 1.0.1 released.
1424 @heading In @code{EGCS} 1.0.1 versus @code{EGCS} 1.0:
1425 @ifclear USERVISONLY
1428 Fix run-time crash involving @code{NAMELIST} on 64-bit
1429 machines such as Alphas.
1433 @c 1997-12-03: EGCS 1.0 released.
1434 @heading In @code{EGCS} 1.0 versus @command{g77} 0.5.21:
1437 Version 1.0 of @code{egcs}
1438 contains several regressions against
1439 version 0.5.21 of @command{g77},
1441 ``vanilla'' @command{gcc} back end instead of patching
1442 it to fix a few bugs and improve performance in a
1445 Features that have been dropped from this version
1446 of @command{g77} due to their being implemented
1447 via @command{g77}-specific patches to the @command{gcc}
1448 back end in previous releases include:
1452 Support for the C-language @code{restrict} keyword.
1455 Support for the @option{-W} option warning about
1456 integer division by zero.
1459 The Intel x86-specific option @option{-malign-double}
1460 applying to stack-allocated data
1461 as well as statically-allocate data.
1464 @ifclear USERVISONLY
1465 Note that the @file{gcc/f/gbe/} subdirectory has been removed
1466 from this distribution as a result of @command{g77}
1467 being fully integrated with
1468 the @code{egcs} variant of the @command{gcc} back end.
1471 @ifclear USERVISONLY
1473 Fix code generation for iterative @code{DO} loops that
1474 have one or more references to the iteration variable,
1475 or to aliases of it, in their control expressions.
1476 For example, @samp{DO 10 J=2,J} now is compiled correctly.
1479 @ifclear USERVISONLY
1481 Fix @code{DTime} intrinsic so as not to truncate
1482 results to integer values (on some systems).
1485 @ifclear USERVISONLY
1487 @c Toon Moene discovered these.
1488 Some Fortran code, miscompiled
1489 by @command{g77} built on @command{gcc} version 2.8.1
1490 on m68k-next-nextstep3 configurations
1491 when using the @option{-O2} option,
1492 is now compiled correctly.
1493 It is believed that a C function known to miscompile
1494 on that configuration
1495 when using the @samp{-O2 -funroll-loops} options
1496 also is now compiled correctly.
1499 @ifclear USERVISONLY
1501 Remove support for non-@code{egcs} versions of @command{gcc}.
1504 @cindex -@w{}-driver option
1505 @cindex @command{g77} options, -@w{}-driver
1506 @cindex options, -@w{}-driver
1508 Remove support for the @option{--driver} option,
1509 as @command{g77} now does all the driving,
1510 just like @command{gcc}.
1513 Allow any numeric argument to intrinsics
1514 @code{Int2} and @code{Int8}.
1516 @ifclear USERVISONLY
1518 Improve diagnostic messages from @code{libf2c}
1519 so it is more likely that the printing of the
1520 active format string is limited to the string,
1521 with no trailing garbage being printed.
1523 (Unlike @command{f2c}, @command{g77} did not append
1524 a null byte to its compiled form of every
1525 format string specified via a @code{FORMAT} statement.
1526 However, @code{f2c} would exhibit the problem
1527 anyway for a statement like @samp{PRINT '(I)garbage', 1}
1528 by printing @samp{(I)garbage} as the format string.)
1531 @ifclear USERVISONLY
1533 Upgrade to @code{libf2c} as of 1997-09-23.
1534 This fixes a formatted-I/O bug that afflicted
1535 64-bit systems with 32-bit integers
1536 (such as Digital Alpha running GNU/Linux).
1540 @c 1997-09-09: 0.5.21 released.
1543 @ifclear USERVISONLY
1545 Fix a code-generation bug introduced by 0.5.20
1546 caused by loop unrolling (by specifying
1547 @option{-funroll-loops} or similar).
1548 This bug afflicted all code compiled by
1549 version 2.7.2.2.f.2 of @command{gcc} (C, C++,
1550 Fortran, and so on).
1553 @ifclear USERVISONLY
1555 Fix a code-generation bug manifested when
1556 combining local @code{EQUIVALENCE} with a
1557 @code{DATA} statement that follows
1558 the first executable statement (or is
1559 treated as an executable-context statement
1560 as a result of using the @option{-fpedantic}
1564 @ifclear USERVISONLY
1566 Fix a compiler crash that occurred when an
1567 integer division by a constant zero is detected.
1568 Instead, when the @option{-W} option is specified,
1569 the @command{gcc} back end issues a warning about such a case.
1570 This bug afflicted all code compiled by
1571 version 2.7.2.2.f.2 of @command{gcc} (C, C++,
1572 Fortran, and so on).
1576 When the @option{-W} option is specified, @command{gcc}, @command{g77},
1577 and other GNU compilers that incorporate the @command{gcc}
1578 back end as modified by @command{g77}, issue
1579 a warning about integer division by constant zero.
1582 @ifclear USERVISONLY
1584 Fix a compiler crash that occurred in some cases
1585 of procedure inlining.
1586 (Such cases became more frequent in 0.5.20.)
1589 @ifclear USERVISONLY
1591 Fix a compiler crash resulting from using @code{DATA}
1592 or similar to initialize a @code{COMPLEX} variable or
1596 @ifclear USERVISONLY
1598 Fix compiler crashes involving use of @code{AND}, @code{OR},
1599 or @code{XOR} intrinsics.
1602 @ifclear USERVISONLY
1604 Fix compiler bug triggered when using a @code{COMMON}
1605 or @code{EQUIVALENCE} variable
1606 as the target of an @code{ASSIGN}
1607 or assigned-@code{GOTO} statement.
1610 @ifclear USERVISONLY
1612 Fix compiler crashes due to using the name of a some
1613 non-standard intrinsics (such as @code{FTell} or
1614 @code{FPutC}) as such and as the name of a procedure
1616 Such dual use of a name in a program is allowed by
1620 @c @command{g77}'s version of @code{libf2c} has been modified
1621 @c so that the external names of library's procedures do not
1622 @c conflict with names used for Fortran procedures compiled
1623 @c by @command{g77}.
1624 @c An additional layer of jacket procedures has been added
1625 @c to @code{libf2c} to map the old names to the new names,
1626 @c for automatic use by programs that interface to the
1627 @c library procedures via the external-procedure mechanism.
1629 @c For example, the intrinsic @code{FPUTC} previously was
1630 @c implemented by @command{g77} as a call to the @code{libf2c}
1631 @c routine @code{fputc_}.
1632 @c This would conflict with a Fortran procedure named @code{FPUTC}
1633 @c (using default compiler options), and this conflict
1634 @c would cause a crash under certain circumstances.
1636 @c Now, the intrinsic @code{FPUTC} calls @code{G77_fputc_0},
1637 @c which does not conflict with the @code{fputc_} external
1638 @c that implements a Fortran procedure named @code{FPUTC}.
1640 @c Programs that refer to @code{FPUTC} as an external procedure
1641 @c without supplying their own implementation will link to
1642 @c the new @code{libf2c} routine @code{fputc_}, which is
1643 @c simply a jacket routine that calls @code{G77_fputc_0}.
