1 NOTE: This file is automatically generated from the files `news0.texi'
2 and `news.texi'. `NEWS' is *not* a source file, although it is
3 normally included within source distributions.
5 This file lists recent changes to the GNU Fortran compiler.
6 Copyright (C) 1995-1999 Free Software Foundation, Inc. You may copy,
7 distribute, and modify it freely as long as you preserve this copyright
8 notice and permission notice.
10 News About GNU Fortran
11 **********************
13 Changes made to recent versions of GNU Fortran are listed below,
14 with the most recent version first.
16 The changes are generally listed in order:
18 1. Code-generation and run-time-library bug-fixes
20 2. Compiler and run-time-library crashes involving valid code that
25 4. Fixes and enhancements to existing features
29 6. Internal improvements
33 This order is not strict--for example, some items involve a
34 combination of these elements.
36 Note that two variants of `g77' are tracked below. The `egcs'
37 variant is described vis-a-vis previous versions of `egcs' and/or an
38 official FSF version, as appropriate.
40 Therefore, `egcs' versions sometimes have multiple listings to help
41 clarify how they differ from other versions, though this can make
42 getting a complete picture of what a particular `egcs' version contains
43 somewhat more difficult.
48 * Fix `g77' so it no longer crashes when compiling I/O statements
49 using keywords that define `INTEGER' values, such as `IOSTAT=J',
50 where J is other than default `INTEGER' (such as `INTEGER*2').
51 Instead, it issues a diagnostic.
53 * The `-ax' option is now obeyed when compiling Fortran programs.
54 (It is passed to the `f771' driver.)
56 * Source file names with the suffixes `.FOR' and `.FPP' now are
57 recognized by `g77' as if they ended in `.for' and `.fpp',
60 * `g77' now warns about a reference to a function when the
61 corresponding *subsequent* function program unit disagrees with
62 the reference concerning the type of the function.
64 * Improve documentation and indexing, including information on Year
65 2000 (Y2K) compliance.
67 In 0.5.24 and `egcs' 1.1.2 (versus 0.5.23 and 1.1.1):
68 =====================================================
70 * Fix the `IDate' Intrinsic (VXT) so the returned year is in the
71 documented, non-Y2K-compliant range of 0-99, instead of being
72 returned as 100 in the year 2000.
74 *Note IDate Intrinsic (VXT)::, for more information.
76 * Fix the `Date_and_Time' intrinsic (in `libg2c') to return the
77 milliseconds value properly in VALUES(8).
79 * Fix the `LStat' intrinsic (in `libg2c') to return device-ID
80 information properly in SARRAY(7).
82 * Improve documentation.
84 In 0.5.24 and `egcs' 1.1.1 (versus 0.5.23 and 1.1):
85 ===================================================
87 * Fix `libg2c' so it performs an implicit `ENDFILE' operation (as
88 appropriate) whenever a `REWIND' is done.
90 (This bug was introduced in 0.5.23 and `egcs' 1.1 in `g77''s
93 * Fix `libg2c' so it no longer crashes with a spurious diagnostic
94 upon doing any I/O following a direct formatted write.
96 (This bug was introduced in 0.5.23 and `egcs' 1.1 in `g77''s
99 * Fix `g77' so it no longer crashes compiling references to the
100 `Rand' intrinsic on some systems.
102 * Fix `g77' portion of installation process so it works better on
103 some systems (those with shells requiring `else true' clauses on
104 `if' constructs for the completion code to be set properly).
106 In `egcs' 1.1 (versus 0.5.24):
107 ==============================
109 * Fix `g77' crash compiling code containing the construct
110 `CMPLX(0.)' or similar.
112 * Fix `g77' crash (or apparently infinite run-time) when compiling
113 certain complicated expressions involving `COMPLEX' arithmetic
114 (especially multiplication).
116 * Fix a code-generation bug that afflicted Intel x86 targets when
117 `-O2' was specified compiling, for example, an old version of the
120 The x87 coprocessor stack was being mismanaged in cases involving
121 assigned `GOTO' and `ASSIGN'.
123 * Align static double-precision variables and arrays on Intel x86
124 targets regardless of whether `-malign-double' is specified.
126 Generally, this affects only local variables and arrays having the
127 `SAVE' attribute or given initial values via `DATA'.
129 In `egcs' 1.1 (versus `egcs' 1.0.3):
130 ====================================
132 * Fix bugs in the `libU77' intrinsic `HostNm' that wrote one byte
133 beyond the end of its `CHARACTER' argument, and in the `libU77'
134 intrinsics `GMTime' and `LTime' that overwrote their arguments.
136 * Assumed arrays with negative bounds (such as `REAL A(-1:*)') no
137 longer elicit spurious diagnostics from `g77', even on systems
138 with pointers having different sizes than integers.
140 This bug is not known to have existed in any recent version of
141 `gcc'. It was introduced in an early release of `egcs'.
143 * Valid combinations of `EXTERNAL', passing that external as a dummy
144 argument without explicitly giving it a type, and, in a subsequent
145 program unit, referencing that external as an external function
146 with a different type no longer crash `g77'.
148 * `CASE DEFAULT' no longer crashes `g77'.
150 * The `-Wunused' option no longer issues a spurious warning about
151 the "master" procedure generated by `g77' for procedures
152 containing `ENTRY' statements.
154 * Support `FORMAT(I<EXPR>)' when EXPR is a compile-time constant
155 `INTEGER' expression.
157 * Fix `g77' `-g' option so procedures that use `ENTRY' can be
158 stepped through, line by line, in `gdb'.
160 * Allow any `REAL' argument to intrinsics `Second' and `CPU_Time'.
162 * Use `tempnam', if available, to open scratch files (as in
163 `OPEN(STATUS='SCRATCH')') so that the `TMPDIR' environment
164 variable, if present, is used.
166 * `g77''s version of `libf2c' separates out the setting of global
167 state (such as command-line arguments and signal handling) from
168 `main.o' into distinct, new library archive members.
170 This should make it easier to write portable applications that
171 have their own (non-Fortran) `main()' routine properly set up the
172 `libf2c' environment, even when `libf2c' (now `libg2c') is a
175 * `g77' no longer installs the `f77' command and `f77.1' man page in
176 the `/usr' or `/usr/local' heirarchy, even if the `f77-install-ok'
177 file exists in the source or build directory. See the
178 installation documentation for more information.
180 * `g77' no longer installs the `libf2c.a' library and `f2c.h'
181 include file in the `/usr' or `/usr/local' heirarchy, even if the
182 `f2c-install-ok' or `f2c-exists-ok' files exist in the source or
183 build directory. See the installation documentation for more
186 * The `libf2c.a' library produced by `g77' has been renamed to
187 `libg2c.a'. It is installed only in the `gcc' "private" directory
188 heirarchy, `gcc-lib'. This allows system administrators and users
189 to choose which version of the `libf2c' library from `netlib' they
190 wish to use on a case-by-case basis. See the installation
191 documentation for more information.
193 * The `f2c.h' include (header) file produced by `g77' has been
194 renamed to `g2c.h'. It is installed only in the `gcc' "private"
195 directory heirarchy, `gcc-lib'. This allows system administrators
196 and users to choose which version of the include file from
197 `netlib' they wish to use on a case-by-case basis. See the
198 installation documentation for more information.
200 * The `g77' command now expects the run-time library to be named
201 `libg2c.a' instead of `libf2c.a', to ensure that a version other
202 than the one built and installed as part of the same `g77' version
205 * During the configuration and build process, `g77' creates
206 subdirectories it needs only as it needs them. Other cleaning up
207 of the configuration and build process has been performed as well.
209 * `install-info' now used to update the directory of Info
210 documentation to contain an entry for `g77' (during installation).
212 * Some diagnostics have been changed from warnings to errors, to
213 prevent inadvertent use of the resulting, probably buggy, programs.
214 These mostly include diagnostics about use of unsupported features
215 in the `OPEN', `INQUIRE', `READ', and `WRITE' statements, and
216 about truncations of various sorts of constants.
218 * Improve compilation of `FORMAT' expressions so that a null byte is
219 appended to the last operand if it is a constant. This provides a
220 cleaner run-time diagnostic as provided by `libf2c' for statements
221 like `PRINT '(I1', 42'.
