1 This file lists recent changes to the GNU Fortran compiler. Copyright
2 (C) 1995, 1996 Free Software Foundation, Inc. You may copy,
3 distribute, and modify it freely as long as you preserve this copyright
4 notice and permission notice.
9 Changes made to recent versions of GNU Fortran are listed below,
10 with the most recent version first.
12 The changes are generally listed in order:
14 1. Code-generation and run-time-library bug-fixes
16 2. Compiler and run-time-library crashes involving valid code that
21 4. Fixes and enhancements to existing features
25 6. Internal improvements
29 This order is not strict--for example, some items involve a
30 combination of these elements.
32 Note that two variants of `g77' are tracked below. The `egcs'
33 variant is described vis-a-vis previous versions of `egcs' and/or an
34 official FSF version, as appropriate.
36 Therefore, `egcs' versions sometimes have multiple listings to help
37 clarify how they differ from other versions, though this can make
38 getting a complete picture of what a particular `egcs' version contains
39 somewhat more difficult.
41 An online, "live" version of this document (derived directly from
42 the up-to-date mainline version of `g77' within `egcs') is available at
43 `http://egcs.cygnus.com/onlinedocs/g77_news.html'.
45 In 0.5.24 and `egcs' 1.1.2 (versus 0.5.23 and 1.1.1):
46 =====================================================
48 * Fix the `IDate' Intrinsic (VXT) so the returned year is in the
49 documented, non-Y2K-compliant range of 0-99, instead of being
50 returned as 100 in the year 2000.
52 * Fix the `Date_and_Time' intrinsic (in `libg2c') to return the
53 milliseconds value properly in VALUES(8).
55 * Fix the `LStat' intrinsic (in `libg2c') to return device-ID
56 information properly in SARRAY(7).
58 * Improve documentation.
60 In 0.5.24 and `egcs' 1.1.1 (versus 0.5.23 and 1.1):
61 ===================================================
63 * Fix `libg2c' so it performs an implicit `ENDFILE' operation (as
64 appropriate) whenever a `REWIND' is done.
66 (This bug was introduced in 0.5.23 and `egcs' 1.1 in `g77''s
69 * Fix `libg2c' so it no longer crashes with a spurious diagnostic
70 upon doing any I/O following a direct formatted write.
72 (This bug was introduced in 0.5.23 and `egcs' 1.1 in `g77''s
75 * Fix `g77' so it no longer crashes compiling references to the
76 `Rand' intrinsic on some systems.
78 * Fix `g77' portion of installation process so it works better on
79 some systems (those with shells requiring `else true' clauses on
80 `if' constructs for the completion code to be set properly).
82 In `egcs' 1.1 (versus 0.5.24):
83 ==============================
85 * Fix `g77' crash compiling code containing the construct
86 `CMPLX(0.)' or similar.
88 * Fix `g77' crash (or apparently infinite run-time) when compiling
89 certain complicated expressions involving `COMPLEX' arithmetic
90 (especially multiplication).
92 * Fix a code-generation bug that afflicted Intel x86 targets when
93 `-O2' was specified compiling, for example, an old version of the
96 The x87 coprocessor stack was being mismanaged in cases involving
97 assigned `GOTO' and `ASSIGN'.
99 * Align static double-precision variables and arrays on Intel x86
100 targets regardless of whether `-malign-double' is specified.
102 Generally, this affects only local variables and arrays having the
103 `SAVE' attribute or given initial values via `DATA'.
105 In `egcs' 1.1 (versus `egcs' 1.0.3):
106 ====================================
108 * Fix bugs in the `libU77' intrinsic `HostNm' that wrote one byte
109 beyond the end of its `CHARACTER' argument, and in the `libU77'
110 intrinsics `GMTime' and `LTime' that overwrote their arguments.
112 * Assumed arrays with negative bounds (such as `REAL A(-1:*)') no
113 longer elicit spurious diagnostics from `g77', even on systems
114 with pointers having different sizes than integers.
116 This bug is not known to have existed in any recent version of
117 `gcc'. It was introduced in an early release of `egcs'.
119 * Valid combinations of `EXTERNAL', passing that external as a dummy
120 argument without explicitly giving it a type, and, in a subsequent
121 program unit, referencing that external as an external function
122 with a different type no longer crash `g77'.
124 * `CASE DEFAULT' no longer crashes `g77'.
126 * The `-Wunused' option no longer issues a spurious warning about
127 the "master" procedure generated by `g77' for procedures
128 containing `ENTRY' statements.
130 * Support `FORMAT(I<EXPR>)' when EXPR is a compile-time constant
131 `INTEGER' expression.
133 * Fix `g77' `-g' option so procedures that use `ENTRY' can be
134 stepped through, line by line, in `gdb'.
136 * Allow any `REAL' argument to intrinsics `Second' and `CPU_Time'.
138 * Use `tempnam', if available, to open scratch files (as in
139 `OPEN(STATUS='SCRATCH')') so that the `TMPDIR' environment
140 variable, if present, is used.
142 * `g77''s version of `libf2c' separates out the setting of global
143 state (such as command-line arguments and signal handling) from
144 `main.o' into distinct, new library archive members.
146 This should make it easier to write portable applications that
147 have their own (non-Fortran) `main()' routine properly set up the
148 `libf2c' environment, even when `libf2c' (now `libg2c') is a
151 * `g77' no longer installs the `f77' command and `f77.1' man page in
152 the `/usr' or `/usr/local' heirarchy, even if the `f77-install-ok'
153 file exists in the source or build directory. See the
154 installation documentation for more information.
156 * `g77' no longer installs the `libf2c.a' library and `f2c.h'
157 include file in the `/usr' or `/usr/local' heirarchy, even if the
158 `f2c-install-ok' or `f2c-exists-ok' files exist in the source or
159 build directory. See the installation documentation for more
162 * The `libf2c.a' library produced by `g77' has been renamed to
163 `libg2c.a'. It is installed only in the `gcc' "private" directory
164 heirarchy, `gcc-lib'. This allows system administrators and users
165 to choose which version of the `libf2c' library from `netlib' they
166 wish to use on a case-by-case basis. See the installation
167 documentation for more information.
169 * The `f2c.h' include (header) file produced by `g77' has been
170 renamed to `g2c.h'. It is installed only in the `gcc' "private"
171 directory heirarchy, `gcc-lib'. This allows system administrators
172 and users to choose which version of the include file from
173 `netlib' they wish to use on a case-by-case basis. See the
174 installation documentation for more information.
176 * The `g77' command now expects the run-time library to be named
177 `libg2c.a' instead of `libf2c.a', to ensure that a version other
178 than the one built and installed as part of the same `g77' version
181 * During the configuration and build process, `g77' creates
182 subdirectories it needs only as it needs them. Other cleaning up
183 of the configuration and build process has been performed as well.
185 * `install-info' now used to update the directory of Info
186 documentation to contain an entry for `g77' (during installation).
188 * Some diagnostics have been changed from warnings to errors, to
189 prevent inadvertent use of the resulting, probably buggy, programs.
190 These mostly include diagnostics about use of unsupported features
191 in the `OPEN', `INQUIRE', `READ', and `WRITE' statements, and
192 about truncations of various sorts of constants.
194 * Improve compilation of `FORMAT' expressions so that a null byte is
195 appended to the last operand if it is a constant. This provides a
196 cleaner run-time diagnostic as provided by `libf2c' for statements
197 like `PRINT '(I1', 42'.
199 * Improve documentation and indexing.
201 * The upgrade to `libf2c' as of 1998-06-18 should fix a variety of
202 problems, including those involving some uses of the `T' format
203 specifier, and perhaps some build (porting) problems as well.
205 In 0.5.24 and `egcs' 1.1 (versus 0.5.23):
206 =========================================
208 * `g77' no longer produces incorrect code and initial values for
209 `EQUIVALENCE' and `COMMON' aggregates that, due to "unnatural"
210 ordering of members vis-a-vis their types, require initial padding.
212 * `g77' no longer crashes when compiling code containing
213 specification statements such as `INTEGER(KIND=7) PTR'.