1645 @ifclear USERVISONLY
1647 Place automatic arrays on the stack, even if
1648 @code{SAVE} or the @option{-fno-automatic} option
1650 This avoids a compiler crash in some cases.
1653 @ifclear USERVISONLY
1655 The @option{-malign-double} option now reliably aligns
1656 @code{DOUBLE PRECISION} optimally on Pentium and
1657 Pentium Pro architectures (586 and 686 in @command{gcc}).
1661 New option @option{-Wno-globals} disables warnings
1662 about ``suspicious'' use of a name both as a global
1663 name and as the implicit name of an intrinsic, and
1664 warnings about disagreements over the number or natures of
1665 arguments passed to global procedures, or the
1666 natures of the procedures themselves.
1668 The default is to issue such warnings, which are
1669 new as of this version of @command{g77}.
1672 New option @option{-fno-globals} disables diagnostics
1673 about potentially fatal disagreements
1674 analysis problems, such as disagreements over the
1675 number or natures of arguments passed to global
1676 procedures, or the natures of those procedures themselves.
1678 The default is to issue such diagnostics and flag
1679 the compilation as unsuccessful.
1680 With this option, the diagnostics are issued as
1681 warnings, or, if @option{-Wno-globals} is specified,
1682 are not issued at all.
1684 This option also disables inlining of global procedures,
1685 to avoid compiler crashes resulting from coding errors
1686 that these diagnostics normally would identify.
1688 @ifclear USERVISONLY
1690 Diagnose cases where a reference to a procedure
1691 disagrees with the type of that procedure, or
1692 where disagreements about the number or nature
1694 This avoids a compiler crash.
1697 @ifclear USERVISONLY
1699 Fix parsing bug whereby @command{g77} rejected a
1700 second initialization specification immediately
1701 following the first's closing @samp{/} without
1702 an intervening comma in a @code{DATA} statement,
1703 and the second specification was an implied-DO list.
1706 @ifclear USERVISONLY
1708 Improve performance of the @command{gcc} back end so
1709 certain complicated expressions involving @code{COMPLEX}
1710 arithmetic (especially multiplication) don't appear to
1711 take forever to compile.
1714 @ifclear USERVISONLY
1716 Fix a couple of profiling-related bugs in @command{gcc}
1720 @ifclear USERVISONLY
1722 Integrate GNU Ada's (GNAT's) changes to the back end,
1723 which consist almost entirely of bug fixes.
1724 These fixes are circa version 3.10p of GNAT.
1727 @ifclear USERVISONLY
1729 Include some other @command{gcc} fixes that seem useful in
1730 @command{g77}'s version of @command{gcc}.
1731 (See @file{gcc/ChangeLog} for details---compare it
1732 to that file in the vanilla @code{gcc-2.7.2.3.tar.gz}
1737 Fix @code{libU77} routines that accept file and other names
1738 to strip trailing blanks from them, for consistency
1739 with other implementations.
1740 Blanks may be forcibly appended to such names by
1741 appending a single null character (@samp{CHAR(0)})
1742 to the significant trailing blanks.
1745 Fix @code{CHMOD} intrinsic to work with file names
1746 that have embedded blanks, commas, and so on.
1749 Fix @code{SIGNAL} intrinsic so it accepts an
1750 optional third @code{Status} argument.
1752 @ifclear USERVISONLY
1754 Fix @code{IDATE()} intrinsic subroutine (VXT form)
1755 so it accepts arguments in the correct order.
1756 Documentation fixed accordingly, and for
1757 @code{GMTIME()} and @code{LTIME()} as well.
1761 Make many changes to @code{libU77} intrinsics to
1762 support existing code more directly.
1764 Such changes include allowing both subroutine and
1765 function forms of many routines, changing @code{MCLOCK()}
1766 and @code{TIME()} to return @code{INTEGER(KIND=1)} values,
1767 introducing @code{MCLOCK8()} and @code{TIME8()} to
1768 return @code{INTEGER(KIND=2)} values,
1769 and placing functions that are intended to perform
1770 side effects in a new intrinsic group, @code{badu77}.
1772 @ifclear USERVISONLY
1774 Improve @code{libU77} so it is more portable.
1778 Add options @option{-fbadu77-intrinsics-delete},
1779 @option{-fbadu77-intrinsics-hide}, and so on.
1781 @ifclear USERVISONLY
1783 Fix crashes involving diagnosed or invalid code.
1786 @ifclear USERVISONLY
1788 @command{g77} and @command{gcc} now do a somewhat better
1789 job detecting and diagnosing arrays that are too
1790 large to handle before these cause diagnostics
1791 during the assembler or linker phase, a compiler
1792 crash, or generation of incorrect code.
1795 @ifclear USERVISONLY
1797 Make some fixes to alias analysis code.
1800 @ifclear USERVISONLY
1802 Add support for @code{restrict} keyword in @command{gcc}
1806 @ifclear USERVISONLY
1808 Support @command{gcc} version 2.7.2.3
1809 (modified by @command{g77} into version 2.7.2.3.f.1),
1811 support for prior versions of @command{gcc}.
1814 @ifclear USERVISONLY
1816 Incorporate GNAT's patches to the @command{gcc} back
1817 end into @command{g77}'s, so GNAT users do not need
1818 to apply GNAT's patches to build both GNAT and @command{g77}
1819 from the same source tree.
1822 @ifclear USERVISONLY
1824 Modify @command{make} rules and related code so that
1825 generation of Info documentation doesn't require
1826 compilation using @command{gcc}.
1827 Now, any ANSI C compiler should be adequate to
1828 produce the @command{g77} documentation (in particular,
1829 the tables of intrinsics) from scratch.
1833 Add @code{INT2} and @code{INT8} intrinsics.
1836 Add @code{CPU_TIME} intrinsic.
1839 Add @code{ALARM} intrinsic.
1842 @code{CTIME} intrinsic now accepts any @code{INTEGER}
1843 argument, not just @code{INTEGER(KIND=2)}.
1845 @ifclear USERVISONLY
1847 Warn when explicit type declaration disagrees with
1848 the type of an intrinsic invocation.
1851 @ifclear USERVISONLY
1853 Support @samp{*f771} entry in @command{gcc} @file{specs} file.
1856 @ifclear USERVISONLY
1858 Fix typo in @command{make} rule @command{g77-cross}, used only for
1862 @ifclear USERVISONLY
1864 Fix @code{libf2c} build procedure to re-archive library
1865 if previous attempt to archive was interrupted.
1868 @ifclear USERVISONLY
1870 Change @command{gcc} to unroll loops only during the last
1871 invocation (of as many as two invocations) of loop
1875 @ifclear USERVISONLY
1877 Improve handling of @option{-fno-f2c} so that code that
1878 attempts to pass an intrinsic as an actual argument,
1879 such as @samp{CALL FOO(ABS)}, is rejected due to the fact
1880 that the run-time-library routine is, effectively,
1881 compiled with @option{-ff2c} in effect.
1884 @ifclear USERVISONLY
1886 Fix @command{g77} driver to recognize @option{-fsyntax-only}
1887 as an option that inhibits linking, just like @option{-c} or
1888 @option{-S}, and to recognize and properly handle the
1889 @option{-nostdlib}, @option{-M}, @option{-MM}, @option{-nodefaultlibs},
1890 and @option{-Xlinker} options.