223 * Improve documentation and indexing.
225 * The upgrade to `libf2c' as of 1998-06-18 should fix a variety of
226 problems, including those involving some uses of the `T' format
227 specifier, and perhaps some build (porting) problems as well.
229 In 0.5.24 and `egcs' 1.1 (versus 0.5.23):
230 =========================================
232 * `g77' no longer produces incorrect code and initial values for
233 `EQUIVALENCE' and `COMMON' aggregates that, due to "unnatural"
234 ordering of members vis-a-vis their types, require initial padding.
236 * `g77' no longer crashes when compiling code containing
237 specification statements such as `INTEGER(KIND=7) PTR'.
239 * `g77' no longer crashes when compiling code such as `J = SIGNAL(1,
242 * `g77' now treats `%LOC(EXPR)' and `LOC(EXPR)' as "ordinary"
243 expressions when they are used as arguments in procedure calls.
244 This change applies only to global (filewide) analysis, making it
245 consistent with how `g77' actually generates code for these cases.
247 Previously, `g77' treated these expressions as denoting special
248 "pointer" arguments for the purposes of filewide analysis.
250 * The `g77' driver now ensures that `-lg2c' is specified in the link
251 phase prior to any occurrence of `-lm'. This prevents
252 accidentally linking to a routine in the SunOS4 `-lm' library when
253 the generated code wants to link to the one in `libf2c' (`libg2c').
255 * `g77' emits more debugging information when `-g' is used.
257 This new information allows, for example, `which __g77_length_a'
258 to be used in `gdb' to determine the type of the phantom length
259 argument supplied with `CHARACTER' variables.
261 This information pertains to internally-generated type, variable,
262 and other information, not to the longstanding deficiencies
263 vis-a-vis `COMMON' and `EQUIVALENCE'.
265 * The F90 `Date_and_Time' intrinsic now is supported.
267 * The F90 `System_Clock' intrinsic allows the optional arguments
268 (except for the `Count' argument) to be omitted.
270 * Upgrade to `libf2c' as of 1998-06-18.
272 * Improve documentation and indexing.
274 In 0.5.23 (versus 0.5.22):
275 ==========================
277 * This release contains several regressions against version 0.5.22
278 of `g77', due to using the "vanilla" `gcc' back end instead of
279 patching it to fix a few bugs and improve performance in a few
282 *Note Actual Bugs We Haven't Fixed Yet: Actual Bugs, available in
283 plain-text format in `gcc/f/BUGS', for information on the known
284 bugs in this version, including the regressions.
286 Features that have been dropped from this version of `g77' due to
287 their being implemented via `g77'-specific patches to the `gcc'
288 back end in previous releases include:
290 - Support for `__restrict__' keyword, the options
291 `-fargument-alias', `-fargument-noalias', and
292 `-fargument-noalias-global', and the corresponding
295 (`egcs' has the alias-analysis code, but not the
296 `__restrict__' keyword. `egcs' `g77' users benefit from the
297 alias-analysis code despite the lack of the `__restrict__'
298 keyword, which is a C-language construct.)
300 - Support for the GNU compiler options `-fmove-all-movables',
301 `-freduce-all-givs', and `-frerun-loop-opt'.
303 (`egcs' supports these options. `g77' users of `egcs'
304 benefit from them even if they are not explicitly specified,
305 because the defaults are optimized for `g77' users.)
307 - Support for the `-W' option warning about integer division by
310 - The Intel x86-specific option `-malign-double' applying to
311 stack-allocated data as well as statically-allocate data.
313 Note that the `gcc/f/gbe/' subdirectory has been removed from this
314 distribution as a result of `g77' no longer including patches for
317 * Fix bugs in the `libU77' intrinsic `HostNm' that wrote one byte
318 beyond the end of its `CHARACTER' argument, and in the `libU77'
319 intrinsics `GMTime' and `LTime' that overwrote their arguments.
321 * Support `gcc' version 2.8, and remove support for prior versions
324 * Remove support for the `--driver' option, as `g77' now does all
325 the driving, just like `gcc'.
327 * `CASE DEFAULT' no longer crashes `g77'.
329 * Valid combinations of `EXTERNAL', passing that external as a dummy
330 argument without explicitly giving it a type, and, in a subsequent
331 program unit, referencing that external as an external function
332 with a different type no longer crash `g77'.
334 * `g77' no longer installs the `f77' command and `f77.1' man page in
335 the `/usr' or `/usr/local' heirarchy, even if the `f77-install-ok'
336 file exists in the source or build directory. See the
337 installation documentation for more information.
339 * `g77' no longer installs the `libf2c.a' library and `f2c.h'
340 include file in the `/usr' or `/usr/local' heirarchy, even if the
341 `f2c-install-ok' or `f2c-exists-ok' files exist in the source or
342 build directory. See the installation documentation for more
345 * The `libf2c.a' library produced by `g77' has been renamed to
346 `libg2c.a'. It is installed only in the `gcc' "private" directory
347 heirarchy, `gcc-lib'. This allows system administrators and users
348 to choose which version of the `libf2c' library from `netlib' they
349 wish to use on a case-by-case basis. See the installation
350 documentation for more information.
352 * The `f2c.h' include (header) file produced by `g77' has been
353 renamed to `g2c.h'. It is installed only in the `gcc' "private"
354 directory heirarchy, `gcc-lib'. This allows system administrators
355 and users to choose which version of the include file from
356 `netlib' they wish to use on a case-by-case basis. See the
357 installation documentation for more information.
359 * The `g77' command now expects the run-time library to be named
360 `libg2c.a' instead of `libf2c.a', to ensure that a version other
361 than the one built and installed as part of the same `g77' version
364 * The `-Wunused' option no longer issues a spurious warning about
365 the "master" procedure generated by `g77' for procedures
366 containing `ENTRY' statements.
368 * `g77''s version of `libf2c' separates out the setting of global
369 state (such as command-line arguments and signal handling) from
370 `main.o' into distinct, new library archive members.
372 This should make it easier to write portable applications that
373 have their own (non-Fortran) `main()' routine properly set up the
374 `libf2c' environment, even when `libf2c' (now `libg2c') is a
377 * During the configuration and build process, `g77' creates
378 subdirectories it needs only as it needs them, thus avoiding
379 unnecessary creation of, for example, `stage1/f/runtime' when
380 doing a non-bootstrap build. Other cleaning up of the
381 configuration and build process has been performed as well.
383 * `install-info' now used to update the directory of Info
384 documentation to contain an entry for `g77' (during installation).
386 * Some diagnostics have been changed from warnings to errors, to
387 prevent inadvertent use of the resulting, probably buggy, programs.
388 These mostly include diagnostics about use of unsupported features
389 in the `OPEN', `INQUIRE', `READ', and `WRITE' statements, and
390 about truncations of various sorts of constants.
392 * Improve documentation and indexing.
394 * Upgrade to `libf2c' as of 1998-04-20.
396 This should fix a variety of problems, including those involving
397 some uses of the `T' format specifier, and perhaps some build
398 (porting) problems as well.
400 In 0.5.22 (versus 0.5.21):
401 ==========================
403 * Fix code generation for iterative `DO' loops that have one or more
404 references to the iteration variable, or to aliases of it, in
405 their control expressions. For example, `DO 10 J=2,J' now is
408 * Fix a code-generation bug that afflicted Intel x86 targets when
409 `-O2' was specified compiling, for example, an old version of the
412 The x87 coprocessor stack was being mismanaged in cases involving
413 assigned `GOTO' and `ASSIGN'.
415 * Fix `DTime' intrinsic so as not to truncate results to integer
416 values (on some systems).
418 * Fix `Signal' intrinsic so it offers portable support for 64-bit
419 systems (such as Digital Alphas running GNU/Linux).
421 * Fix run-time crash involving `NAMELIST' on 64-bit machines such as
424 * Fix `g77' version of `libf2c' so it no longer produces a spurious
425 `I/O recursion' diagnostic at run time when an I/O operation (such
426 as `READ *,I') is interrupted in a manner that causes the program
427 to be terminated via the `f_exit' routine (such as via `C-c').
429 * Fix `g77' crash triggered by `CASE' statement with an omitted
430 lower or upper bound.