215 * `g77' no longer crashes when compiling code such as `J = SIGNAL(1,
218 * `g77' now treats `%LOC(EXPR)' and `LOC(EXPR)' as "ordinary"
219 expressions when they are used as arguments in procedure calls.
220 This change applies only to global (filewide) analysis, making it
221 consistent with how `g77' actually generates code for these cases.
223 Previously, `g77' treated these expressions as denoting special
224 "pointer" arguments for the purposes of filewide analysis.
226 * The `g77' driver now ensures that `-lg2c' is specified in the link
227 phase prior to any occurrence of `-lm'. This prevents
228 accidentally linking to a routine in the SunOS4 `-lm' library when
229 the generated code wants to link to the one in `libf2c' (`libg2c').
231 * `g77' emits more debugging information when `-g' is used.
233 This new information allows, for example, `which __g77_length_a'
234 to be used in `gdb' to determine the type of the phantom length
235 argument supplied with `CHARACTER' variables.
237 This information pertains to internally-generated type, variable,
238 and other information, not to the longstanding deficiencies
239 vis-a-vis `COMMON' and `EQUIVALENCE'.
241 * The F90 `Date_and_Time' intrinsic now is supported.
243 * The F90 `System_Clock' intrinsic allows the optional arguments
244 (except for the `Count' argument) to be omitted.
246 * Upgrade to `libf2c' as of 1998-06-18.
248 * Improve documentation and indexing.
250 In 0.5.23 (versus 0.5.22):
251 ==========================
253 * This release contains several regressions against version 0.5.22
254 of `g77', due to using the "vanilla" `gcc' back end instead of
255 patching it to fix a few bugs and improve performance in a few
258 See `egcs/gcc/f/BUGS', for information on the known bugs in this
259 version, including the regressions.
261 Features that have been dropped from this version of `g77' due to
262 their being implemented via `g77'-specific patches to the `gcc'
263 back end in previous releases include:
265 - Support for `__restrict__' keyword, the options
266 `-fargument-alias', `-fargument-noalias', and
267 `-fargument-noalias-global', and the corresponding
270 (`egcs' has the alias-analysis code, but not the
271 `__restrict__' keyword. `egcs' `g77' users benefit from the
272 alias-analysis code despite the lack of the `__restrict__'
273 keyword, which is a C-language construct.)
275 - Support for the GNU compiler options `-fmove-all-movables',
276 `-freduce-all-givs', and `-frerun-loop-opt'.
278 (`egcs' supports these options. `g77' users of `egcs'
279 benefit from them even if they are not explicitly specified,
280 because the defaults are optimized for `g77' users.)
282 - Support for the `-W' option warning about integer division by
285 - The Intel x86-specific option `-malign-double' applying to
286 stack-allocated data as well as statically-allocate data.
288 Note that the `gcc/f/gbe/' subdirectory has been removed from this
289 distribution as a result of `g77' no longer including patches for
292 * Fix bugs in the `libU77' intrinsic `HostNm' that wrote one byte
293 beyond the end of its `CHARACTER' argument, and in the `libU77'
294 intrinsics `GMTime' and `LTime' that overwrote their arguments.
296 * Support `gcc' version 2.8, and remove support for prior versions
299 * Remove support for the `--driver' option, as `g77' now does all
300 the driving, just like `gcc'.
302 * `CASE DEFAULT' no longer crashes `g77'.
304 * Valid combinations of `EXTERNAL', passing that external as a dummy
305 argument without explicitly giving it a type, and, in a subsequent
306 program unit, referencing that external as an external function
307 with a different type no longer crash `g77'.
309 * `g77' no longer installs the `f77' command and `f77.1' man page in
310 the `/usr' or `/usr/local' heirarchy, even if the `f77-install-ok'
311 file exists in the source or build directory. See the
312 installation documentation for more information.
314 * `g77' no longer installs the `libf2c.a' library and `f2c.h'
315 include file in the `/usr' or `/usr/local' heirarchy, even if the
316 `f2c-install-ok' or `f2c-exists-ok' files exist in the source or
317 build directory. See the installation documentation for more
320 * The `libf2c.a' library produced by `g77' has been renamed to
321 `libg2c.a'. It is installed only in the `gcc' "private" directory
322 heirarchy, `gcc-lib'. This allows system administrators and users
323 to choose which version of the `libf2c' library from `netlib' they
324 wish to use on a case-by-case basis. See the installation
325 documentation for more information.
327 * The `f2c.h' include (header) file produced by `g77' has been
328 renamed to `g2c.h'. It is installed only in the `gcc' "private"
329 directory heirarchy, `gcc-lib'. This allows system administrators
330 and users to choose which version of the include file from
331 `netlib' they wish to use on a case-by-case basis. See the
332 installation documentation for more information.
334 * The `g77' command now expects the run-time library to be named
335 `libg2c.a' instead of `libf2c.a', to ensure that a version other
336 than the one built and installed as part of the same `g77' version
339 * The `-Wunused' option no longer issues a spurious warning about
340 the "master" procedure generated by `g77' for procedures
341 containing `ENTRY' statements.
343 * `g77''s version of `libf2c' separates out the setting of global
344 state (such as command-line arguments and signal handling) from
345 `main.o' into distinct, new library archive members.
347 This should make it easier to write portable applications that
348 have their own (non-Fortran) `main()' routine properly set up the
349 `libf2c' environment, even when `libf2c' (now `libg2c') is a
352 * During the configuration and build process, `g77' creates
353 subdirectories it needs only as it needs them, thus avoiding
354 unnecessary creation of, for example, `stage1/f/runtime' when
355 doing a non-bootstrap build. Other cleaning up of the
356 configuration and build process has been performed as well.
358 * `install-info' now used to update the directory of Info
359 documentation to contain an entry for `g77' (during installation).
361 * Some diagnostics have been changed from warnings to errors, to
362 prevent inadvertent use of the resulting, probably buggy, programs.
363 These mostly include diagnostics about use of unsupported features
364 in the `OPEN', `INQUIRE', `READ', and `WRITE' statements, and
365 about truncations of various sorts of constants.
367 * Improve documentation and indexing.
369 * Upgrade to `libf2c' as of 1998-04-20.
371 This should fix a variety of problems, including those involving
372 some uses of the `T' format specifier, and perhaps some build
373 (porting) problems as well.
375 In 0.5.22 (versus 0.5.21):
376 ==========================
378 * Fix code generation for iterative `DO' loops that have one or more
379 references to the iteration variable, or to aliases of it, in
380 their control expressions. For example, `DO 10 J=2,J' now is
383 * Fix a code-generation bug that afflicted Intel x86 targets when
384 `-O2' was specified compiling, for example, an old version of the
387 The x87 coprocessor stack was being mismanaged in cases involving
388 assigned `GOTO' and `ASSIGN'.
390 * Fix `DTime' intrinsic so as not to truncate results to integer
391 values (on some systems).
393 * Fix `Signal' intrinsic so it offers portable support for 64-bit
394 systems (such as Digital Alphas running GNU/Linux).
396 * Fix run-time crash involving `NAMELIST' on 64-bit machines such as
399 * Fix `g77' version of `libf2c' so it no longer produces a spurious
400 `I/O recursion' diagnostic at run time when an I/O operation (such
401 as `READ *,I') is interrupted in a manner that causes the program
402 to be terminated via the `f_exit' routine (such as via `C-c').
404 * Fix `g77' crash triggered by `CASE' statement with an omitted
405 lower or upper bound.
407 * Fix `g77' crash compiling references to `CPU_Time' intrinsic.
409 * Fix `g77' crash (or apparently infinite run-time) when compiling
410 certain complicated expressions involving `COMPLEX' arithmetic
411 (especially multiplication).
413 * Fix `g77' crash on statements such as `PRINT *,
414 (REAL(Z(I)),I=1,2)', where `Z' is `DOUBLE COMPLEX'.
418 * Support `FORMAT(I<EXPR>)' when EXPR is a compile-time constant
419 `INTEGER' expression.