1893 @ifclear USERVISONLY
1895 Upgrade to @code{libf2c} as of 1997-08-16.
1898 @ifclear USERVISONLY
1900 Modify @code{libf2c} to consistently and clearly diagnose
1901 recursive I/O (at run time).
1905 @command{g77} driver now prints version information (such as produced
1906 by @kbd{g77 -v}) to @code{stderr} instead of @code{stdout}.
1909 The @samp{.r} suffix now designates a Ratfor source file,
1910 to be preprocessed via the @command{ratfor} command, available
1913 @ifclear USERVISONLY
1915 Fix some aspects of how @command{gcc} determines what kind of
1916 system is being configured and what kinds are supported.
1917 For example, GNU Linux/Alpha ELF systems now are directly
1921 @ifclear USERVISONLY
1923 Improve diagnostics.
1926 @ifclear USERVISONLY
1928 Improve documentation and indexing.
1931 @ifclear USERVISONLY
1933 Include all pertinent files for @code{libf2c} that come
1934 from @code{netlib.bell-labs.com}; give any such files
1935 that aren't quite accurate in @command{g77}'s version of
1936 @code{libf2c} the suffix @samp{.netlib}.
1939 @ifclear USERVISONLY
1941 Reserve @code{INTEGER(KIND=0)} for future use.
1945 @c 1997-02-28: 0.5.20 released.
1949 The @option{-fno-typeless-boz} option is now the default.
1951 This option specifies that non-decimal-radix
1952 constants using the prefixed-radix form (such as @samp{Z'1234'})
1953 are to be interpreted as @code{INTEGER(KIND=1)} constants.
1954 Specify @option{-ftypeless-boz} to cause such
1955 constants to be interpreted as typeless.
1957 (Version 0.5.19 introduced @option{-fno-typeless-boz} and
1961 @xref{Fortran Dialect Options,,Options Controlling Fortran Dialect},
1962 for information on the @option{-ftypeless-boz} option.
1966 Options @option{-ff90-intrinsics-enable} and
1967 @option{-fvxt-intrinsics-enable} now are the
1970 Some programs might use names that clash with
1971 intrinsic names defined (and now enabled) by these
1972 options or by the new @code{libU77} intrinsics.
1973 Users of such programs might need to compile them
1974 differently (using, for example, @option{-ff90-intrinsics-disable})
1975 or, better yet, insert appropriate @code{EXTERNAL}
1976 statements specifying that these names are not intended
1977 to be names of intrinsics.
1980 The @code{ALWAYS_FLUSH} macro is no longer defined when
1981 building @code{libf2c}, which should result in improved
1982 I/O performance, especially over NFS.
1984 @emph{Note:} If you have code that depends on the behavior
1985 of @code{libf2c} when built with @code{ALWAYS_FLUSH} defined,
1986 you will have to modify @code{libf2c} accordingly before
1987 building it from this and future versions of @command{g77}.
1990 @xref{Output Assumed To Flush}, for more information.
1994 Dave Love's implementation of @code{libU77} has been
1995 added to the version of @code{libf2c} distributed with
1996 and built as part of @command{g77}.
1997 @command{g77} now knows about the routines in this library
2001 New option @option{-fvxt} specifies that the
2002 source file is written in VXT Fortran, instead of GNU Fortran.
2005 @xref{VXT Fortran}, for more information on the constructs
2006 recognized when the @option{-fvxt} option is specified.
2010 The @option{-fvxt-not-f90} option has been deleted,
2011 along with its inverse, @option{-ff90-not-vxt}.
2013 If you used one of these deleted options, you should
2014 re-read the pertinent documentation to determine which
2015 options, if any, are appropriate for compiling your
2016 code with this version of @command{g77}.
2019 @xref{Other Dialects}, for more information.
2023 The @option{-fugly} option now issues a warning, as it
2024 likely will be removed in a future version.
2026 (Enabling all the @option{-fugly-*} options is unlikely
2027 to be feasible, or sensible, in the future,
2028 so users should learn to specify only those
2029 @option{-fugly-*} options they really need for a
2030 particular source file.)
2033 The @option{-fugly-assumed} option, introduced in
2034 version 0.5.19, has been changed to
2035 better accommodate old and new code.
2038 @xref{Ugly Assumed-Size Arrays}, for more information.
2041 @ifclear USERVISONLY
2043 Make a number of fixes to the @command{g77} front end and
2044 the @command{gcc} back end to better support Alpha (AXP)
2046 This includes providing at least one bug-fix to the
2047 @command{gcc} back end for Alphas.
2051 Related to supporting Alpha (AXP) machines, the @code{LOC()}
2052 intrinsic and @code{%LOC()} construct now return
2053 values of @code{INTEGER(KIND=0)} type,
2054 as defined by the GNU Fortran language.
2056 This type is wide enough
2057 (holds the same number of bits)
2058 as the character-pointer type on the machine.
2060 On most machines, this won't make a difference,
2061 whereas, on Alphas and other systems with 64-bit pointers,
2062 the @code{INTEGER(KIND=0)} type is equivalent to @code{INTEGER(KIND=2)}
2063 (often referred to as @code{INTEGER*8})
2064 instead of the more common @code{INTEGER(KIND=1)}
2065 (often referred to as @code{INTEGER*4}).
2068 Emulate @code{COMPLEX} arithmetic in the @command{g77} front
2069 end, to avoid bugs in @code{complex} support in the
2070 @command{gcc} back end.
2071 New option @option{-fno-emulate-complex}
2072 causes @command{g77} to revert the 0.5.19 behavior.
2074 @ifclear USERVISONLY
2076 Fix bug whereby @samp{REAL A(1)}, for example, caused
2077 a compiler crash if @option{-fugly-assumed} was in effect
2078 and @var{A} was a local (automatic) array.
2079 That case is no longer affected by the new
2080 handling of @option{-fugly-assumed}.
2083 @ifclear USERVISONLY
2085 Fix @command{g77} command driver so that @samp{g77 -o foo.f}
2086 no longer deletes @file{foo.f} before issuing other
2087 diagnostics, and so the @option{-x} option is properly
2091 @ifclear USERVISONLY
2093 Enable inlining of subroutines and functions by the @command{gcc}
2095 This works as it does for @command{gcc} itself---program units
2096 may be inlined for invocations that follow them in the same
2097 program unit, as long as the appropriate compile-time
2098 options are specified.
2102 Dummy arguments are no longer assumed to potentially alias
2104 other dummy arguments or @code{COMMON} areas when any of
2105 these are defined (assigned to) by Fortran code.
2107 This can result in faster and/or smaller programs when
2108 compiling with optimization enabled, though on some
2109 systems this effect is observed only when @option{-fforce-addr}
2112 New options @option{-falias-check}, @option{-fargument-alias},
2113 @option{-fargument-noalias},
2114 and @option{-fno-argument-noalias-global} control the
2115 way @command{g77} handles potential aliasing.
2118 @xref{Aliasing Assumed To Work}, for detailed information on why the
2119 new defaults might result in some programs no longer working the way they
2120 did when compiled by previous versions of @command{g77}.