432 * Fix `g77' crash compiling references to `CPU_Time' intrinsic.
434 * Fix `g77' crash (or apparently infinite run-time) when compiling
435 certain complicated expressions involving `COMPLEX' arithmetic
436 (especially multiplication).
438 * Fix `g77' crash on statements such as `PRINT *,
439 (REAL(Z(I)),I=1,2)', where `Z' is `DOUBLE COMPLEX'.
443 * Support `FORMAT(I<EXPR>)' when EXPR is a compile-time constant
444 `INTEGER' expression.
446 * Fix `g77' `-g' option so procedures that use `ENTRY' can be
447 stepped through, line by line, in `gdb'.
449 * Fix a profiling-related bug in `gcc' back end for Intel x86
452 * Allow any `REAL' argument to intrinsics `Second' and `CPU_Time'.
454 * Allow any numeric argument to intrinsics `Int2' and `Int8'.
456 * Use `tempnam', if available, to open scratch files (as in
457 `OPEN(STATUS='SCRATCH')') so that the `TMPDIR' environment
458 variable, if present, is used.
460 * Rename the `gcc' keyword `restrict' to `__restrict__', to avoid
461 rejecting valid, existing, C programs. Support for `restrict' is
462 now more like support for `complex'.
464 * Fix `-fpedantic' to not reject procedure invocations such as
465 `I=J()' and `CALL FOO()'.
467 * Fix `-fugly-comma' to affect invocations of only external
468 procedures. Restore rejection of gratuitous trailing omitted
469 arguments to intrinsics, as in `I=MAX(3,4,,)'.
471 * Fix compiler so it accepts `-fgnu-intrinsics-*' and
472 `-fbadu77-intrinsics-*' options.
474 * Improve diagnostic messages from `libf2c' so it is more likely
475 that the printing of the active format string is limited to the
476 string, with no trailing garbage being printed.
478 (Unlike `f2c', `g77' did not append a null byte to its compiled
479 form of every format string specified via a `FORMAT' statement.
480 However, `f2c' would exhibit the problem anyway for a statement
481 like `PRINT '(I)garbage', 1' by printing `(I)garbage' as the
484 * Improve compilation of `FORMAT' expressions so that a null byte is
485 appended to the last operand if it is a constant. This provides a
486 cleaner run-time diagnostic as provided by `libf2c' for statements
487 like `PRINT '(I1', 42'.
489 * Fix various crashes involving code with diagnosed errors.
491 * Fix cross-compilation bug when configuring `libf2c'.
493 * Improve diagnostics.
495 * Improve documentation and indexing.
497 * Upgrade to `libf2c' as of 1997-09-23. This fixes a formatted-I/O
498 bug that afflicted 64-bit systems with 32-bit integers (such as
499 Digital Alpha running GNU/Linux).
501 In `egcs' 1.0.2 (versus `egcs' 1.0.1):
502 ======================================
504 * Fix `g77' crash triggered by `CASE' statement with an omitted
505 lower or upper bound.
507 * Fix `g77' crash on statements such as `PRINT *,
508 (REAL(Z(I)),I=1,2)', where `Z' is `DOUBLE COMPLEX'.
510 * Fix `-fPIC' (such as compiling for ELF targets) on the Intel x86
511 architecture target so invalid assembler code is no longer
514 * Fix `-fpedantic' to not reject procedure invocations such as
515 `I=J()' and `CALL FOO()'.
517 * Fix `-fugly-comma' to affect invocations of only external
518 procedures. Restore rejection of gratuitous trailing omitted
519 arguments to intrinsics, as in `I=MAX(3,4,,)'.
521 * Fix compiler so it accepts `-fgnu-intrinsics-*' and
522 `-fbadu77-intrinsics-*' options.
524 In `egcs' 1.0.1 (versus `egcs' 1.0):
525 ====================================
527 * Fix run-time crash involving `NAMELIST' on 64-bit machines such as
530 In `egcs' 1.0 (versus 0.5.21):
531 ==============================
533 * Version 1.0 of `egcs' contains several regressions against version
534 0.5.21 of `g77', due to using the "vanilla" `gcc' back end instead
535 of patching it to fix a few bugs and improve performance in a few
538 *Note Actual Bugs We Haven't Fixed Yet: Actual Bugs, available in
539 plain-text format in `gcc/f/BUGS', for information on the known
540 bugs in this version, including the regressions.
542 Features that have been dropped from this version of `g77' due to
543 their being implemented via `g77'-specific patches to the `gcc'
544 back end in previous releases include:
546 - Support for the C-language `restrict' keyword.
548 - Support for the `-W' option warning about integer division by
551 - The Intel x86-specific option `-malign-double' applying to
552 stack-allocated data as well as statically-allocate data.
554 Note that the `gcc/f/gbe/' subdirectory has been removed from this
555 distribution as a result of `g77' being fully integrated with the
556 `egcs' variant of the `gcc' back end.
558 * Fix code generation for iterative `DO' loops that have one or more
559 references to the iteration variable, or to aliases of it, in
560 their control expressions. For example, `DO 10 J=2,J' now is
563 * Fix `DTime' intrinsic so as not to truncate results to integer
564 values (on some systems).
566 * Some Fortran code, miscompiled by `g77' built on `gcc' version
567 2.8.1 on m68k-next-nextstep3 configurations when using the `-O2'
568 option, is now compiled correctly. It is believed that a C
569 function known to miscompile on that configuration when using the
570 `-O2 -funroll-loops' options also is now compiled correctly.
572 * Remove support for non-`egcs' versions of `gcc'.
574 * Remove support for the `--driver' option, as `g77' now does all
575 the driving, just like `gcc'.
577 * Allow any numeric argument to intrinsics `Int2' and `Int8'.
579 * Improve diagnostic messages from `libf2c' so it is more likely
580 that the printing of the active format string is limited to the
581 string, with no trailing garbage being printed.
583 (Unlike `f2c', `g77' did not append a null byte to its compiled
584 form of every format string specified via a `FORMAT' statement.
585 However, `f2c' would exhibit the problem anyway for a statement
586 like `PRINT '(I)garbage', 1' by printing `(I)garbage' as the
589 * Upgrade to `libf2c' as of 1997-09-23. This fixes a formatted-I/O
590 bug that afflicted 64-bit systems with 32-bit integers (such as
591 Digital Alpha running GNU/Linux).
596 * Fix a code-generation bug introduced by 0.5.20 caused by loop
597 unrolling (by specifying `-funroll-loops' or similar). This bug
598 afflicted all code compiled by version 2.7.2.2.f.2 of `gcc' (C,
599 C++, Fortran, and so on).
601 * Fix a code-generation bug manifested when combining local
602 `EQUIVALENCE' with a `DATA' statement that follows the first
603 executable statement (or is treated as an executable-context
604 statement as a result of using the `-fpedantic' option).
606 * Fix a compiler crash that occured when an integer division by a
607 constant zero is detected. Instead, when the `-W' option is
608 specified, the `gcc' back end issues a warning about such a case.
609 This bug afflicted all code compiled by version 2.7.2.2.f.2 of
610 `gcc' (C, C++, Fortran, and so on).
612 * Fix a compiler crash that occurred in some cases of procedure
613 inlining. (Such cases became more frequent in 0.5.20.)
615 * Fix a compiler crash resulting from using `DATA' or similar to
616 initialize a `COMPLEX' variable or array to zero.
618 * Fix compiler crashes involving use of `AND', `OR', or `XOR'
621 * Fix compiler bug triggered when using a `COMMON' or `EQUIVALENCE'
622 variable as the target of an `ASSIGN' or assigned-`GOTO' statement.
624 * Fix compiler crashes due to using the name of a some non-standard
625 intrinsics (such as `FTELL' or `FPUTC') as such and as the name of
626 a procedure or common block. Such dual use of a name in a program
627 is allowed by the standard.
629 * Place automatic arrays on the stack, even if `SAVE' or the
630 `-fno-automatic' option is in effect. This avoids a compiler
633 * The `-malign-double' option now reliably aligns `DOUBLE PRECISION'
634 optimally on Pentium and Pentium Pro architectures (586 and 686 in
637 * New option `-Wno-globals' disables warnings about "suspicious" use
638 of a name both as a global name and as the implicit name of an
639 intrinsic, and warnings about disagreements over the number or
640 natures of arguments passed to global procedures, or the natures
641 of the procedures themselves.