421 * Fix `g77' `-g' option so procedures that use `ENTRY' can be
422 stepped through, line by line, in `gdb'.
424 * Fix a profiling-related bug in `gcc' back end for Intel x86
427 * Allow any `REAL' argument to intrinsics `Second' and `CPU_Time'.
429 * Allow any numeric argument to intrinsics `Int2' and `Int8'.
431 * Use `tempnam', if available, to open scratch files (as in
432 `OPEN(STATUS='SCRATCH')') so that the `TMPDIR' environment
433 variable, if present, is used.
435 * Rename the `gcc' keyword `restrict' to `__restrict__', to avoid
436 rejecting valid, existing, C programs. Support for `restrict' is
437 now more like support for `complex'.
439 * Fix `-fpedantic' to not reject procedure invocations such as
440 `I=J()' and `CALL FOO()'.
442 * Fix `-fugly-comma' to affect invocations of only external
443 procedures. Restore rejection of gratuitous trailing omitted
444 arguments to intrinsics, as in `I=MAX(3,4,,)'.
446 * Fix compiler so it accepts `-fgnu-intrinsics-*' and
447 `-fbadu77-intrinsics-*' options.
449 * Improve diagnostic messages from `libf2c' so it is more likely
450 that the printing of the active format string is limited to the
451 string, with no trailing garbage being printed.
453 (Unlike `f2c', `g77' did not append a null byte to its compiled
454 form of every format string specified via a `FORMAT' statement.
455 However, `f2c' would exhibit the problem anyway for a statement
456 like `PRINT '(I)garbage', 1' by printing `(I)garbage' as the
459 * Improve compilation of `FORMAT' expressions so that a null byte is
460 appended to the last operand if it is a constant. This provides a
461 cleaner run-time diagnostic as provided by `libf2c' for statements
462 like `PRINT '(I1', 42'.
464 * Fix various crashes involving code with diagnosed errors.
466 * Fix cross-compilation bug when configuring `libf2c'.
468 * Improve diagnostics.
470 * Improve documentation and indexing.
472 * Upgrade to `libf2c' as of 1997-09-23. This fixes a formatted-I/O
473 bug that afflicted 64-bit systems with 32-bit integers (such as
474 Digital Alpha running GNU/Linux).
476 In `egcs' 1.0.2 (versus `egcs' 1.0.1):
477 ======================================
479 * Fix `g77' crash triggered by `CASE' statement with an omitted
480 lower or upper bound.
482 * Fix `g77' crash on statements such as `PRINT *,
483 (REAL(Z(I)),I=1,2)', where `Z' is `DOUBLE COMPLEX'.
485 * Fix `-fPIC' (such as compiling for ELF targets) on the Intel x86
486 architecture target so invalid assembler code is no longer
489 * Fix `-fpedantic' to not reject procedure invocations such as
490 `I=J()' and `CALL FOO()'.
492 * Fix `-fugly-comma' to affect invocations of only external
493 procedures. Restore rejection of gratuitous trailing omitted
494 arguments to intrinsics, as in `I=MAX(3,4,,)'.
496 * Fix compiler so it accepts `-fgnu-intrinsics-*' and
497 `-fbadu77-intrinsics-*' options.
499 In `egcs' 1.0.1 (versus `egcs' 1.0):
500 ====================================
502 * Fix run-time crash involving `NAMELIST' on 64-bit machines such as
505 In `egcs' 1.0 (versus 0.5.21):
506 ==============================
508 * Version 1.0 of `egcs' contains several regressions against version
509 0.5.21 of `g77', due to using the "vanilla" `gcc' back end instead
510 of patching it to fix a few bugs and improve performance in a few
513 See `egcs/gcc/f/BUGS', for information on the known bugs in this
514 version, including the regressions.
516 Features that have been dropped from this version of `g77' due to
517 their being implemented via `g77'-specific patches to the `gcc'
518 back end in previous releases include:
520 - Support for the C-language `restrict' keyword.
522 - Support for the `-W' option warning about integer division by
525 - The Intel x86-specific option `-malign-double' applying to
526 stack-allocated data as well as statically-allocate data.
528 Note that the `gcc/f/gbe/' subdirectory has been removed from this
529 distribution as a result of `g77' being fully integrated with the
530 `egcs' variant of the `gcc' back end.
532 * Fix code generation for iterative `DO' loops that have one or more
533 references to the iteration variable, or to aliases of it, in
534 their control expressions. For example, `DO 10 J=2,J' now is
537 * Fix `DTime' intrinsic so as not to truncate results to integer
538 values (on some systems).
540 * Remove support for non-`egcs' versions of `gcc'.
542 * Remove support for the `--driver' option, as `g77' now does all
543 the driving, just like `gcc'.
545 * Allow any numeric argument to intrinsics `Int2' and `Int8'.
547 * Improve diagnostic messages from `libf2c' so it is more likely
548 that the printing of the active format string is limited to the
549 string, with no trailing garbage being printed.
551 (Unlike `f2c', `g77' did not append a null byte to its compiled
552 form of every format string specified via a `FORMAT' statement.
553 However, `f2c' would exhibit the problem anyway for a statement
554 like `PRINT '(I)garbage', 1' by printing `(I)garbage' as the
557 * Upgrade to `libf2c' as of 1997-09-23. This fixes a formatted-I/O
558 bug that afflicted 64-bit systems with 32-bit integers (such as
559 Digital Alpha running GNU/Linux).
564 * Fix a code-generation bug introduced by 0.5.20 caused by loop
565 unrolling (by specifying `-funroll-loops' or similar). This bug
566 afflicted all code compiled by version 2.7.2.2.f.2 of `gcc' (C,
567 C++, Fortran, and so on).
569 * Fix a code-generation bug manifested when combining local
570 `EQUIVALENCE' with a `DATA' statement that follows the first
571 executable statement (or is treated as an executable-context
572 statement as a result of using the `-fpedantic' option).
574 * Fix a compiler crash that occured when an integer division by a
575 constant zero is detected. Instead, when the `-W' option is
576 specified, the `gcc' back end issues a warning about such a case.
577 This bug afflicted all code compiled by version 2.7.2.2.f.2 of
578 `gcc' (C, C++, Fortran, and so on).
580 * Fix a compiler crash that occurred in some cases of procedure
581 inlining. (Such cases became more frequent in 0.5.20.)
583 * Fix a compiler crash resulting from using `DATA' or similar to
584 initialize a `COMPLEX' variable or array to zero.
586 * Fix compiler crashes involving use of `AND', `OR', or `XOR'
589 * Fix compiler bug triggered when using a `COMMON' or `EQUIVALENCE'
590 variable as the target of an `ASSIGN' or assigned-`GOTO' statement.
592 * Fix compiler crashes due to using the name of a some non-standard
593 intrinsics (such as `FTELL' or `FPUTC') as such and as the name of
594 a procedure or common block. Such dual use of a name in a program
595 is allowed by the standard.
597 * Place automatic arrays on the stack, even if `SAVE' or the
598 `-fno-automatic' option is in effect. This avoids a compiler
601 * The `-malign-double' option now reliably aligns `DOUBLE PRECISION'
602 optimally on Pentium and Pentium Pro architectures (586 and 686 in
605 * New option `-Wno-globals' disables warnings about "suspicious" use
606 of a name both as a global name and as the implicit name of an
607 intrinsic, and warnings about disagreements over the number or
608 natures of arguments passed to global procedures, or the natures
609 of the procedures themselves.
611 The default is to issue such warnings, which are new as of this
614 * New option `-fno-globals' disables diagnostics about potentially
615 fatal disagreements analysis problems, such as disagreements over
616 the number or natures of arguments passed to global procedures, or
617 the natures of those procedures themselves.
619 The default is to issue such diagnostics and flag the compilation
620 as unsuccessful. With this option, the diagnostics are issued as
621 warnings, or, if `-Wno-globals' is specified, are not issued at
624 This option also disables inlining of global procedures, to avoid
625 compiler crashes resulting from coding errors that these
626 diagnostics normally would identify.
628 * Diagnose cases where a reference to a procedure disagrees with the
629 type of that procedure, or where disagreements about the number or
630 nature of arguments exist. This avoids a compiler crash.