2123 @ifclear USERVISONLY
2125 The @code{CONJG()} and @code{DCONJG()} intrinsics now
2126 are compiled in-line.
2129 @ifclear USERVISONLY
2131 The bug-fix for 0.5.19.1 has been re-done.
2132 The @command{g77} compiler has been changed back to
2133 assume @code{libf2c} has no aliasing problems in
2134 its implementations of the @code{COMPLEX} (and
2135 @code{DOUBLE COMPLEX}) intrinsics.
2136 The @code{libf2c} has been changed to have no such
2139 As a result, 0.5.20 is expected to offer improved performance
2140 over 0.5.19.1, perhaps as good as 0.5.19 in most
2141 or all cases, due to this change alone.
2143 @emph{Note:} This change requires version 0.5.20 of
2144 @code{libf2c}, at least, when linking code produced
2145 by any versions of @command{g77} other than 0.5.19.1.
2146 Use @samp{g77 -v} to determine the version numbers
2147 of the @code{libF77}, @code{libI77}, and @code{libU77}
2148 components of the @code{libf2c} library.
2149 (If these version numbers are not printed---in
2150 particular, if the linker complains about unresolved
2151 references to names like @samp{g77__fvers__}---that
2152 strongly suggests your installation has an obsolete
2153 version of @code{libf2c}.)
2157 New option @option{-fugly-assign} specifies that the
2158 same memory locations are to be used to hold the
2159 values assigned by both statements @samp{I = 3} and
2160 @samp{ASSIGN 10 TO I}, for example.
2161 (Normally, @command{g77} uses a separate memory location
2162 to hold assigned statement labels.)
2165 @xref{Ugly Assigned Labels}, for more information.
2169 @code{FORMAT} and @code{ENTRY} statements now are allowed to
2170 precede @code{IMPLICIT NONE} statements.
2172 @ifclear USERVISONLY
2174 Produce diagnostic for unsupported @code{SELECT CASE} on
2175 @code{CHARACTER} type, instead of crashing, at compile time.
2178 @ifclear USERVISONLY
2180 Fix crashes involving diagnosed or invalid code.
2183 @ifclear USERVISONLY
2185 Change approach to building @code{libf2c} archive
2186 (@file{libf2c.a}) so that members are added to it
2187 only when truly necessary, so the user that installs
2188 an already-built @command{g77} doesn't need to have write
2189 access to the build tree (whereas the user doing the
2190 build might not have access to install new software
2194 @ifclear USERVISONLY
2196 Support @command{gcc} version 2.7.2.2
2197 (modified by @command{g77} into version 2.7.2.2.f.2),
2199 support for prior versions of @command{gcc}.
2202 @ifclear USERVISONLY
2204 Upgrade to @code{libf2c} as of 1997-02-08, and
2205 fix up some of the build procedures.
2208 @ifclear USERVISONLY
2210 Improve general build procedures for @command{g77},
2211 fixing minor bugs (such as deletion of any file
2212 named @file{f771} in the parent directory of @code{gcc/}).
2216 Enable full support of @code{INTEGER(KIND=2)}
2217 (often referred to as @code{INTEGER*8})
2219 @code{libf2c} and @file{f2c.h} so that @command{f2c} users
2220 may make full use of its features via the @command{g77}
2221 version of @file{f2c.h} and the @code{INTEGER(KIND=2)}
2222 support routines in the @command{g77} version of @code{libf2c}.
2225 Improve @command{g77} driver and @code{libf2c} so that @samp{g77 -v}
2226 yields version information on the library.
2229 The @code{SNGL} and @code{FLOAT} intrinsics now are
2230 specific intrinsics, instead of synonyms for the
2231 generic intrinsic @code{REAL}.
2234 New intrinsics have been added.
2235 These are @code{REALPART}, @code{IMAGPART},
2237 @code{LONG}, and @code{SHORT}.
2240 A new group of intrinsics, @code{gnu}, has been added
2241 to contain the new @code{REALPART}, @code{IMAGPART},
2242 and @code{COMPLEX} intrinsics.
2243 An old group, @code{dcp}, has been removed.
2246 Complain about industry-wide ambiguous references
2247 @samp{REAL(@var{expr})} and @samp{AIMAG(@var{expr})},
2248 where @var{expr} is @code{DOUBLE COMPLEX} (or any
2249 complex type other than @code{COMPLEX}), unless
2250 @option{-ff90} option specifies Fortran 90 interpretation
2251 or new @option{-fugly-complex} option, in conjunction with
2252 @option{-fnot-f90}, specifies @command{f2c} interpretation.
2254 @ifclear USERVISONLY
2256 Make improvements to diagnostics.
2259 @ifclear USERVISONLY
2261 Speed up compiler a bit.
2264 @ifclear USERVISONLY
2266 Improvements to documentation and indexing, including
2267 a new chapter containing information on one, later
2268 more, diagnostics that users are directed to pull
2269 up automatically via a message in the diagnostic itself.
2271 (Hence the menu item @code{M} for the node
2272 @code{Diagnostics} in the top-level menu of
2273 the Info documentation.)
2277 @ifclear DOC-OLDNEWS
2278 @heading In previous versions:
2280 Information on previous versions is archived
2281 in @file{@value{path-g77}/news.texi}
2282 following the test of the @code{DOC-OLDNEWS} macro.
2286 @c 1997-02-01: 0.5.19.1 released.
2287 @heading In 0.5.19.1:
2290 Code-generation bugs afflicting operations on complex
2291 data have been fixed.
2293 These bugs occurred when assigning the result of an
2294 operation to a complex variable (or array element)
2295 that also served as an input to that operation.
2297 The operations affected by this bug were: @code{CONJG()},
2298 @code{DCONJG()}, @code{CCOS()}, @code{CDCOS()},
2299 @code{CLOG()}, @code{CDLOG()}, @code{CSIN()}, @code{CDSIN()},
2300 @code{CSQRT()}, @code{CDSQRT()}, complex division, and
2301 raising a @code{DOUBLE COMPLEX} operand to an @code{INTEGER}
2303 (The related generic and @samp{Z}-prefixed intrinsics,
2304 such as @code{ZSIN()}, also were affected.)
2306 For example, @samp{C = CSQRT(C)}, @samp{Z = Z/C}, and @samp{Z = Z**I}
2307 (where @samp{C} is @code{COMPLEX} and @samp{Z} is
2308 @code{DOUBLE COMPLEX}) have been fixed.
2311 @c 1996-12-07: 0.5.19 released.
2315 Fix @code{FORMAT} statement parsing so negative values for
2316 specifiers such as @code{P} (e.g. @samp{FORMAT(-1PF8.1)})
2317 are correctly processed as negative.
2320 Fix @code{SIGNAL} intrinsic so it once again accepts a
2321 procedure as its second argument.
2324 A temporary kludge option provides bare-bones information on
2325 @code{COMMON} and @code{EQUIVALENCE} members at debug time.
2328 New @option{-fonetrip} option specifies FORTRAN-66-style
2329 one-trip @code{DO} loops.
2332 New @option{-fno-silent} option causes names of program units
2333 to be printed as they are compiled, in a fashion similar to
2334 UNIX @command{f77} and @command{f2c}.
2337 New @option{-fugly-assumed} option specifies that arrays
2338 dimensioned via @samp{DIMENSION X(1)}, for example, are to be
2339 treated as assumed-size.