643 The default is to issue such warnings, which are new as of this
646 * New option `-fno-globals' disables diagnostics about potentially
647 fatal disagreements analysis problems, such as disagreements over
648 the number or natures of arguments passed to global procedures, or
649 the natures of those procedures themselves.
651 The default is to issue such diagnostics and flag the compilation
652 as unsuccessful. With this option, the diagnostics are issued as
653 warnings, or, if `-Wno-globals' is specified, are not issued at
656 This option also disables inlining of global procedures, to avoid
657 compiler crashes resulting from coding errors that these
658 diagnostics normally would identify.
660 * Diagnose cases where a reference to a procedure disagrees with the
661 type of that procedure, or where disagreements about the number or
662 nature of arguments exist. This avoids a compiler crash.
664 * Fix parsing bug whereby `g77' rejected a second initialization
665 specification immediately following the first's closing `/' without
666 an intervening comma in a `DATA' statement, and the second
667 specification was an implied-DO list.
669 * Improve performance of the `gcc' back end so certain complicated
670 expressions involving `COMPLEX' arithmetic (especially
671 multiplication) don't appear to take forever to compile.
673 * Fix a couple of profiling-related bugs in `gcc' back end.
675 * Integrate GNU Ada's (GNAT's) changes to the back end, which
676 consist almost entirely of bug fixes. These fixes are circa
677 version 3.10p of GNAT.
679 * Include some other `gcc' fixes that seem useful in `g77''s version
680 of `gcc'. (See `gcc/ChangeLog' for details--compare it to that
681 file in the vanilla `gcc-2.7.2.3.tar.gz' distribution.)
683 * Fix `libU77' routines that accept file and other names to strip
684 trailing blanks from them, for consistency with other
685 implementations. Blanks may be forcibly appended to such names by
686 appending a single null character (`CHAR(0)') to the significant
689 * Fix `CHMOD' intrinsic to work with file names that have embedded
690 blanks, commas, and so on.
692 * Fix `SIGNAL' intrinsic so it accepts an optional third `Status'
695 * Fix `IDATE()' intrinsic subroutine (VXT form) so it accepts
696 arguments in the correct order. Documentation fixed accordingly,
697 and for `GMTIME()' and `LTIME()' as well.
699 * Make many changes to `libU77' intrinsics to support existing code
702 Such changes include allowing both subroutine and function forms
703 of many routines, changing `MCLOCK()' and `TIME()' to return
704 `INTEGER(KIND=1)' values, introducing `MCLOCK8()' and `TIME8()' to
705 return `INTEGER(KIND=2)' values, and placing functions that are
706 intended to perform side effects in a new intrinsic group,
709 * Improve `libU77' so it is more portable.
711 * Add options `-fbadu77-intrinsics-delete',
712 `-fbadu77-intrinsics-hide', and so on.
714 * Fix crashes involving diagnosed or invalid code.
716 * `g77' and `gcc' now do a somewhat better job detecting and
717 diagnosing arrays that are too large to handle before these cause
718 diagnostics during the assembler or linker phase, a compiler
719 crash, or generation of incorrect code.
721 * Make some fixes to alias analysis code.
723 * Add support for `restrict' keyword in `gcc' front end.
725 * Support `gcc' version 2.7.2.3 (modified by `g77' into version
726 2.7.2.3.f.1), and remove support for prior versions of `gcc'.
728 * Incorporate GNAT's patches to the `gcc' back end into `g77''s, so
729 GNAT users do not need to apply GNAT's patches to build both GNAT
730 and `g77' from the same source tree.
732 * Modify `make' rules and related code so that generation of Info
733 documentation doesn't require compilation using `gcc'. Now, any
734 ANSI C compiler should be adequate to produce the `g77'
735 documentation (in particular, the tables of intrinsics) from
738 * Add `INT2' and `INT8' intrinsics.
740 * Add `CPU_TIME' intrinsic.
742 * Add `ALARM' intrinsic.
744 * `CTIME' intrinsic now accepts any `INTEGER' argument, not just
747 * Warn when explicit type declaration disagrees with the type of an
748 intrinsic invocation.
750 * Support `*f771' entry in `gcc' `specs' file.
752 * Fix typo in `make' rule `g77-cross', used only for cross-compiling.
754 * Fix `libf2c' build procedure to re-archive library if previous
755 attempt to archive was interrupted.
757 * Change `gcc' to unroll loops only during the last invocation (of
758 as many as two invocations) of loop optimization.
760 * Improve handling of `-fno-f2c' so that code that attempts to pass
761 an intrinsic as an actual argument, such as `CALL FOO(ABS)', is
762 rejected due to the fact that the run-time-library routine is,
763 effectively, compiled with `-ff2c' in effect.
765 * Fix `g77' driver to recognize `-fsyntax-only' as an option that
766 inhibits linking, just like `-c' or `-S', and to recognize and
767 properly handle the `-nostdlib', `-M', `-MM', `-nodefaultlibs',
768 and `-Xlinker' options.
770 * Upgrade to `libf2c' as of 1997-08-16.
772 * Modify `libf2c' to consistently and clearly diagnose recursive I/O
775 * `g77' driver now prints version information (such as produced by
776 `g77 -v') to `stderr' instead of `stdout'.
778 * The `.r' suffix now designates a Ratfor source file, to be
779 preprocessed via the `ratfor' command, available separately.
781 * Fix some aspects of how `gcc' determines what kind of system is
782 being configured and what kinds are supported. For example, GNU
783 Linux/Alpha ELF systems now are directly supported.
785 * Improve diagnostics.
787 * Improve documentation and indexing.
789 * Include all pertinent files for `libf2c' that come from
790 `netlib.bell-labs.com'; give any such files that aren't quite
791 accurate in `g77''s version of `libf2c' the suffix `.netlib'.
793 * Reserve `INTEGER(KIND=0)' for future use.
798 * The `-fno-typeless-boz' option is now the default.
800 This option specifies that non-decimal-radix constants using the
801 prefixed-radix form (such as `Z'1234'') are to be interpreted as
802 `INTEGER' constants. Specify `-ftypeless-boz' to cause such
803 constants to be interpreted as typeless.
805 (Version 0.5.19 introduced `-fno-typeless-boz' and its inverse.)
807 * Options `-ff90-intrinsics-enable' and `-fvxt-intrinsics-enable'
808 now are the defaults.
810 Some programs might use names that clash with intrinsic names
811 defined (and now enabled) by these options or by the new `libU77'
812 intrinsics. Users of such programs might need to compile them
813 differently (using, for example, `-ff90-intrinsics-disable') or,
814 better yet, insert appropriate `EXTERNAL' statements specifying
815 that these names are not intended to be names of intrinsics.
817 * The `ALWAYS_FLUSH' macro is no longer defined when building
818 `libf2c', which should result in improved I/O performance,
821 *Note:* If you have code that depends on the behavior of `libf2c'
822 when built with `ALWAYS_FLUSH' defined, you will have to modify
823 `libf2c' accordingly before building it from this and future
826 * Dave Love's implementation of `libU77' has been added to the
827 version of `libf2c' distributed with and built as part of `g77'.
828 `g77' now knows about the routines in this library as intrinsics.
830 * New option `-fvxt' specifies that the source file is written in
831 VXT Fortran, instead of GNU Fortran.
833 * The `-fvxt-not-f90' option has been deleted, along with its
834 inverse, `-ff90-not-vxt'.
836 If you used one of these deleted options, you should re-read the
837 pertinent documentation to determine which options, if any, are
838 appropriate for compiling your code with this version of `g77'.
840 * The `-fugly' option now issues a warning, as it likely will be
841 removed in a future version.
843 (Enabling all the `-fugly-*' options is unlikely to be feasible,
844 or sensible, in the future, so users should learn to specify only
845 those `-fugly-*' options they really need for a particular source
848 * The `-fugly-assumed' option, introduced in version 0.5.19, has
849 been changed to better accommodate old and new code.
851 * Make a number of fixes to the `g77' front end and the `gcc' back
852 end to better support Alpha (AXP) machines. This includes
853 providing at least one bug-fix to the `gcc' back end for Alphas.