632 * Fix parsing bug whereby `g77' rejected a second initialization
633 specification immediately following the first's closing `/' without
634 an intervening comma in a `DATA' statement, and the second
635 specification was an implied-DO list.
637 * Improve performance of the `gcc' back end so certain complicated
638 expressions involving `COMPLEX' arithmetic (especially
639 multiplication) don't appear to take forever to compile.
641 * Fix a couple of profiling-related bugs in `gcc' back end.
643 * Integrate GNU Ada's (GNAT's) changes to the back end, which
644 consist almost entirely of bug fixes. These fixes are circa
645 version 3.10p of GNAT.
647 * Include some other `gcc' fixes that seem useful in `g77''s version
648 of `gcc'. (See `gcc/ChangeLog' for details--compare it to that
649 file in the vanilla `gcc-2.7.2.3.tar.gz' distribution.)
651 * Fix `libU77' routines that accept file and other names to strip
652 trailing blanks from them, for consistency with other
653 implementations. Blanks may be forcibly appended to such names by
654 appending a single null character (`CHAR(0)') to the significant
657 * Fix `CHMOD' intrinsic to work with file names that have embedded
658 blanks, commas, and so on.
660 * Fix `SIGNAL' intrinsic so it accepts an optional third `Status'
663 * Fix `IDATE()' intrinsic subroutine (VXT form) so it accepts
664 arguments in the correct order. Documentation fixed accordingly,
665 and for `GMTIME()' and `LTIME()' as well.
667 * Make many changes to `libU77' intrinsics to support existing code
670 Such changes include allowing both subroutine and function forms
671 of many routines, changing `MCLOCK()' and `TIME()' to return
672 `INTEGER(KIND=1)' values, introducing `MCLOCK8()' and `TIME8()' to
673 return `INTEGER(KIND=2)' values, and placing functions that are
674 intended to perform side effects in a new intrinsic group,
677 * Improve `libU77' so it is more portable.
679 * Add options `-fbadu77-intrinsics-delete',
680 `-fbadu77-intrinsics-hide', and so on.
682 * Fix crashes involving diagnosed or invalid code.
684 * `g77' and `gcc' now do a somewhat better job detecting and
685 diagnosing arrays that are too large to handle before these cause
686 diagnostics during the assembler or linker phase, a compiler
687 crash, or generation of incorrect code.
689 * Make some fixes to alias analysis code.
691 * Add support for `restrict' keyword in `gcc' front end.
693 * Support `gcc' version 2.7.2.3 (modified by `g77' into version
694 2.7.2.3.f.1), and remove support for prior versions of `gcc'.
696 * Incorporate GNAT's patches to the `gcc' back end into `g77''s, so
697 GNAT users do not need to apply GNAT's patches to build both GNAT
698 and `g77' from the same source tree.
700 * Modify `make' rules and related code so that generation of Info
701 documentation doesn't require compilation using `gcc'. Now, any
702 ANSI C compiler should be adequate to produce the `g77'
703 documentation (in particular, the tables of intrinsics) from
706 * Add `INT2' and `INT8' intrinsics.
708 * Add `CPU_TIME' intrinsic.
710 * Add `ALARM' intrinsic.
712 * `CTIME' intrinsic now accepts any `INTEGER' argument, not just
715 * Warn when explicit type declaration disagrees with the type of an
716 intrinsic invocation.
718 * Support `*f771' entry in `gcc' `specs' file.
720 * Fix typo in `make' rule `g77-cross', used only for cross-compiling.
722 * Fix `libf2c' build procedure to re-archive library if previous
723 attempt to archive was interrupted.
725 * Change `gcc' to unroll loops only during the last invocation (of
726 as many as two invocations) of loop optimization.
728 * Improve handling of `-fno-f2c' so that code that attempts to pass
729 an intrinsic as an actual argument, such as `CALL FOO(ABS)', is
730 rejected due to the fact that the run-time-library routine is,
731 effectively, compiled with `-ff2c' in effect.
733 * Fix `g77' driver to recognize `-fsyntax-only' as an option that
734 inhibits linking, just like `-c' or `-S', and to recognize and
735 properly handle the `-nostdlib', `-M', `-MM', `-nodefaultlibs',
736 and `-Xlinker' options.
738 * Upgrade to `libf2c' as of 1997-08-16.
740 * Modify `libf2c' to consistently and clearly diagnose recursive I/O
743 * `g77' driver now prints version information (such as produced by
744 `g77 -v') to `stderr' instead of `stdout'.
746 * The `.r' suffix now designates a Ratfor source file, to be
747 preprocessed via the `ratfor' command, available separately.
749 * Fix some aspects of how `gcc' determines what kind of system is
750 being configured and what kinds are supported. For example, GNU
751 Linux/Alpha ELF systems now are directly supported.
753 * Improve diagnostics.
755 * Improve documentation and indexing.
757 * Include all pertinent files for `libf2c' that come from
758 `netlib.bell-labs.com'; give any such files that aren't quite
759 accurate in `g77''s version of `libf2c' the suffix `.netlib'.
761 * Reserve `INTEGER(KIND=0)' for future use.
766 * The `-fno-typeless-boz' option is now the default.
768 This option specifies that non-decimal-radix constants using the
769 prefixed-radix form (such as `Z'1234'') are to be interpreted as
770 `INTEGER' constants. Specify `-ftypeless-boz' to cause such
771 constants to be interpreted as typeless.
773 (Version 0.5.19 introduced `-fno-typeless-boz' and its inverse.)
775 * Options `-ff90-intrinsics-enable' and `-fvxt-intrinsics-enable'
776 now are the defaults.
778 Some programs might use names that clash with intrinsic names
779 defined (and now enabled) by these options or by the new `libU77'
780 intrinsics. Users of such programs might need to compile them
781 differently (using, for example, `-ff90-intrinsics-disable') or,
782 better yet, insert appropriate `EXTERNAL' statements specifying
783 that these names are not intended to be names of intrinsics.
785 * The `ALWAYS_FLUSH' macro is no longer defined when building
786 `libf2c', which should result in improved I/O performance,
789 *Note:* If you have code that depends on the behavior of `libf2c'
790 when built with `ALWAYS_FLUSH' defined, you will have to modify
791 `libf2c' accordingly before building it from this and future
794 * Dave Love's implementation of `libU77' has been added to the
795 version of `libf2c' distributed with and built as part of `g77'.
796 `g77' now knows about the routines in this library as intrinsics.
798 * New option `-fvxt' specifies that the source file is written in
799 VXT Fortran, instead of GNU Fortran.
801 * The `-fvxt-not-f90' option has been deleted, along with its
802 inverse, `-ff90-not-vxt'.
804 If you used one of these deleted options, you should re-read the
805 pertinent documentation to determine which options, if any, are
806 appropriate for compiling your code with this version of `g77'.
808 * The `-fugly' option now issues a warning, as it likely will be
809 removed in a future version.
811 (Enabling all the `-fugly-*' options is unlikely to be feasible,
812 or sensible, in the future, so users should learn to specify only
813 those `-fugly-*' options they really need for a particular source
816 * The `-fugly-assumed' option, introduced in version 0.5.19, has
817 been changed to better accommodate old and new code.
819 * Make a number of fixes to the `g77' front end and the `gcc' back
820 end to better support Alpha (AXP) machines. This includes
821 providing at least one bug-fix to the `gcc' back end for Alphas.
823 * Related to supporting Alpha (AXP) machines, the `LOC()' intrinsic
824 and `%LOC()' construct now return values of integer type that is
825 the same width (holds the same number of bits) as the pointer type
828 On most machines, this won't make a difference, whereas on Alphas,
829 the type these constructs return is `INTEGER*8' instead of the
830 more common `INTEGER*4'.
832 * Emulate `COMPLEX' arithmetic in the `g77' front end, to avoid bugs
833 in `complex' support in the `gcc' back end. New option
834 `-fno-emulate-complex' causes `g77' to revert the 0.5.19 behavior.