2342 New @option{-fno-typeless-boz} option specifies that non-decimal-radix
2343 constants using the prefixed-radix form (such as @samp{Z'1234'})
2344 are to be interpreted as @code{INTEGER} constants.
2347 New @option{-ff66} option is a ``shorthand'' option that specifies
2348 behaviors considered appropriate for FORTRAN 66 programs.
2351 New @option{-ff77} option is a ``shorthand'' option that specifies
2352 behaviors considered appropriate for UNIX @command{f77} programs.
2355 New @option{-fugly-comma} and @option{-fugly-logint} options provided
2356 to perform some of what @option{-fugly} used to do.
2357 @option{-fugly} and @option{-fno-ugly} are now ``shorthand'' options,
2358 in that they do nothing more than enable (or disable) other
2359 @option{-fugly-*} options.
2362 Fix parsing of assignment statements involving targets that
2363 are substrings of elements of @code{CHARACTER} arrays having
2364 names such as @samp{READ}, @samp{WRITE}, @samp{GOTO}, and
2365 @samp{REALFUNCTIONFOO}.
2368 Fix crashes involving diagnosed code.
2371 Fix handling of local @code{EQUIVALENCE} areas so certain cases
2372 of valid Fortran programs are not misdiagnosed as improperly
2373 extending the area backwards.
2376 Support @command{gcc} version 2.7.2.1.
2379 Upgrade to @code{libf2c} as of 1996-09-26, and
2380 fix up some of the build procedures.
2383 Change code generation for list-directed I/O so it allows
2384 for new versions of @code{libf2c} that might return nonzero
2385 status codes for some operations previously assumed to always
2388 This change not only affects how @code{IOSTAT=} variables
2389 are set by list-directed I/O, it also affects whether
2390 @code{END=} and @code{ERR=} labels are reached by these
2394 Add intrinsic support for new @code{FTELL} and @code{FSEEK}
2395 procedures in @code{libf2c}.
2398 Modify @code{fseek_()} in @code{libf2c} to be more portable
2399 (though, in practice, there might be no systems where this
2400 matters) and to catch invalid @code{whence} arguments.
2403 Some useless warnings from the @option{-Wunused} option have
2407 Fix a problem building the @file{f771} executable
2408 on AIX systems by linking with the @option{-bbigtoc} option.
2411 Abort configuration if @command{gcc} has not been patched
2412 using the patch file provided in the @file{gcc/f/gbe/}
2416 Add options @option{--help} and @option{--version} to the
2417 @command{g77} command, to conform to GNU coding guidelines.
2418 Also add printing of @command{g77} version number when
2419 the @option{--verbose} (@option{-v}) option is used.
2422 Change internally generated name for local @code{EQUIVALENCE}
2423 areas to one based on the alphabetically sorted first name
2424 in the list of names for entities placed at the beginning
2428 Improvements to documentation and indexing.
2431 @c 1996-04-01: 0.5.18 released.
2435 Add some rudimentary support for @code{INTEGER*1},
2436 @code{INTEGER*2}, @code{INTEGER*8},
2437 and their @code{LOGICAL} equivalents.
2438 (This support works on most, maybe all, @command{gcc} targets.)
2440 Thanks to Scott Snyder (@email{snyder@@d0sgif.fnal.gov})
2441 for providing the patch for this!
2443 Among the missing elements from the support for these
2444 features are full intrinsic support and constants.
2447 Add some rudimentary support for the @code{BYTE} and
2448 @code{WORD} type-declaration statements.
2449 @code{BYTE} corresponds to @code{INTEGER*1},
2450 while @code{WORD} corresponds to @code{INTEGER*2}.
2452 Thanks to Scott Snyder (@email{snyder@@d0sgif.fnal.gov})
2453 for providing the patch for this!
2456 The compiler code handling intrinsics has been largely
2457 rewritten to accommodate the new types.
2458 No new intrinsics or arguments for existing
2459 intrinsics have been added, so there is, at this
2460 point, no intrinsic to convert to @code{INTEGER*8},
2464 Support automatic arrays in procedures.
2467 Reduce space/time requirements for handling large
2468 @emph{sparsely} initialized aggregate arrays.
2469 This improvement applies to only a subset of
2470 the general problem to be addressed in 0.6.
2473 Treat initial values of zero as if they weren't
2474 specified (in DATA and type-declaration statements).
2475 The initial values will be set to zero anyway, but the amount
2476 of compile time processing them will be reduced,
2477 in some cases significantly (though, again, this
2478 is only a subset of the general problem to be
2481 A new option, @option{-fzeros}, is introduced to
2482 enable the traditional treatment of zeros as any
2486 With @option{-ff90} in force, @command{g77} incorrectly
2487 interpreted @samp{REAL(Z)} as returning a @code{REAL}
2488 result, instead of as a @code{DOUBLE PRECISION}
2490 (Here, @samp{Z} is @code{DOUBLE COMPLEX}.)
2492 With @option{-fno-f90} in force, the interpretation remains
2493 unchanged, since this appears to be how at least some
2494 F77 code using the @code{DOUBLE COMPLEX} extension expected
2497 Essentially, @samp{REAL(Z)} in F90 is the same as
2498 @samp{DBLE(Z)}, while in extended F77, it appears to
2499 be the same as @samp{REAL(REAL(Z))}.
2502 An expression involving exponentiation, where both operands
2503 were type @code{INTEGER} and the right-hand operand
2504 was negative, was erroneously evaluated.
2507 Fix bugs involving @code{DATA} implied-@code{DO} constructs
2508 (these involved an errant diagnostic and a crash, both on good
2509 code, one involving subsequent statement-function definition).
2512 Close @code{INCLUDE} files after processing them, so compiling source
2513 files with lots of @code{INCLUDE} statements does not result in
2514 being unable to open @code{INCLUDE} files after all the available
2515 file descriptors are used up.
2518 Speed up compiling, especially of larger programs, and perhaps
2519 slightly reduce memory utilization while compiling (this is
2520 @emph{not} the improvement planned for 0.6 involving large aggregate
2521 areas)---these improvements result from simply turning
2522 off some low-level code to do self-checking that hasn't been
2523 triggered in a long time.
2526 Introduce three new options that
2527 implement optimizations in the @command{gcc} back end (GBE).
2528 These options are @option{-fmove-all-movables}, @option{-freduce-all-givs},
2529 and @option{-frerun-loop-opt}, which are enabled, by default,
2530 for Fortran compilations.
2531 These optimizations are intended to help toon Fortran programs.
2534 Patch the GBE to do a better job optimizing certain
2535 kinds of references to array elements.
2538 Due to patches to the GBE, the version number of @command{gcc}
2539 also is patched to make it easier to manage installations,
2540 especially useful if it turns out a @command{g77} change to the
2543 The @command{g77}-modified version number is the @command{gcc}
2544 version number with the string @samp{.f.@var{n}} appended,
2545 where @samp{f} identifies the version as enhanced for
2546 Fortran, and @var{n} is @samp{1} for the first Fortran
2547 patch for that version of @command{gcc}, @samp{2} for the
2550 So, this introduces version 2.7.2.f.1 of @command{gcc}.
2553 Make several improvements and fixes to diagnostics, including
2554 the removal of two that were inappropriate or inadequate.