855 * Related to supporting Alpha (AXP) machines, the `LOC()' intrinsic
856 and `%LOC()' construct now return values of integer type that is
857 the same width (holds the same number of bits) as the pointer type
860 On most machines, this won't make a difference, whereas on Alphas,
861 the type these constructs return is `INTEGER*8' instead of the
862 more common `INTEGER*4'.
864 * Emulate `COMPLEX' arithmetic in the `g77' front end, to avoid bugs
865 in `complex' support in the `gcc' back end. New option
866 `-fno-emulate-complex' causes `g77' to revert the 0.5.19 behavior.
868 * Fix bug whereby `REAL A(1)', for example, caused a compiler crash
869 if `-fugly-assumed' was in effect and A was a local (automatic)
870 array. That case is no longer affected by the new handling of
873 * Fix `g77' command driver so that `g77 -o foo.f' no longer deletes
874 `foo.f' before issuing other diagnostics, and so the `-x' option
877 * Enable inlining of subroutines and functions by the `gcc' back end.
878 This works as it does for `gcc' itself--program units may be
879 inlined for invocations that follow them in the same program unit,
880 as long as the appropriate compile-time options are specified.
882 * Dummy arguments are no longer assumed to potentially alias
883 (overlap) other dummy arguments or `COMMON' areas when any of
884 these are defined (assigned to) by Fortran code.
886 This can result in faster and/or smaller programs when compiling
887 with optimization enabled, though on some systems this effect is
888 observed only when `-fforce-addr' also is specified.
890 New options `-falias-check', `-fargument-alias',
891 `-fargument-noalias', and `-fno-argument-noalias-global' control
892 the way `g77' handles potential aliasing.
894 * The `CONJG()' and `DCONJG()' intrinsics now are compiled in-line.
896 * The bug-fix for 0.5.19.1 has been re-done. The `g77' compiler has
897 been changed back to assume `libf2c' has no aliasing problems in
898 its implementations of the `COMPLEX' (and `DOUBLE COMPLEX')
899 intrinsics. The `libf2c' has been changed to have no such
902 As a result, 0.5.20 is expected to offer improved performance over
903 0.5.19.1, perhaps as good as 0.5.19 in most or all cases, due to
906 *Note:* This change requires version 0.5.20 of `libf2c', at least,
907 when linking code produced by any versions of `g77' other than
908 0.5.19.1. Use `g77 -v' to determine the version numbers of the
909 `libF77', `libI77', and `libU77' components of the `libf2c'
910 library. (If these version numbers are not printed--in
911 particular, if the linker complains about unresolved references to
912 names like `g77__fvers__'--that strongly suggests your
913 installation has an obsolete version of `libf2c'.)
915 * New option `-fugly-assign' specifies that the same memory
916 locations are to be used to hold the values assigned by both
917 statements `I = 3' and `ASSIGN 10 TO I', for example. (Normally,
918 `g77' uses a separate memory location to hold assigned statement
921 * `FORMAT' and `ENTRY' statements now are allowed to precede
922 `IMPLICIT NONE' statements.
924 * Produce diagnostic for unsupported `SELECT CASE' on `CHARACTER'
925 type, instead of crashing, at compile time.
927 * Fix crashes involving diagnosed or invalid code.
929 * Change approach to building `libf2c' archive (`libf2c.a') so that
930 members are added to it only when truly necessary, so the user
931 that installs an already-built `g77' doesn't need to have write
932 access to the build tree (whereas the user doing the build might
933 not have access to install new software on the system).
935 * Support `gcc' version 2.7.2.2 (modified by `g77' into version
936 2.7.2.2.f.2), and remove support for prior versions of `gcc'.
938 * Upgrade to `libf2c' as of 1997-02-08, and fix up some of the build
941 * Improve general build procedures for `g77', fixing minor bugs
942 (such as deletion of any file named `f771' in the parent directory
945 * Enable full support of `INTEGER*8' available in `libf2c' and
946 `f2c.h' so that `f2c' users may make full use of its features via
947 the `g77' version of `f2c.h' and the `INTEGER*8' support routines
948 in the `g77' version of `libf2c'.
950 * Improve `g77' driver and `libf2c' so that `g77 -v' yields version
951 information on the library.
953 * The `SNGL' and `FLOAT' intrinsics now are specific intrinsics,
954 instead of synonyms for the generic intrinsic `REAL'.
956 * New intrinsics have been added. These are `REALPART', `IMAGPART',
957 `COMPLEX', `LONG', and `SHORT'.
959 * A new group of intrinsics, `gnu', has been added to contain the
960 new `REALPART', `IMAGPART', and `COMPLEX' intrinsics. An old
961 group, `dcp', has been removed.
963 * Complain about industry-wide ambiguous references `REAL(EXPR)' and
964 `AIMAG(EXPR)', where EXPR is `DOUBLE COMPLEX' (or any complex type
965 other than `COMPLEX'), unless `-ff90' option specifies Fortran 90
966 interpretation or new `-fugly-complex' option, in conjunction with
967 `-fnot-f90', specifies `f2c' interpretation.
969 * Make improvements to diagnostics.
971 * Speed up compiler a bit.
973 * Improvements to documentation and indexing, including a new
974 chapter containing information on one, later more, diagnostics
975 that users are directed to pull up automatically via a message in
976 the diagnostic itself.
978 (Hence the menu item `M' for the node `Diagnostics' in the
979 top-level menu of the Info documentation.)
984 * Code-generation bugs afflicting operations on complex data have
987 These bugs occurred when assigning the result of an operation to a
988 complex variable (or array element) that also served as an input
991 The operations affected by this bug were: `CONJG()', `DCONJG()',
992 `CCOS()', `CDCOS()', `CLOG()', `CDLOG()', `CSIN()', `CDSIN()',
993 `CSQRT()', `CDSQRT()', complex division, and raising a `DOUBLE
994 COMPLEX' operand to an `INTEGER' power. (The related generic and
995 `Z'-prefixed intrinsics, such as `ZSIN()', also were affected.)
997 For example, `C = CSQRT(C)', `Z = Z/C', and `Z = Z**I' (where `C'
998 is `COMPLEX' and `Z' is `DOUBLE COMPLEX') have been fixed.
1003 * Fix `FORMAT' statement parsing so negative values for specifiers
1004 such as `P' (e.g. `FORMAT(-1PF8.1)') are correctly processed as
1007 * Fix `SIGNAL' intrinsic so it once again accepts a procedure as its
1010 * A temporary kludge option provides bare-bones information on
1011 `COMMON' and `EQUIVALENCE' members at debug time.
1013 * New `-fonetrip' option specifies FORTRAN-66-style one-trip `DO'
1016 * New `-fno-silent' option causes names of program units to be
1017 printed as they are compiled, in a fashion similar to UNIX `f77'
1020 * New `-fugly-assumed' option specifies that arrays dimensioned via
1021 `DIMENSION X(1)', for example, are to be treated as assumed-size.
1023 * New `-fno-typeless-boz' option specifies that non-decimal-radix
1024 constants using the prefixed-radix form (such as `Z'1234'') are to
1025 be interpreted as `INTEGER' constants.
1027 * New `-ff66' option is a "shorthand" option that specifies
1028 behaviors considered appropriate for FORTRAN 66 programs.
1030 * New `-ff77' option is a "shorthand" option that specifies
1031 behaviors considered appropriate for UNIX `f77' programs.
1033 * New `-fugly-comma' and `-fugly-logint' options provided to perform
1034 some of what `-fugly' used to do. `-fugly' and `-fno-ugly' are
1035 now "shorthand" options, in that they do nothing more than enable
1036 (or disable) other `-fugly-*' options.
1038 * Fix parsing of assignment statements involving targets that are
1039 substrings of elements of `CHARACTER' arrays having names such as
1040 `READ', `WRITE', `GOTO', and `REALFUNCTIONFOO'.
1042 * Fix crashes involving diagnosed code.
1044 * Fix handling of local `EQUIVALENCE' areas so certain cases of
1045 valid Fortran programs are not misdiagnosed as improperly
1046 extending the area backwards.
1048 * Support `gcc' version 2.7.2.1.