836 * Fix bug whereby `REAL A(1)', for example, caused a compiler crash
837 if `-fugly-assumed' was in effect and A was a local (automatic)
838 array. That case is no longer affected by the new handling of
841 * Fix `g77' command driver so that `g77 -o foo.f' no longer deletes
842 `foo.f' before issuing other diagnostics, and so the `-x' option
845 * Enable inlining of subroutines and functions by the `gcc' back end.
846 This works as it does for `gcc' itself--program units may be
847 inlined for invocations that follow them in the same program unit,
848 as long as the appropriate compile-time options are specified.
850 * Dummy arguments are no longer assumed to potentially alias
851 (overlap) other dummy arguments or `COMMON' areas when any of
852 these are defined (assigned to) by Fortran code.
854 This can result in faster and/or smaller programs when compiling
855 with optimization enabled, though on some systems this effect is
856 observed only when `-fforce-addr' also is specified.
858 New options `-falias-check', `-fargument-alias',
859 `-fargument-noalias', and `-fno-argument-noalias-global' control
860 the way `g77' handles potential aliasing.
862 * The `CONJG()' and `DCONJG()' intrinsics now are compiled in-line.
864 * The bug-fix for 0.5.19.1 has been re-done. The `g77' compiler has
865 been changed back to assume `libf2c' has no aliasing problems in
866 its implementations of the `COMPLEX' (and `DOUBLE COMPLEX')
867 intrinsics. The `libf2c' has been changed to have no such
870 As a result, 0.5.20 is expected to offer improved performance over
871 0.5.19.1, perhaps as good as 0.5.19 in most or all cases, due to
874 *Note:* This change requires version 0.5.20 of `libf2c', at least,
875 when linking code produced by any versions of `g77' other than
876 0.5.19.1. Use `g77 -v' to determine the version numbers of the
877 `libF77', `libI77', and `libU77' components of the `libf2c'
878 library. (If these version numbers are not printed--in
879 particular, if the linker complains about unresolved references to
880 names like `g77__fvers__'--that strongly suggests your
881 installation has an obsolete version of `libf2c'.)
883 * New option `-fugly-assign' specifies that the same memory
884 locations are to be used to hold the values assigned by both
885 statements `I = 3' and `ASSIGN 10 TO I', for example. (Normally,
886 `g77' uses a separate memory location to hold assigned statement
889 * `FORMAT' and `ENTRY' statements now are allowed to precede
890 `IMPLICIT NONE' statements.
892 * Produce diagnostic for unsupported `SELECT CASE' on `CHARACTER'
893 type, instead of crashing, at compile time.
895 * Fix crashes involving diagnosed or invalid code.
897 * Change approach to building `libf2c' archive (`libf2c.a') so that
898 members are added to it only when truly necessary, so the user
899 that installs an already-built `g77' doesn't need to have write
900 access to the build tree (whereas the user doing the build might
901 not have access to install new software on the system).
903 * Support `gcc' version 2.7.2.2 (modified by `g77' into version
904 2.7.2.2.f.2), and remove support for prior versions of `gcc'.
906 * Upgrade to `libf2c' as of 1997-02-08, and fix up some of the build
909 * Improve general build procedures for `g77', fixing minor bugs
910 (such as deletion of any file named `f771' in the parent directory
913 * Enable full support of `INTEGER*8' available in `libf2c' and
914 `f2c.h' so that `f2c' users may make full use of its features via
915 the `g77' version of `f2c.h' and the `INTEGER*8' support routines
916 in the `g77' version of `libf2c'.
918 * Improve `g77' driver and `libf2c' so that `g77 -v' yields version
919 information on the library.
921 * The `SNGL' and `FLOAT' intrinsics now are specific intrinsics,
922 instead of synonyms for the generic intrinsic `REAL'.
924 * New intrinsics have been added. These are `REALPART', `IMAGPART',
925 `COMPLEX', `LONG', and `SHORT'.
927 * A new group of intrinsics, `gnu', has been added to contain the
928 new `REALPART', `IMAGPART', and `COMPLEX' intrinsics. An old
929 group, `dcp', has been removed.
931 * Complain about industry-wide ambiguous references `REAL(EXPR)' and
932 `AIMAG(EXPR)', where EXPR is `DOUBLE COMPLEX' (or any complex type
933 other than `COMPLEX'), unless `-ff90' option specifies Fortran 90
934 interpretation or new `-fugly-complex' option, in conjunction with
935 `-fnot-f90', specifies `f2c' interpretation.
937 * Make improvements to diagnostics.
939 * Speed up compiler a bit.
941 * Improvements to documentation and indexing, including a new
942 chapter containing information on one, later more, diagnostics
943 that users are directed to pull up automatically via a message in
944 the diagnostic itself.
946 (Hence the menu item `M' for the node `Diagnostics' in the
947 top-level menu of the Info documentation.)
952 * Code-generation bugs afflicting operations on complex data have
955 These bugs occurred when assigning the result of an operation to a
956 complex variable (or array element) that also served as an input
959 The operations affected by this bug were: `CONJG()', `DCONJG()',
960 `CCOS()', `CDCOS()', `CLOG()', `CDLOG()', `CSIN()', `CDSIN()',
961 `CSQRT()', `CDSQRT()', complex division, and raising a `DOUBLE
962 COMPLEX' operand to an `INTEGER' power. (The related generic and
963 `Z'-prefixed intrinsics, such as `ZSIN()', also were affected.)
965 For example, `C = CSQRT(C)', `Z = Z/C', and `Z = Z**I' (where `C'
966 is `COMPLEX' and `Z' is `DOUBLE COMPLEX') have been fixed.
971 * Fix `FORMAT' statement parsing so negative values for specifiers
972 such as `P' (e.g. `FORMAT(-1PF8.1)') are correctly processed as
975 * Fix `SIGNAL' intrinsic so it once again accepts a procedure as its
978 * A temporary kludge option provides bare-bones information on
979 `COMMON' and `EQUIVALENCE' members at debug time.
981 * New `-fonetrip' option specifies FORTRAN-66-style one-trip `DO'
984 * New `-fno-silent' option causes names of program units to be
985 printed as they are compiled, in a fashion similar to UNIX `f77'
988 * New `-fugly-assumed' option specifies that arrays dimensioned via
989 `DIMENSION X(1)', for example, are to be treated as assumed-size.
991 * New `-fno-typeless-boz' option specifies that non-decimal-radix
992 constants using the prefixed-radix form (such as `Z'1234'') are to
993 be interpreted as `INTEGER' constants.
995 * New `-ff66' option is a "shorthand" option that specifies
996 behaviors considered appropriate for FORTRAN 66 programs.
998 * New `-ff77' option is a "shorthand" option that specifies
999 behaviors considered appropriate for UNIX `f77' programs.
1001 * New `-fugly-comma' and `-fugly-logint' options provided to perform
1002 some of what `-fugly' used to do. `-fugly' and `-fno-ugly' are
1003 now "shorthand" options, in that they do nothing more than enable
1004 (or disable) other `-fugly-*' options.
1006 * Fix parsing of assignment statements involving targets that are
1007 substrings of elements of `CHARACTER' arrays having names such as
1008 `READ', `WRITE', `GOTO', and `REALFUNCTIONFOO'.
1010 * Fix crashes involving diagnosed code.
1012 * Fix handling of local `EQUIVALENCE' areas so certain cases of
1013 valid Fortran programs are not misdiagnosed as improperly
1014 extending the area backwards.
1016 * Support `gcc' version 2.7.2.1.
1018 * Upgrade to `libf2c' as of 1996-09-26, and fix up some of the build
1021 * Change code generation for list-directed I/O so it allows for new
1022 versions of `libf2c' that might return non-zero status codes for
1023 some operations previously assumed to always return zero.
1025 This change not only affects how `IOSTAT=' variables are set by
1026 list-directed I/O, it also affects whether `END=' and `ERR='
1027 labels are reached by these operations.
1029 * Add intrinsic support for new `FTELL' and `FSEEK' procedures in
1032 * Modify `fseek_()' in `libf2c' to be more portable (though, in
1033 practice, there might be no systems where this matters) and to
1034 catch invalid `whence' arguments.