2557 Warning about two successive arithmetic operators, produced
2558 by @option{-Wsurprising}, now produced @emph{only} when both
2559 operators are, indeed, arithmetic (not relational/boolean).
2562 @option{-Wsurprising} now warns about the remaining cases
2563 of using non-integral variables for implied-@code{DO}
2564 loops, instead of these being rejected unless @option{-fpedantic}
2565 or @option{-fugly} specified.
2568 Allow @code{SAVE} of a local variable or array, even after
2569 it has been given an initial value via @code{DATA}, for example.
2572 Introduce an Info version of @command{g77} documentation, which
2573 supersedes @file{gcc/f/CREDITS}, @file{gcc/f/DOC}, and
2574 @file{gcc/f/PROJECTS}.
2575 These files will be removed in a future release.
2576 The files @file{gcc/f/BUGS}, @file{gcc/f/INSTALL}, and
2577 @file{gcc/f/NEWS} now are automatically built from
2578 the texinfo source when distributions are made.
2580 This effort was inspired by a first pass at translating
2581 @file{g77-0.5.16/f/DOC} that was contributed to Craig by
2582 David Ronis (@email{ronis@@onsager.chem.mcgill.ca}).
2585 New @option{-fno-second-underscore} option to specify
2586 that, when @option{-funderscoring} is in effect, a second
2587 underscore is not to be appended to Fortran names already
2588 containing an underscore.
2591 Change the way iterative @code{DO} loops work to follow
2593 In particular, calculation of the iteration count is
2594 still done by converting the start, end, and increment
2595 parameters to the type of the @code{DO} variable, but
2596 the result of the calculation is always converted to
2597 the default @code{INTEGER} type.
2599 (This should have no effect on existing code compiled
2600 by @command{g77}, but code written to assume that use
2601 of a @emph{wider} type for the @code{DO} variable
2602 will result in an iteration count being fully calculated
2603 using that wider type (wider
2604 than default @code{INTEGER}) must be rewritten.)
2607 Support @command{gcc} version 2.7.2.
2610 Upgrade to @code{libf2c} as of 1996-03-23, and
2611 fix up some of the build procedures.
2613 Note that the email addresses related to @command{f2c}
2614 have changed---the distribution site now is
2615 named @code{netlib.bell-labs.com}, and the
2616 maintainer's new address is @email{dmg@@bell-labs.com}.
2619 @c 1995-11-18: 0.5.17 released.
2623 @strong{Fix serious bug} in @samp{g77 -v} command that can cause removal of a
2624 system's @file{/dev/null} special file if run by user @code{root}.
2626 @strong{All users} of version 0.5.16 should ensure that
2627 they have not removed @file{/dev/null} or replaced it with an ordinary
2628 file (e.g. by comparing the output of @samp{ls -l /dev/null} with
2629 @samp{ls -l /dev/zero}.
2630 If the output isn't basically the
2631 same, contact your system
2632 administrator about restoring @file{/dev/null} to its proper status).
2634 This bug is particularly insidious because removing @file{/dev/null} as
2635 a special file can go undetected for quite a while, aside from
2636 various applications and programs exhibiting sudden, strange
2639 I sincerely apologize for not realizing the
2640 implications of the fact that when @samp{g77 -v} runs the @command{ld} command
2641 with @samp{-o /dev/null} that @command{ld} tries to @emph{remove} the executable
2642 it is supposed to build (especially if it reports unresolved
2643 references, which it should in this case)!
2646 Fix crash on @samp{CHARACTER*(*) FOO} in a main or block data program unit.
2649 Fix crash that can occur when diagnostics given outside of any
2650 program unit (such as when input file contains @samp{@@foo}).
2653 Fix crashes, infinite loops (hangs), and such involving diagnosed code.
2656 Fix @code{ASSIGN}'ed variables so they can be @code{SAVE}'d or dummy arguments,
2657 and issue clearer error message in cases where target of @code{ASSIGN}
2658 or @code{ASSIGN}ed @code{GOTO}/@code{FORMAT} is too small (which should
2662 Make @code{libf2c} build procedures work on more systems again by
2663 eliminating unnecessary invocations of @samp{ld -r -x} and @command{mv}.
2666 Fix omission of @option{-funix-intrinsics-@dots{}} options in list of permitted
2667 options to compiler.
2670 Fix failure to always diagnose missing type declaration for
2671 @code{IMPLICIT NONE}.
2674 Fix compile-time performance problem (which could sometimes
2675 crash the compiler, cause a hang, or whatever, due to a bug
2676 in the back end) involving exponentiation with a large @code{INTEGER}
2677 constant for the right-hand operator (e.g. @samp{I**32767}).
2680 Fix build procedures so cross-compiling @command{g77} (the @command{fini}
2681 utility in particular) is properly built using the host compiler.
2684 Add new @option{-Wsurprising} option to warn about constructs that are
2685 interpreted by the Fortran standard (and @command{g77}) in ways that
2686 are surprising to many programmers.
2689 Add @code{ERF()} and @code{ERFC()} as generic intrinsics mapping to existing
2690 @code{ERF}/@code{DERF} and @code{ERFC}/@code{DERFC} specific intrinsics.
2692 @emph{Note:} You should
2693 specify @samp{INTRINSIC ERF,ERFC} in any code where you might use
2694 these as generic intrinsics, to improve likelihood of diagnostics
2695 (instead of subtle run-time bugs) when using a compiler that
2696 doesn't support these as intrinsics (e.g. @command{f2c}).
2699 Remove from @option{-fno-pedantic} the diagnostic about @code{DO}
2700 with non-@code{INTEGER} index variable; issue that under
2701 @option{-Wsurprising} instead.
2704 Clarify some diagnostics that say things like ``ignored'' when that's
2708 Clarify diagnostic on use of @code{.EQ.}/@code{.NE.} on @code{LOGICAL}
2712 Minor improvements to code generation for various operations on
2713 @code{LOGICAL} operands.
2716 Minor improvement to code generation for some @code{DO} loops on some
2720 Support @command{gcc} version 2.7.1.
2723 Upgrade to @code{libf2c} as of 1995-11-15.
2726 @c 1995-08-30: 0.5.16 released.
2730 Fix a code-generation bug involving complicated @code{EQUIVALENCE} statements
2731 not involving @code{COMMON}.
2734 Fix code-generation bugs involving invoking ``gratis'' library procedures
2735 in @code{libf2c} from code compiled with @option{-fno-f2c} by making these
2736 procedures known to @command{g77} as intrinsics (not affected by -fno-f2c).
2737 This is known to fix code invoking @code{ERF()}, @code{ERFC()},
2738 @code{DERF()}, and @code{DERFC()}.
2741 Update @code{libf2c} to include netlib patches through 1995-08-16, and
2742 @code{#define} @code{WANT_LEAD_0} to 1 to make @command{g77}-compiled code more
2743 consistent with other Fortran implementations by outputting
2744 leading zeros in formatted and list-directed output.