1050 * Upgrade to `libf2c' as of 1996-09-26, and fix up some of the build
1053 * Change code generation for list-directed I/O so it allows for new
1054 versions of `libf2c' that might return non-zero status codes for
1055 some operations previously assumed to always return zero.
1057 This change not only affects how `IOSTAT=' variables are set by
1058 list-directed I/O, it also affects whether `END=' and `ERR='
1059 labels are reached by these operations.
1061 * Add intrinsic support for new `FTELL' and `FSEEK' procedures in
1064 * Modify `fseek_()' in `libf2c' to be more portable (though, in
1065 practice, there might be no systems where this matters) and to
1066 catch invalid `whence' arguments.
1068 * Some useless warnings from the `-Wunused' option have been
1071 * Fix a problem building the `f771' executable on AIX systems by
1072 linking with the `-bbigtoc' option.
1074 * Abort configuration if `gcc' has not been patched using the patch
1075 file provided in the `gcc/f/gbe/' subdirectory.
1077 * Add options `--help' and `--version' to the `g77' command, to
1078 conform to GNU coding guidelines. Also add printing of `g77'
1079 version number when the `--verbose' (`-v') option is used.
1081 * Change internally generated name for local `EQUIVALENCE' areas to
1082 one based on the alphabetically sorted first name in the list of
1083 names for entities placed at the beginning of the areas.
1085 * Improvements to documentation and indexing.
1090 * Add some rudimentary support for `INTEGER*1', `INTEGER*2',
1091 `INTEGER*8', and their `LOGICAL' equivalents. (This support works
1092 on most, maybe all, `gcc' targets.)
1094 Thanks to Scott Snyder (<snyder@d0sgif.fnal.gov>) for providing
1097 Among the missing elements from the support for these features are
1098 full intrinsic support and constants.
1100 * Add some rudimentary support for the `BYTE' and `WORD'
1101 type-declaration statements. `BYTE' corresponds to `INTEGER*1',
1102 while `WORD' corresponds to `INTEGER*2'.
1104 Thanks to Scott Snyder (<snyder@d0sgif.fnal.gov>) for providing
1107 * The compiler code handling intrinsics has been largely rewritten
1108 to accommodate the new types. No new intrinsics or arguments for
1109 existing intrinsics have been added, so there is, at this point,
1110 no intrinsic to convert to `INTEGER*8', for example.
1112 * Support automatic arrays in procedures.
1114 * Reduce space/time requirements for handling large *sparsely*
1115 initialized aggregate arrays. This improvement applies to only a
1116 subset of the general problem to be addressed in 0.6.
1118 * Treat initial values of zero as if they weren't specified (in DATA
1119 and type-declaration statements). The initial values will be set
1120 to zero anyway, but the amount of compile time processing them
1121 will be reduced, in some cases significantly (though, again, this
1122 is only a subset of the general problem to be addressed in 0.6).
1124 A new option, `-fzeros', is introduced to enable the traditional
1125 treatment of zeros as any other value.
1127 * With `-ff90' in force, `g77' incorrectly interpreted `REAL(Z)' as
1128 returning a `REAL' result, instead of as a `DOUBLE PRECISION'
1129 result. (Here, `Z' is `DOUBLE COMPLEX'.)
1131 With `-fno-f90' in force, the interpretation remains unchanged,
1132 since this appears to be how at least some F77 code using the
1133 `DOUBLE COMPLEX' extension expected it to work.
1135 Essentially, `REAL(Z)' in F90 is the same as `DBLE(Z)', while in
1136 extended F77, it appears to be the same as `REAL(REAL(Z))'.
1138 * An expression involving exponentiation, where both operands were
1139 type `INTEGER' and the right-hand operand was negative, was
1140 erroneously evaluated.
1142 * Fix bugs involving `DATA' implied-`DO' constructs (these involved
1143 an errant diagnostic and a crash, both on good code, one involving
1144 subsequent statement-function definition).
1146 * Close `INCLUDE' files after processing them, so compiling source
1147 files with lots of `INCLUDE' statements does not result in being
1148 unable to open `INCLUDE' files after all the available file
1149 descriptors are used up.
1151 * Speed up compiling, especially of larger programs, and perhaps
1152 slightly reduce memory utilization while compiling (this is *not*
1153 the improvement planned for 0.6 involving large aggregate
1154 areas)--these improvements result from simply turning off some
1155 low-level code to do self-checking that hasn't been triggered in a
1158 * Introduce three new options that implement optimizations in the
1159 `gcc' back end (GBE). These options are `-fmove-all-movables',
1160 `-freduce-all-givs', and `-frerun-loop-opt', which are enabled, by
1161 default, for Fortran compilations. These optimizations are
1162 intended to help toon Fortran programs.
1164 * Patch the GBE to do a better job optimizing certain kinds of
1165 references to array elements.
1167 * Due to patches to the GBE, the version number of `gcc' also is
1168 patched to make it easier to manage installations, especially
1169 useful if it turns out a `g77' change to the GBE has a bug.
1171 The `g77'-modified version number is the `gcc' version number with
1172 the string `.f.N' appended, where `f' identifies the version as
1173 enhanced for Fortran, and N is `1' for the first Fortran patch for
1174 that version of `gcc', `2' for the second, and so on.
1176 So, this introduces version 2.7.2.f.1 of `gcc'.
1178 * Make several improvements and fixes to diagnostics, including the
1179 removal of two that were inappropriate or inadequate.
1181 * Warning about two successive arithmetic operators, produced by
1182 `-Wsurprising', now produced *only* when both operators are,
1183 indeed, arithmetic (not relational/boolean).
1185 * `-Wsurprising' now warns about the remaining cases of using
1186 non-integral variables for implied-`DO' loops, instead of these
1187 being rejected unless `-fpedantic' or `-fugly' specified.
1189 * Allow `SAVE' of a local variable or array, even after it has been
1190 given an initial value via `DATA', for example.
1192 * Introduce an Info version of `g77' documentation, which supercedes
1193 `gcc/f/CREDITS', `gcc/f/DOC', and `gcc/f/PROJECTS'. These files
1194 will be removed in a future release. The files `gcc/f/BUGS',
1195 `gcc/f/INSTALL', and `gcc/f/NEWS' now are automatically built from
1196 the texinfo source when distributions are made.
1198 This effort was inspired by a first pass at translating
1199 `g77-0.5.16/f/DOC' that was contributed to Craig by David Ronis
1200 (<ronis@onsager.chem.mcgill.ca>).
1202 * New `-fno-second-underscore' option to specify that, when
1203 `-funderscoring' is in effect, a second underscore is not to be
1204 appended to Fortran names already containing an underscore.
1206 * Change the way iterative `DO' loops work to follow the F90
1207 standard. In particular, calculation of the iteration count is
1208 still done by converting the start, end, and increment parameters
1209 to the type of the `DO' variable, but the result of the
1210 calculation is always converted to the default `INTEGER' type.
1212 (This should have no effect on existing code compiled by `g77',
1213 but code written to assume that use of a *wider* type for the `DO'
1214 variable will result in an iteration count being fully calculated
1215 using that wider type (wider than default `INTEGER') must be
1218 * Support `gcc' version 2.7.2.
1220 * Upgrade to `libf2c' as of 1996-03-23, and fix up some of the build
1223 Note that the email addresses related to `f2c' have changed--the
1224 distribution site now is named `netlib.bell-labs.com', and the
1225 maintainer's new address is <dmg@bell-labs.com>.
1230 * *Fix serious bug* in `g77 -v' command that can cause removal of a
1231 system's `/dev/null' special file if run by user `root'.
1233 *All users* of version 0.5.16 should ensure that they have not
1234 removed `/dev/null' or replaced it with an ordinary file (e.g. by
1235 comparing the output of `ls -l /dev/null' with `ls -l /dev/zero'.
1236 If the output isn't basically the same, contact your system
1237 administrator about restoring `/dev/null' to its proper status).
1239 This bug is particularly insidious because removing `/dev/null' as
1240 a special file can go undetected for quite a while, aside from
1241 various applications and programs exhibiting sudden, strange
1244 I sincerely apologize for not realizing the implications of the
1245 fact that when `g77 -v' runs the `ld' command with `-o /dev/null'
1246 that `ld' tries to *remove* the executable it is supposed to build
1247 (especially if it reports unresolved references, which it should
1250 * Fix crash on `CHARACTER*(*) FOO' in a main or block data program
1253 * Fix crash that can occur when diagnostics given outside of any
1254 program unit (such as when input file contains `@foo').