1036 * Some useless warnings from the `-Wunused' option have been
1039 * Fix a problem building the `f771' executable on AIX systems by
1040 linking with the `-bbigtoc' option.
1042 * Abort configuration if `gcc' has not been patched using the patch
1043 file provided in the `gcc/f/gbe/' subdirectory.
1045 * Add options `--help' and `--version' to the `g77' command, to
1046 conform to GNU coding guidelines. Also add printing of `g77'
1047 version number when the `--verbose' (`-v') option is used.
1049 * Change internally generated name for local `EQUIVALENCE' areas to
1050 one based on the alphabetically sorted first name in the list of
1051 names for entities placed at the beginning of the areas.
1053 * Improvements to documentation and indexing.
1058 * Add some rudimentary support for `INTEGER*1', `INTEGER*2',
1059 `INTEGER*8', and their `LOGICAL' equivalents. (This support works
1060 on most, maybe all, `gcc' targets.)
1062 Thanks to Scott Snyder (<snyder@d0sgif.fnal.gov>) for providing
1065 Among the missing elements from the support for these features are
1066 full intrinsic support and constants.
1068 * Add some rudimentary support for the `BYTE' and `WORD'
1069 type-declaration statements. `BYTE' corresponds to `INTEGER*1',
1070 while `WORD' corresponds to `INTEGER*2'.
1072 Thanks to Scott Snyder (<snyder@d0sgif.fnal.gov>) for providing
1075 * The compiler code handling intrinsics has been largely rewritten
1076 to accommodate the new types. No new intrinsics or arguments for
1077 existing intrinsics have been added, so there is, at this point,
1078 no intrinsic to convert to `INTEGER*8', for example.
1080 * Support automatic arrays in procedures.
1082 * Reduce space/time requirements for handling large *sparsely*
1083 initialized aggregate arrays. This improvement applies to only a
1084 subset of the general problem to be addressed in 0.6.
1086 * Treat initial values of zero as if they weren't specified (in DATA
1087 and type-declaration statements). The initial values will be set
1088 to zero anyway, but the amount of compile time processing them
1089 will be reduced, in some cases significantly (though, again, this
1090 is only a subset of the general problem to be addressed in 0.6).
1092 A new option, `-fzeros', is introduced to enable the traditional
1093 treatment of zeros as any other value.
1095 * With `-ff90' in force, `g77' incorrectly interpreted `REAL(Z)' as
1096 returning a `REAL' result, instead of as a `DOUBLE PRECISION'
1097 result. (Here, `Z' is `DOUBLE COMPLEX'.)
1099 With `-fno-f90' in force, the interpretation remains unchanged,
1100 since this appears to be how at least some F77 code using the
1101 `DOUBLE COMPLEX' extension expected it to work.
1103 Essentially, `REAL(Z)' in F90 is the same as `DBLE(Z)', while in
1104 extended F77, it appears to be the same as `REAL(REAL(Z))'.
1106 * An expression involving exponentiation, where both operands were
1107 type `INTEGER' and the right-hand operand was negative, was
1108 erroneously evaluated.
1110 * Fix bugs involving `DATA' implied-`DO' constructs (these involved
1111 an errant diagnostic and a crash, both on good code, one involving
1112 subsequent statement-function definition).
1114 * Close `INCLUDE' files after processing them, so compiling source
1115 files with lots of `INCLUDE' statements does not result in being
1116 unable to open `INCLUDE' files after all the available file
1117 descriptors are used up.
1119 * Speed up compiling, especially of larger programs, and perhaps
1120 slightly reduce memory utilization while compiling (this is *not*
1121 the improvement planned for 0.6 involving large aggregate
1122 areas)--these improvements result from simply turning off some
1123 low-level code to do self-checking that hasn't been triggered in a
1126 * Introduce three new options that implement optimizations in the
1127 `gcc' back end (GBE). These options are `-fmove-all-movables',
1128 `-freduce-all-givs', and `-frerun-loop-opt', which are enabled, by
1129 default, for Fortran compilations. These optimizations are
1130 intended to help toon Fortran programs.
1132 * Patch the GBE to do a better job optimizing certain kinds of
1133 references to array elements.
1135 * Due to patches to the GBE, the version number of `gcc' also is
1136 patched to make it easier to manage installations, especially
1137 useful if it turns out a `g77' change to the GBE has a bug.
1139 The `g77'-modified version number is the `gcc' version number with
1140 the string `.f.N' appended, where `f' identifies the version as
1141 enhanced for Fortran, and N is `1' for the first Fortran patch for
1142 that version of `gcc', `2' for the second, and so on.
1144 So, this introduces version 2.7.2.f.1 of `gcc'.
1146 * Make several improvements and fixes to diagnostics, including the
1147 removal of two that were inappropriate or inadequate.
1149 * Warning about two successive arithmetic operators, produced by
1150 `-Wsurprising', now produced *only* when both operators are,
1151 indeed, arithmetic (not relational/boolean).
1153 * `-Wsurprising' now warns about the remaining cases of using
1154 non-integral variables for implied-`DO' loops, instead of these
1155 being rejected unless `-fpedantic' or `-fugly' specified.
1157 * Allow `SAVE' of a local variable or array, even after it has been
1158 given an initial value via `DATA', for example.
1160 * Introduce an Info version of `g77' documentation, which supercedes
1161 `gcc/f/CREDITS', `gcc/f/DOC', and `gcc/f/PROJECTS'. These files
1162 will be removed in a future release. The files `gcc/f/BUGS',
1163 `gcc/f/INSTALL', and `gcc/f/NEWS' now are automatically built from
1164 the texinfo source when distributions are made.
1166 This effort was inspired by a first pass at translating
1167 `g77-0.5.16/f/DOC' that was contributed to Craig by David Ronis
1168 (<ronis@onsager.chem.mcgill.ca>).
1170 * New `-fno-second-underscore' option to specify that, when
1171 `-funderscoring' is in effect, a second underscore is not to be
1172 appended to Fortran names already containing an underscore.
1174 * Change the way iterative `DO' loops work to follow the F90
1175 standard. In particular, calculation of the iteration count is
1176 still done by converting the start, end, and increment parameters
1177 to the type of the `DO' variable, but the result of the
1178 calculation is always converted to the default `INTEGER' type.
1180 (This should have no effect on existing code compiled by `g77',
1181 but code written to assume that use of a *wider* type for the `DO'
1182 variable will result in an iteration count being fully calculated
1183 using that wider type (wider than default `INTEGER') must be
1186 * Support `gcc' version 2.7.2.
1188 * Upgrade to `libf2c' as of 1996-03-23, and fix up some of the build
1191 Note that the email addresses related to `f2c' have changed--the
1192 distribution site now is named `netlib.bell-labs.com', and the
1193 maintainer's new address is <dmg@bell-labs.com>.
1198 * *Fix serious bug* in `g77 -v' command that can cause removal of a
1199 system's `/dev/null' special file if run by user `root'.
1201 *All users* of version 0.5.16 should ensure that they have not
1202 removed `/dev/null' or replaced it with an ordinary file (e.g. by
1203 comparing the output of `ls -l /dev/null' with `ls -l /dev/zero'.
1204 If the output isn't basically the same, contact your system
1205 administrator about restoring `/dev/null' to its proper status).
1207 This bug is particularly insidious because removing `/dev/null' as
1208 a special file can go undetected for quite a while, aside from
1209 various applications and programs exhibiting sudden, strange
1212 I sincerely apologize for not realizing the implications of the
1213 fact that when `g77 -v' runs the `ld' command with `-o /dev/null'
1214 that `ld' tries to *remove* the executable it is supposed to build
1215 (especially if it reports unresolved references, which it should
1218 * Fix crash on `CHARACTER*(*) FOO' in a main or block data program
1221 * Fix crash that can occur when diagnostics given outside of any
1222 program unit (such as when input file contains `@foo').