2747 Fix a code-generation bug involving adjustable dummy arrays with high
2748 bounds whose primaries are changed during procedure execution, and
2749 which might well improve code-generation performance for such arrays
2750 compared to @command{f2c} plus @command{gcc} (but apparently only when using
2751 @file{gcc-2.7.0} or later).
2754 Fix a code-generation bug involving invocation of @code{COMPLEX} and
2755 @code{DOUBLE COMPLEX} @code{FUNCTION}s and doing @code{COMPLEX} and
2756 @code{DOUBLE COMPLEX} divides, when the result
2757 of the invocation or divide is assigned directly to a variable
2758 that overlaps one or more of the arguments to the invocation or divide.
2761 Fix crash by not generating new optimal code for @samp{X**I} if @samp{I} is
2762 nonconstant and the expression is used to dimension a dummy
2763 array, since the @command{gcc} back end does not support the necessary
2764 mechanics (and the @command{gcc} front end rejects the equivalent
2765 construct, as it turns out).
2768 Fix crash on expressions like @samp{COMPLEX**INTEGER}.
2771 Fix crash on expressions like @samp{(1D0,2D0)**2}, i.e. raising a
2772 @code{DOUBLE COMPLEX} constant to an @code{INTEGER} constant power.
2775 Fix crashes and such involving diagnosed code.
2778 Diagnose, instead of crashing on, statement function definitions
2779 having duplicate dummy argument names.
2782 Fix bug causing rejection of good code involving statement function
2786 Fix bug resulting in debugger not knowing size of local equivalence
2787 area when any member of area has initial value (via @code{DATA},
2791 Fix installation bug that prevented installation of @command{g77} driver.
2792 Provide for easy selection of whether to install copy of @command{g77}
2793 as @command{f77} to replace the broken code.
2796 Fix @command{gcc} driver (affects @command{g77} thereby) to not
2797 gratuitously invoke the
2798 @code{f771} program (e.g. when @option{-E} is specified).
2801 Fix diagnostic to point to correct source line when it immediately
2802 follows an @code{INCLUDE} statement.
2805 Support more compiler options in @command{gcc}/@command{g77} when
2806 compiling Fortran files.
2807 These options include @option{-p}, @option{-pg}, @option{-aux-info}, @option{-P},
2808 correct setting of version-number macros for preprocessing, full
2809 recognition of @option{-O0}, and
2810 automatic insertion of configuration-specific linker specs.
2813 Add new intrinsics that interface to existing routines in @code{libf2c}:
2814 @code{ABORT}, @code{DERF}, @code{DERFC}, @code{ERF}, @code{ERFC}, @code{EXIT},
2815 @code{FLUSH}, @code{GETARG}, @code{GETENV}, @code{IARGC},
2816 @code{SIGNAL}, and @code{SYSTEM}.
2817 Note that @code{ABORT}, @code{EXIT}, @code{FLUSH}, @code{SIGNAL}, and
2818 @code{SYSTEM} are intrinsic subroutines, not functions (since they
2819 have side effects), so to get the return values from @code{SIGNAL}
2820 and @code{SYSTEM}, append a final argument specifying an @code{INTEGER}
2821 variable or array element (e.g. @samp{CALL SYSTEM('rm foo',ISTAT)}).
2824 Add new intrinsic group named @code{unix} to contain the new intrinsics,
2825 and by default enable this new group.
2828 Move @code{LOC()} intrinsic out of the @code{vxt} group to the new
2832 Improve @command{g77} so that @samp{g77 -v} by itself (or with
2833 certain other options, including @option{-B}, @option{-b}, @option{-i},
2834 @option{-nostdlib}, and @option{-V}) reports lots more useful
2835 version info, and so that long-form options @command{gcc} accepts are
2836 understood by @command{g77} as well (even in truncated, unambiguous forms).
2839 Add new @command{g77} option @option{--driver=name} to specify driver when
2840 default, @command{gcc}, isn't appropriate.
2843 Add support for @samp{#} directives (as output by the preprocessor) in the
2844 compiler, and enable generation of those directives by the
2845 preprocessor (when compiling @samp{.F} files) so diagnostics and debugging
2846 info are more useful to users of the preprocessor.
2849 Produce better diagnostics, more like @command{gcc}, with info such as
2850 @samp{In function `foo':} and @samp{In file included from...:}.
2853 Support @command{gcc}'s @option{-fident} and @option{-fno-ident} options.
2856 When @option{-Wunused} in effect, don't warn about local variables used as
2857 statement-function dummy arguments or @code{DATA} implied-@code{DO} iteration
2858 variables, even though, strictly speaking, these are not uses
2859 of the variables themselves.
2862 When @samp{-W -Wunused} in effect, don't warn about unused dummy arguments
2863 at all, since there's no way to turn this off for individual
2864 cases (@command{g77} might someday start warning about these)---applies
2865 to @command{gcc} versions 2.7.0 and later, since earlier versions didn't
2866 warn about unused dummy arguments.
2869 New option @option{-fno-underscoring} that inhibits transformation of names
2870 (by appending one or two underscores) so users may experiment
2871 with implications of such an environment.
2874 Minor improvement to @file{gcc/f/info} module to make it easier to build
2875 @command{g77} using the native (non-@command{gcc}) compiler on certain machines
2876 (but definitely not all machines nor all non-@command{gcc} compilers).
2878 do not report bugs showing problems compilers have with
2879 macros defined in @file{gcc/f/target.h} and used in places like
2880 @file{gcc/f/expr.c}.
2883 Add warning to be printed for each invocation of the compiler
2884 if the target machine @code{INTEGER}, @code{REAL}, or @code{LOGICAL} size
2886 since @command{g77} is known to not work well for such cases.
2889 Lots of new documentation (though work is still needed to put it into
2890 canonical GNU format).
2893 Build @code{libf2c} with @option{-g0}, not @option{-g2}, in effect
2894 (by default), to produce
2895 smaller library without lots of debugging clutter.
2898 @c 1995-05-19: 0.5.15 released.
2902 Fix bad code generation involving @samp{X**I} and temporary, internal variables
2903 generated by @command{g77} and the back end (such as for @code{DO} loops).
2906 Fix crash given @samp{CHARACTER A;DATA A/.TRUE./}.
2909 Replace crash with diagnostic given @samp{CHARACTER A;DATA A/1.0/}.
2912 Fix crash or other erratic behavior when null character constant
2913 (@samp{''}) is encountered.
2916 Fix crash or other erratic behavior involving diagnosed code.
2919 Fix code generation for external functions returning type @code{REAL} when
2920 the @option{-ff2c} option is in force (which it is by default) so that
2921 @command{f2c} compatibility is indeed provided.
2924 Disallow @samp{COMMON I(10)} if @samp{I} has previously been specified
2925 with an array declarator.
2928 New @option{-ffixed-line-length-@var{n}} option, where @var{n} is the
2930 of a typical fixed-form line, defaulting to 72 columns, such
2931 that characters beyond column @var{n} are ignored, or @var{n} is @samp{none},
2932 meaning no characters are ignored.
2933 does not affect lines
2934 with @samp{&} in column 1, which are always processed as if
2935 @option{-ffixed-line-length-none} was in effect.
2938 No longer generate better code for some kinds of array references,
2939 as @command{gcc} back end is to be fixed to do this even better, and it
2940 turned out to slow down some code in some cases after all.