1256 * Fix crashes, infinite loops (hangs), and such involving diagnosed
1259 * Fix `ASSIGN''ed variables so they can be `SAVE''d or dummy
1260 arguments, and issue clearer error message in cases where target
1261 of `ASSIGN' or `ASSIGN'ed `GOTO'/`FORMAT' is too small (which
1262 should never happen).
1264 * Make `libf2c' build procedures work on more systems again by
1265 eliminating unnecessary invocations of `ld -r -x' and `mv'.
1267 * Fix omission of `-funix-intrinsics-...' options in list of
1268 permitted options to compiler.
1270 * Fix failure to always diagnose missing type declaration for
1273 * Fix compile-time performance problem (which could sometimes crash
1274 the compiler, cause a hang, or whatever, due to a bug in the back
1275 end) involving exponentiation with a large `INTEGER' constant for
1276 the right-hand operator (e.g. `I**32767').
1278 * Fix build procedures so cross-compiling `g77' (the `fini' utility
1279 in particular) is properly built using the host compiler.
1281 * Add new `-Wsurprising' option to warn about constructs that are
1282 interpreted by the Fortran standard (and `g77') in ways that are
1283 surprising to many programmers.
1285 * Add `ERF()' and `ERFC()' as generic intrinsics mapping to existing
1286 `ERF'/`DERF' and `ERFC'/`DERFC' specific intrinsics.
1288 *Note:* You should specify `INTRINSIC ERF,ERFC' in any code where
1289 you might use these as generic intrinsics, to improve likelihood
1290 of diagnostics (instead of subtle run-time bugs) when using a
1291 compiler that doesn't support these as intrinsics (e.g. `f2c').
1293 * Remove from `-fno-pedantic' the diagnostic about `DO' with
1294 non-`INTEGER' index variable; issue that under `-Wsurprising'
1297 * Clarify some diagnostics that say things like "ignored" when that's
1300 * Clarify diagnostic on use of `.EQ.'/`.NE.' on `LOGICAL' operands.
1302 * Minor improvements to code generation for various operations on
1305 * Minor improvement to code generation for some `DO' loops on some
1308 * Support `gcc' version 2.7.1.
1310 * Upgrade to `libf2c' as of 1995-11-15.
1315 * Fix a code-generation bug involving complicated `EQUIVALENCE'
1316 statements not involving `COMMON'.
1318 * Fix code-generation bugs involving invoking "gratis" library
1319 procedures in `libf2c' from code compiled with `-fno-f2c' by
1320 making these procedures known to `g77' as intrinsics (not affected
1321 by -fno-f2c). This is known to fix code invoking `ERF()',
1322 `ERFC()', `DERF()', and `DERFC()'.
1324 * Update `libf2c' to include netlib patches through 1995-08-16, and
1325 `#define' `WANT_LEAD_0' to 1 to make `g77'-compiled code more
1326 consistent with other Fortran implementations by outputting
1327 leading zeros in formatted and list-directed output.
1329 * Fix a code-generation bug involving adjustable dummy arrays with
1330 high bounds whose primaries are changed during procedure
1331 execution, and which might well improve code-generation
1332 performance for such arrays compared to `f2c' plus `gcc' (but
1333 apparently only when using `gcc-2.7.0' or later).
1335 * Fix a code-generation bug involving invocation of `COMPLEX' and
1336 `DOUBLE COMPLEX' `FUNCTION's and doing `COMPLEX' and `DOUBLE
1337 COMPLEX' divides, when the result of the invocation or divide is
1338 assigned directly to a variable that overlaps one or more of the
1339 arguments to the invocation or divide.
1341 * Fix crash by not generating new optimal code for `X**I' if `I' is
1342 nonconstant and the expression is used to dimension a dummy array,
1343 since the `gcc' back end does not support the necessary mechanics
1344 (and the `gcc' front end rejects the equivalent construct, as it
1347 * Fix crash on expressions like `COMPLEX**INTEGER'.
1349 * Fix crash on expressions like `(1D0,2D0)**2', i.e. raising a
1350 `DOUBLE COMPLEX' constant to an `INTEGER' constant power.
1352 * Fix crashes and such involving diagnosed code.
1354 * Diagnose, instead of crashing on, statement function definitions
1355 having duplicate dummy argument names.
1357 * Fix bug causing rejection of good code involving statement function
1360 * Fix bug resulting in debugger not knowing size of local equivalence
1361 area when any member of area has initial value (via `DATA', for
1364 * Fix installation bug that prevented installation of `g77' driver.
1365 Provide for easy selection of whether to install copy of `g77' as
1366 `f77' to replace the broken code.
1368 * Fix `gcc' driver (affects `g77' thereby) to not gratuitously
1369 invoke the `f771' program (e.g. when `-E' is specified).
1371 * Fix diagnostic to point to correct source line when it immediately
1372 follows an `INCLUDE' statement.
1374 * Support more compiler options in `gcc'/`g77' when compiling
1375 Fortran files. These options include `-p', `-pg', `-aux-info',
1376 `-P', correct setting of version-number macros for preprocessing,
1377 full recognition of `-O0', and automatic insertion of
1378 configuration-specific linker specs.
1380 * Add new intrinsics that interface to existing routines in `libf2c':
1381 `ABORT', `DERF', `DERFC', `ERF', `ERFC', `EXIT', `FLUSH',
1382 `GETARG', `GETENV', `IARGC', `SIGNAL', and `SYSTEM'. Note that
1383 `ABORT', `EXIT', `FLUSH', `SIGNAL', and `SYSTEM' are intrinsic
1384 subroutines, not functions (since they have side effects), so to
1385 get the return values from `SIGNAL' and `SYSTEM', append a final
1386 argument specifying an `INTEGER' variable or array element (e.g.
1387 `CALL SYSTEM('rm foo',ISTAT)').
1389 * Add new intrinsic group named `unix' to contain the new intrinsics,
1390 and by default enable this new group.
1392 * Move `LOC()' intrinsic out of the `vxt' group to the new `unix'
1395 * Improve `g77' so that `g77 -v' by itself (or with certain other
1396 options, including `-B', `-b', `-i', `-nostdlib', and `-V')
1397 reports lots more useful version info, and so that long-form
1398 options `gcc' accepts are understood by `g77' as well (even in
1399 truncated, unambiguous forms).
1401 * Add new `g77' option `--driver=name' to specify driver when
1402 default, `gcc', isn't appropriate.
1404 * Add support for `#' directives (as output by the preprocessor) in
1405 the compiler, and enable generation of those directives by the
1406 preprocessor (when compiling `.F' files) so diagnostics and
1407 debugging info are more useful to users of the preprocessor.
1409 * Produce better diagnostics, more like `gcc', with info such as `In
1410 function `foo':' and `In file included from...:'.
1412 * Support `gcc''s `-fident' and `-fno-ident' options.
1414 * When `-Wunused' in effect, don't warn about local variables used as
1415 statement-function dummy arguments or `DATA' implied-`DO' iteration
1416 variables, even though, strictly speaking, these are not uses of
1417 the variables themselves.
1419 * When `-W -Wunused' in effect, don't warn about unused dummy
1420 arguments at all, since there's no way to turn this off for
1421 individual cases (`g77' might someday start warning about
1422 these)--applies to `gcc' versions 2.7.0 and later, since earlier
1423 versions didn't warn about unused dummy arguments.
1425 * New option `-fno-underscoring' that inhibits transformation of
1426 names (by appending one or two underscores) so users may experiment
1427 with implications of such an environment.
1429 * Minor improvement to `gcc/f/info' module to make it easier to build
1430 `g77' using the native (non-`gcc') compiler on certain machines
1431 (but definitely not all machines nor all non-`gcc' compilers).
1432 Please do not report bugs showing problems compilers have with
1433 macros defined in `gcc/f/target.h' and used in places like
1436 * Add warning to be printed for each invocation of the compiler if
1437 the target machine `INTEGER', `REAL', or `LOGICAL' size is not 32
1438 bits, since `g77' is known to not work well for such cases (to be
1439 fixed in Version 0.6--*note Actual Bugs We Haven't Fixed Yet:
1442 * Lots of new documentation (though work is still needed to put it
1443 into canonical GNU format).