1224 * Fix crashes, infinite loops (hangs), and such involving diagnosed
1227 * Fix `ASSIGN''ed variables so they can be `SAVE''d or dummy
1228 arguments, and issue clearer error message in cases where target
1229 of `ASSIGN' or `ASSIGN'ed `GOTO'/`FORMAT' is too small (which
1230 should never happen).
1232 * Make `libf2c' build procedures work on more systems again by
1233 eliminating unnecessary invocations of `ld -r -x' and `mv'.
1235 * Fix omission of `-funix-intrinsics-...' options in list of
1236 permitted options to compiler.
1238 * Fix failure to always diagnose missing type declaration for
1241 * Fix compile-time performance problem (which could sometimes crash
1242 the compiler, cause a hang, or whatever, due to a bug in the back
1243 end) involving exponentiation with a large `INTEGER' constant for
1244 the right-hand operator (e.g. `I**32767').
1246 * Fix build procedures so cross-compiling `g77' (the `fini' utility
1247 in particular) is properly built using the host compiler.
1249 * Add new `-Wsurprising' option to warn about constructs that are
1250 interpreted by the Fortran standard (and `g77') in ways that are
1251 surprising to many programmers.
1253 * Add `ERF()' and `ERFC()' as generic intrinsics mapping to existing
1254 `ERF'/`DERF' and `ERFC'/`DERFC' specific intrinsics.
1256 *Note:* You should specify `INTRINSIC ERF,ERFC' in any code where
1257 you might use these as generic intrinsics, to improve likelihood
1258 of diagnostics (instead of subtle run-time bugs) when using a
1259 compiler that doesn't support these as intrinsics (e.g. `f2c').
1261 * Remove from `-fno-pedantic' the diagnostic about `DO' with
1262 non-`INTEGER' index variable; issue that under `-Wsurprising'
1265 * Clarify some diagnostics that say things like "ignored" when that's
1268 * Clarify diagnostic on use of `.EQ.'/`.NE.' on `LOGICAL' operands.
1270 * Minor improvements to code generation for various operations on
1273 * Minor improvement to code generation for some `DO' loops on some
1276 * Support `gcc' version 2.7.1.
1278 * Upgrade to `libf2c' as of 1995-11-15.
1283 * Fix a code-generation bug involving complicated `EQUIVALENCE'
1284 statements not involving `COMMON'.
1286 * Fix code-generation bugs involving invoking "gratis" library
1287 procedures in `libf2c' from code compiled with `-fno-f2c' by
1288 making these procedures known to `g77' as intrinsics (not affected
1289 by -fno-f2c). This is known to fix code invoking `ERF()',
1290 `ERFC()', `DERF()', and `DERFC()'.
1292 * Update `libf2c' to include netlib patches through 1995-08-16, and
1293 `#define' `WANT_LEAD_0' to 1 to make `g77'-compiled code more
1294 consistent with other Fortran implementations by outputting
1295 leading zeros in formatted and list-directed output.
1297 * Fix a code-generation bug involving adjustable dummy arrays with
1298 high bounds whose primaries are changed during procedure
1299 execution, and which might well improve code-generation
1300 performance for such arrays compared to `f2c' plus `gcc' (but
1301 apparently only when using `gcc-2.7.0' or later).
1303 * Fix a code-generation bug involving invocation of `COMPLEX' and
1304 `DOUBLE COMPLEX' `FUNCTION's and doing `COMPLEX' and `DOUBLE
1305 COMPLEX' divides, when the result of the invocation or divide is
1306 assigned directly to a variable that overlaps one or more of the
1307 arguments to the invocation or divide.
1309 * Fix crash by not generating new optimal code for `X**I' if `I' is
1310 nonconstant and the expression is used to dimension a dummy array,
1311 since the `gcc' back end does not support the necessary mechanics
1312 (and the `gcc' front end rejects the equivalent construct, as it
1315 * Fix crash on expressions like `COMPLEX**INTEGER'.
1317 * Fix crash on expressions like `(1D0,2D0)**2', i.e. raising a
1318 `DOUBLE COMPLEX' constant to an `INTEGER' constant power.
1320 * Fix crashes and such involving diagnosed code.
1322 * Diagnose, instead of crashing on, statement function definitions
1323 having duplicate dummy argument names.
1325 * Fix bug causing rejection of good code involving statement function
1328 * Fix bug resulting in debugger not knowing size of local equivalence
1329 area when any member of area has initial value (via `DATA', for
1332 * Fix installation bug that prevented installation of `g77' driver.
1333 Provide for easy selection of whether to install copy of `g77' as
1334 `f77' to replace the broken code.
1336 * Fix `gcc' driver (affects `g77' thereby) to not gratuitously
1337 invoke the `f771' program (e.g. when `-E' is specified).
1339 * Fix diagnostic to point to correct source line when it immediately
1340 follows an `INCLUDE' statement.
1342 * Support more compiler options in `gcc'/`g77' when compiling
1343 Fortran files. These options include `-p', `-pg', `-aux-info',
1344 `-P', correct setting of version-number macros for preprocessing,
1345 full recognition of `-O0', and automatic insertion of
1346 configuration-specific linker specs.
1348 * Add new intrinsics that interface to existing routines in `libf2c':
1349 `ABORT', `DERF', `DERFC', `ERF', `ERFC', `EXIT', `FLUSH',
1350 `GETARG', `GETENV', `IARGC', `SIGNAL', and `SYSTEM'. Note that
1351 `ABORT', `EXIT', `FLUSH', `SIGNAL', and `SYSTEM' are intrinsic
1352 subroutines, not functions (since they have side effects), so to
1353 get the return values from `SIGNAL' and `SYSTEM', append a final
1354 argument specifying an `INTEGER' variable or array element (e.g.
1355 `CALL SYSTEM('rm foo',ISTAT)').
1357 * Add new intrinsic group named `unix' to contain the new intrinsics,
1358 and by default enable this new group.
1360 * Move `LOC()' intrinsic out of the `vxt' group to the new `unix'
1363 * Improve `g77' so that `g77 -v' by itself (or with certain other
1364 options, including `-B', `-b', `-i', `-nostdlib', and `-V')
1365 reports lots more useful version info, and so that long-form
1366 options `gcc' accepts are understood by `g77' as well (even in
1367 truncated, unambiguous forms).
1369 * Add new `g77' option `--driver=name' to specify driver when
1370 default, `gcc', isn't appropriate.
1372 * Add support for `#' directives (as output by the preprocessor) in
1373 the compiler, and enable generation of those directives by the
1374 preprocessor (when compiling `.F' files) so diagnostics and
1375 debugging info are more useful to users of the preprocessor.
1377 * Produce better diagnostics, more like `gcc', with info such as `In
1378 function `foo':' and `In file included from...:'.
1380 * Support `gcc''s `-fident' and `-fno-ident' options.
1382 * When `-Wunused' in effect, don't warn about local variables used as
1383 statement-function dummy arguments or `DATA' implied-`DO' iteration
1384 variables, even though, strictly speaking, these are not uses of
1385 the variables themselves.
1387 * When `-W -Wunused' in effect, don't warn about unused dummy
1388 arguments at all, since there's no way to turn this off for
1389 individual cases (`g77' might someday start warning about
1390 these)--applies to `gcc' versions 2.7.0 and later, since earlier
1391 versions didn't warn about unused dummy arguments.
1393 * New option `-fno-underscoring' that inhibits transformation of
1394 names (by appending one or two underscores) so users may experiment
1395 with implications of such an environment.
1397 * Minor improvement to `gcc/f/info' module to make it easier to build
1398 `g77' using the native (non-`gcc') compiler on certain machines
1399 (but definitely not all machines nor all non-`gcc' compilers).
1400 Please do not report bugs showing problems compilers have with
1401 macros defined in `gcc/f/target.h' and used in places like
1404 * Add warning to be printed for each invocation of the compiler if
1405 the target machine `INTEGER', `REAL', or `LOGICAL' size is not 32
1406 bits, since `g77' is known to not work well for such cases (to be
1407 fixed in Version 0.6--*note Actual Bugs We Haven't Fixed Yet:
1410 * Lots of new documentation (though work is still needed to put it
1411 into canonical GNU format).