2943 In @code{COMMON} and @code{EQUIVALENCE} areas with any members given initial
2944 values (e.g. via @code{DATA}), uninitialized members now always
2945 initialized to binary zeros (though this is not required by
2946 the standard, and might not be done in future versions
2948 Previously, in some @code{COMMON}/@code{EQUIVALENCE} areas
2949 (essentially those with members of more than one type), the
2950 uninitialized members were initialized to spaces, to
2951 cater to @code{CHARACTER} types, but it seems no existing code expects
2952 that, while much existing code expects binary zeros.
2958 Don't emit bad code when low bound of adjustable array is nonconstant
2959 and thus might vary as an expression at run time.
2962 Emit correct code for calculation of number of trips in @code{DO} loops
2964 where the loop should not execute at all.
2965 (This bug affected cases
2966 where the difference between the begin and end values was less
2967 than the step count, though probably not for floating-point cases.)
2970 Fix crash when extra parentheses surround item in
2971 @code{DATA} implied-@code{DO} list.
2974 Fix crash over minor internal inconsistencies in handling diagnostics,
2975 just substitute dummy strings where necessary.
2978 Fix crash on some systems when compiling call to @code{MVBITS()} intrinsic.
2981 Fix crash on array assignment @samp{TYPE@var{ddd}(@dots{})=@dots{}}, where @var{ddd}
2982 is a string of one or more digits.
2985 Fix crash on @code{DCMPLX()} with a single @code{INTEGER} argument.
2988 Fix various crashes involving code with diagnosed errors.
2991 Support @option{-I} option for @code{INCLUDE} statement, plus @command{gcc}'s
2992 @file{header.gcc} facility for handling systems like MS-DOS.
2995 Allow @code{INCLUDE} statement to be continued across multiple lines,
2996 even allow it to coexist with other statements on the same line.
2999 Incorporate Bellcore fixes to @code{libf2c} through 1995-03-15---this
3000 fixes a bug involving infinite loops reading EOF with empty list-directed
3004 Remove all the @command{g77}-specific auto-configuration scripts, code,
3006 except for temporary substitutes for bsearch() and strtoul(), as
3007 too many configure/build problems were reported in these areas.
3008 People will have to fix their systems' problems themselves, or at
3009 least somewhere other than @command{g77}, which expects a working ANSI C
3010 environment (and, for now, a GNU C compiler to compile @command{g77} itself).
3013 Complain if initialized common redeclared as larger in subsequent program
3017 Warn if blank common initialized, since its size can vary and hence
3018 related warnings that might be helpful won't be seen.
3021 New @option{-fbackslash} option, on by default, that causes @samp{\}
3022 within @code{CHARACTER}
3023 and Hollerith constants to be interpreted a la GNU C.
3025 this behavior is somewhat different from @command{f2c}'s, which supports only
3026 a limited subset of backslash (escape) sequences.
3029 Make @option{-fugly-args} the default.
3032 New @option{-fugly-init} option, on by default, that allows typeless/Hollerith
3033 to be specified as initial values for variables or named constants
3034 (@code{PARAMETER}), and also allows character<->numeric conversion in
3035 those contexts---turn off via @option{-fno-ugly-init}.
3038 New @option{-finit-local-zero} option to initialize
3039 local variables to binary zeros.
3040 This does not affect whether they are @code{SAVE}d, i.e. made
3041 automatic or static.
3044 New @option{-Wimplicit} option to warn about implicitly typed variables, arrays,
3046 (Basically causes all program units to default to @code{IMPLICIT NONE}.)
3049 @option{-Wall} now implies @option{-Wuninitialized} as with @command{gcc}
3050 (i.e. unless @option{-O} not specified, since @option{-Wuninitialized}
3051 requires @option{-O}), and implies @option{-Wunused} as well.
3054 @option{-Wunused} no longer gives spurious messages for unused
3055 @code{EXTERNAL} names (since they are assumed to refer to block data
3056 program units, to make use of libraries more reliable).
3059 Support @code{%LOC()} and @code{LOC()} of character arguments.
3062 Support null (zero-length) character constants and expressions.
3065 Support @command{f2c}'s @code{IMAG()} generic intrinsic.
3068 Support @code{ICHAR()}, @code{IACHAR()}, and @code{LEN()} of
3069 character expressions that are valid in assignments but
3070 not normally as actual arguments.
3073 Support @command{f2c}-style @samp{&} in column 1 to mean continuation line.
3076 Allow @code{NAMELIST}, @code{EXTERNAL}, @code{INTRINSIC}, and @code{VOLATILE}
3077 in @code{BLOCK DATA}, even though these are not allowed by the standard.
3080 Allow @code{RETURN} in main program unit.
3083 Changes to Hollerith-constant support to obey Appendix C of the
3088 Now padded on the right with zeros, not spaces.
3091 Hollerith ``format specifications'' in the form of arrays of
3092 non-character allowed.
3095 Warnings issued when non-space truncation occurs when converting
3099 When specified as actual argument, now passed
3100 by reference to @code{INTEGER} (padded on right with spaces if constant
3101 too small, otherwise fully intact if constant wider the @code{INTEGER}
3102 type) instead of by value.
3105 @strong{Warning:} @command{f2c} differs on the
3106 interpretation of @samp{CALL FOO(1HX)}, which it treats exactly the
3107 same as @samp{CALL FOO('X')}, but which the standard and @command{g77} treat
3108 as @samp{CALL FOO(%REF('X '))} (padded with as many spaces as necessary
3109 to widen to @code{INTEGER}), essentially.
3112 Changes and fixes to typeless-constant support:
3116 Now treated as a typeless double-length @code{INTEGER} value.
3119 Warnings issued when overflow occurs.
3122 Padded on the left with zeros when converting
3126 Should be properly aligned and ordered on
3127 the target machine for whatever type it is turned into.
3130 When specified as actual argument, now passed as reference to
3131 a default @code{INTEGER} constant.
3135 @code{%DESCR()} of a non-@code{CHARACTER} expression now passes a pointer to
3136 the expression plus a length for the expression just as if
3137 it were a @code{CHARACTER} expression.
3138 For example, @samp{CALL FOO(%DESCR(D))}, where
3139 @samp{D} is @code{REAL*8}, is the same as @samp{CALL FOO(D,%VAL(8)))}.
3142 Name of multi-entrypoint master function changed to incorporate
3143 the name of the primary entry point instead of a decimal
3144 value, so the name of the master function for @samp{SUBROUTINE X}
3145 with alternate entry points is now @samp{__g77_masterfun_x}.
3148 Remove redundant message about zero-step-count @code{DO} loops.
3151 Clean up diagnostic messages, shortening many of them.
3154 Fix typo in @command{g77} man page.
3157 Clarify implications of constant-handling bugs in @file{f/BUGS}.
3160 Generate better code for @samp{**} operator with a right-hand operand of
3161 type @code{INTEGER}.
3164 Generate better code for @code{SQRT()} and @code{DSQRT()},
3165 also when @option{-ffast-math}
3166 specified, enable better code generation for @code{SIN()} and @code{COS()}.
3169 Generate better code for some kinds of array references.
3172 Speed up lexing somewhat (this makes the compilation phase noticeably