1445 * Build `libf2c' with `-g0', not `-g2', in effect (by default), to
1446 produce smaller library without lots of debugging clutter.
1451 * Fix bad code generation involving `X**I' and temporary, internal
1452 variables generated by `g77' and the back end (such as for `DO'
1455 * Fix crash given `CHARACTER A;DATA A/.TRUE./'.
1457 * Replace crash with diagnostic given `CHARACTER A;DATA A/1.0/'.
1459 * Fix crash or other erratic behavior when null character constant
1460 (`''') is encountered.
1462 * Fix crash or other erratic behavior involving diagnosed code.
1464 * Fix code generation for external functions returning type `REAL'
1465 when the `-ff2c' option is in force (which it is by default) so
1466 that `f2c' compatibility is indeed provided.
1468 * Disallow `COMMON I(10)' if `I' has previously been specified with
1469 an array declarator.
1471 * New `-ffixed-line-length-N' option, where N is the maximum length
1472 of a typical fixed-form line, defaulting to 72 columns, such that
1473 characters beyond column N are ignored, or N is `none', meaning no
1474 characters are ignored. does not affect lines with `&' in column
1475 1, which are always processed as if `-ffixed-line-length-none' was
1478 * No longer generate better code for some kinds of array references,
1479 as `gcc' back end is to be fixed to do this even better, and it
1480 turned out to slow down some code in some cases after all.
1482 * In `COMMON' and `EQUIVALENCE' areas with any members given initial
1483 values (e.g. via `DATA'), uninitialized members now always
1484 initialized to binary zeros (though this is not required by the
1485 standard, and might not be done in future versions of `g77').
1486 Previously, in some `COMMON'/`EQUIVALENCE' areas (essentially
1487 those with members of more than one type), the uninitialized
1488 members were initialized to spaces, to cater to `CHARACTER' types,
1489 but it seems no existing code expects that, while much existing
1490 code expects binary zeros.
1495 * Don't emit bad code when low bound of adjustable array is
1496 nonconstant and thus might vary as an expression at run time.
1498 * Emit correct code for calculation of number of trips in `DO' loops
1499 for cases where the loop should not execute at all. (This bug
1500 affected cases where the difference between the begin and end
1501 values was less than the step count, though probably not for
1502 floating-point cases.)
1504 * Fix crash when extra parentheses surround item in `DATA'
1507 * Fix crash over minor internal inconsistencies in handling
1508 diagnostics, just substitute dummy strings where necessary.
1510 * Fix crash on some systems when compiling call to `MVBITS()'
1513 * Fix crash on array assignment `TYPEDDD(...)=...', where DDD is a
1514 string of one or more digits.
1516 * Fix crash on `DCMPLX()' with a single `INTEGER' argument.
1518 * Fix various crashes involving code with diagnosed errors.
1520 * Support `-I' option for `INCLUDE' statement, plus `gcc''s
1521 `header.gcc' facility for handling systems like MS-DOS.
1523 * Allow `INCLUDE' statement to be continued across multiple lines,
1524 even allow it to coexist with other statements on the same line.
1526 * Incorporate Bellcore fixes to `libf2c' through 1995-03-15--this
1527 fixes a bug involving infinite loops reading EOF with empty
1528 list-directed I/O list.
1530 * Remove all the `g77'-specific auto-configuration scripts, code,
1531 and so on, except for temporary substitutes for bsearch() and
1532 strtoul(), as too many configure/build problems were reported in
1533 these areas. People will have to fix their systems' problems
1534 themselves, or at least somewhere other than `g77', which expects
1535 a working ANSI C environment (and, for now, a GNU C compiler to
1536 compile `g77' itself).
1538 * Complain if initialized common redeclared as larger in subsequent
1541 * Warn if blank common initialized, since its size can vary and hence
1542 related warnings that might be helpful won't be seen.
1544 * New `-fbackslash' option, on by default, that causes `\' within
1545 `CHARACTER' and Hollerith constants to be interpreted a la GNU C.
1546 Note that this behavior is somewhat different from `f2c''s, which
1547 supports only a limited subset of backslash (escape) sequences.
1549 * Make `-fugly-args' the default.
1551 * New `-fugly-init' option, on by default, that allows
1552 typeless/Hollerith to be specified as initial values for variables
1553 or named constants (`PARAMETER'), and also allows
1554 character<->numeric conversion in those contexts--turn off via
1557 * New `-finit-local-zero' option to initialize local variables to
1558 binary zeros. This does not affect whether they are `SAVE'd, i.e.
1559 made automatic or static.
1561 * New `-Wimplicit' option to warn about implicitly typed variables,
1562 arrays, and functions. (Basically causes all program units to
1563 default to `IMPLICIT NONE'.)
1565 * `-Wall' now implies `-Wuninitialized' as with `gcc' (i.e. unless
1566 `-O' not specified, since `-Wuninitialized' requires `-O'), and
1567 implies `-Wunused' as well.
1569 * `-Wunused' no longer gives spurious messages for unused `EXTERNAL'
1570 names (since they are assumed to refer to block data program
1571 units, to make use of libraries more reliable).
1573 * Support `%LOC()' and `LOC()' of character arguments.
1575 * Support null (zero-length) character constants and expressions.
1577 * Support `f2c''s `IMAG()' generic intrinsic.
1579 * Support `ICHAR()', `IACHAR()', and `LEN()' of character
1580 expressions that are valid in assignments but not normally as
1583 * Support `f2c'-style `&' in column 1 to mean continuation line.
1585 * Allow `NAMELIST', `EXTERNAL', `INTRINSIC', and `VOLATILE' in
1586 `BLOCK DATA', even though these are not allowed by the standard.
1588 * Allow `RETURN' in main program unit.
1590 * Changes to Hollerith-constant support to obey Appendix C of the
1593 - Now padded on the right with zeros, not spaces.
1595 - Hollerith "format specifications" in the form of arrays of
1596 non-character allowed.
1598 - Warnings issued when non-space truncation occurs when
1599 converting to another type.
1601 - When specified as actual argument, now passed by reference to
1602 `INTEGER' (padded on right with spaces if constant too small,
1603 otherwise fully intact if constant wider the `INTEGER' type)
1604 instead of by value.
1606 *Warning:* `f2c' differs on the interpretation of `CALL FOO(1HX)',
1607 which it treats exactly the same as `CALL FOO('X')', but which the
1608 standard and `g77' treat as `CALL FOO(%REF('X '))' (padded with
1609 as many spaces as necessary to widen to `INTEGER'), essentially.
1611 * Changes and fixes to typeless-constant support:
1613 - Now treated as a typeless double-length `INTEGER' value.
1615 - Warnings issued when overflow occurs.
1617 - Padded on the left with zeros when converting to a larger
1620 - Should be properly aligned and ordered on the target machine
1621 for whatever type it is turned into.
1623 - When specified as actual argument, now passed as reference to
1624 a default `INTEGER' constant.
1626 * `%DESCR()' of a non-`CHARACTER' expression now passes a pointer to
1627 the expression plus a length for the expression just as if it were
1628 a `CHARACTER' expression. For example, `CALL FOO(%DESCR(D))',
1629 where `D' is `REAL*8', is the same as `CALL FOO(D,%VAL(8)))'.
1631 * Name of multi-entrypoint master function changed to incorporate
1632 the name of the primary entry point instead of a decimal value, so
1633 the name of the master function for `SUBROUTINE X' with alternate
1634 entry points is now `__g77_masterfun_x'.
1636 * Remove redundant message about zero-step-count `DO' loops.
1638 * Clean up diagnostic messages, shortening many of them.
1640 * Fix typo in `g77' man page.
1642 * Clarify implications of constant-handling bugs in `f/BUGS'.
1644 * Generate better code for `**' operator with a right-hand operand of
1647 * Generate better code for `SQRT()' and `DSQRT()', also when
1648 `-ffast-math' specified, enable better code generation for `SIN()'
1651 * Generate better code for some kinds of array references.
1653 * Speed up lexing somewhat (this makes the compilation phase
OSDN Git Service