1413 * Build `libf2c' with `-g0', not `-g2', in effect (by default), to
1414 produce smaller library without lots of debugging clutter.
1419 * Fix bad code generation involving `X**I' and temporary, internal
1420 variables generated by `g77' and the back end (such as for `DO'
1423 * Fix crash given `CHARACTER A;DATA A/.TRUE./'.
1425 * Replace crash with diagnostic given `CHARACTER A;DATA A/1.0/'.
1427 * Fix crash or other erratic behavior when null character constant
1428 (`''') is encountered.
1430 * Fix crash or other erratic behavior involving diagnosed code.
1432 * Fix code generation for external functions returning type `REAL'
1433 when the `-ff2c' option is in force (which it is by default) so
1434 that `f2c' compatibility is indeed provided.
1436 * Disallow `COMMON I(10)' if `I' has previously been specified with
1437 an array declarator.
1439 * New `-ffixed-line-length-N' option, where N is the maximum length
1440 of a typical fixed-form line, defaulting to 72 columns, such that
1441 characters beyond column N are ignored, or N is `none', meaning no
1442 characters are ignored. does not affect lines with `&' in column
1443 1, which are always processed as if `-ffixed-line-length-none' was
1446 * No longer generate better code for some kinds of array references,
1447 as `gcc' back end is to be fixed to do this even better, and it
1448 turned out to slow down some code in some cases after all.
1450 * In `COMMON' and `EQUIVALENCE' areas with any members given initial
1451 values (e.g. via `DATA'), uninitialized members now always
1452 initialized to binary zeros (though this is not required by the
1453 standard, and might not be done in future versions of `g77').
1454 Previously, in some `COMMON'/`EQUIVALENCE' areas (essentially
1455 those with members of more than one type), the uninitialized
1456 members were initialized to spaces, to cater to `CHARACTER' types,
1457 but it seems no existing code expects that, while much existing
1458 code expects binary zeros.
1463 * Don't emit bad code when low bound of adjustable array is
1464 nonconstant and thus might vary as an expression at run time.
1466 * Emit correct code for calculation of number of trips in `DO' loops
1467 for cases where the loop should not execute at all. (This bug
1468 affected cases where the difference between the begin and end
1469 values was less than the step count, though probably not for
1470 floating-point cases.)
1472 * Fix crash when extra parentheses surround item in `DATA'
1475 * Fix crash over minor internal inconsistencies in handling
1476 diagnostics, just substitute dummy strings where necessary.
1478 * Fix crash on some systems when compiling call to `MVBITS()'
1481 * Fix crash on array assignment `TYPEDDD(...)=...', where DDD is a
1482 string of one or more digits.
1484 * Fix crash on `DCMPLX()' with a single `INTEGER' argument.
1486 * Fix various crashes involving code with diagnosed errors.
1488 * Support `-I' option for `INCLUDE' statement, plus `gcc''s
1489 `header.gcc' facility for handling systems like MS-DOS.
1491 * Allow `INCLUDE' statement to be continued across multiple lines,
1492 even allow it to coexist with other statements on the same line.
1494 * Incorporate Bellcore fixes to `libf2c' through 1995-03-15--this
1495 fixes a bug involving infinite loops reading EOF with empty
1496 list-directed I/O list.
1498 * Remove all the `g77'-specific auto-configuration scripts, code,
1499 and so on, except for temporary substitutes for bsearch() and
1500 strtoul(), as too many configure/build problems were reported in
1501 these areas. People will have to fix their systems' problems
1502 themselves, or at least somewhere other than `g77', which expects
1503 a working ANSI C environment (and, for now, a GNU C compiler to
1504 compile `g77' itself).
1506 * Complain if initialized common redeclared as larger in subsequent
1509 * Warn if blank common initialized, since its size can vary and hence
1510 related warnings that might be helpful won't be seen.
1512 * New `-fbackslash' option, on by default, that causes `\' within
1513 `CHARACTER' and Hollerith constants to be interpreted a la GNU C.
1514 Note that this behavior is somewhat different from `f2c''s, which
1515 supports only a limited subset of backslash (escape) sequences.
1517 * Make `-fugly-args' the default.
1519 * New `-fugly-init' option, on by default, that allows
1520 typeless/Hollerith to be specified as initial values for variables
1521 or named constants (`PARAMETER'), and also allows
1522 character<->numeric conversion in those contexts--turn off via
1525 * New `-finit-local-zero' option to initialize local variables to
1526 binary zeros. This does not affect whether they are `SAVE'd, i.e.
1527 made automatic or static.
1529 * New `-Wimplicit' option to warn about implicitly typed variables,
1530 arrays, and functions. (Basically causes all program units to
1531 default to `IMPLICIT NONE'.)
1533 * `-Wall' now implies `-Wuninitialized' as with `gcc' (i.e. unless
1534 `-O' not specified, since `-Wuninitialized' requires `-O'), and
1535 implies `-Wunused' as well.
1537 * `-Wunused' no longer gives spurious messages for unused `EXTERNAL'
1538 names (since they are assumed to refer to block data program
1539 units, to make use of libraries more reliable).
1541 * Support `%LOC()' and `LOC()' of character arguments.
1543 * Support null (zero-length) character constants and expressions.
1545 * Support `f2c''s `IMAG()' generic intrinsic.
1547 * Support `ICHAR()', `IACHAR()', and `LEN()' of character
1548 expressions that are valid in assignments but not normally as
1551 * Support `f2c'-style `&' in column 1 to mean continuation line.
1553 * Allow `NAMELIST', `EXTERNAL', `INTRINSIC', and `VOLATILE' in
1554 `BLOCK DATA', even though these are not allowed by the standard.
1556 * Allow `RETURN' in main program unit.
1558 * Changes to Hollerith-constant support to obey Appendix C of the
1561 - Now padded on the right with zeros, not spaces.
1563 - Hollerith "format specifications" in the form of arrays of
1564 non-character allowed.
1566 - Warnings issued when non-space truncation occurs when
1567 converting to another type.
1569 - When specified as actual argument, now passed by reference to
1570 `INTEGER' (padded on right with spaces if constant too small,
1571 otherwise fully intact if constant wider the `INTEGER' type)
1572 instead of by value.
1574 *Warning:* `f2c' differs on the interpretation of `CALL FOO(1HX)',
1575 which it treats exactly the same as `CALL FOO('X')', but which the
1576 standard and `g77' treat as `CALL FOO(%REF('X '))' (padded with
1577 as many spaces as necessary to widen to `INTEGER'), essentially.
1579 * Changes and fixes to typeless-constant support:
1581 - Now treated as a typeless double-length `INTEGER' value.
1583 - Warnings issued when overflow occurs.
1585 - Padded on the left with zeros when converting to a larger
1588 - Should be properly aligned and ordered on the target machine
1589 for whatever type it is turned into.
1591 - When specified as actual argument, now passed as reference to
1592 a default `INTEGER' constant.
1594 * `%DESCR()' of a non-`CHARACTER' expression now passes a pointer to
1595 the expression plus a length for the expression just as if it were
1596 a `CHARACTER' expression. For example, `CALL FOO(%DESCR(D))',
1597 where `D' is `REAL*8', is the same as `CALL FOO(D,%VAL(8)))'.
1599 * Name of multi-entrypoint master function changed to incorporate
1600 the name of the primary entry point instead of a decimal value, so
1601 the name of the master function for `SUBROUTINE X' with alternate
1602 entry points is now `__g77_masterfun_x'.
1604 * Remove redundant message about zero-step-count `DO' loops.
1606 * Clean up diagnostic messages, shortening many of them.
1608 * Fix typo in `g77' man page.
1610 * Clarify implications of constant-handling bugs in `f/BUGS'.
1612 * Generate better code for `**' operator with a right-hand operand of
1615 * Generate better code for `SQRT()' and `DSQRT()', also when
1616 `-ffast-math' specified, enable better code generation for `SIN()'
1619 * Generate better code for some kinds of array references.
1621 * Speed up lexing somewhat (this makes the compilation phase