NEWS

   Copyright (C) 2019, 2020, 2021, 2022, 2023, 2024, 2025, 2026
   Free Software Foundation, Inc.
   
   Copying and distribution of this file, with or without modification,
   are permitted in any medium without royalty provided the copyright
   notice and this notice are preserved.

Changes from 5.4.x to 6.0.0
---------------------------

1. The --lint-old option is now gone. "New" awk is almost 40 years
   old, it's about time to stop worrying about old awk.

2. Similarly, the scantily-documented --persist option is now gone.

3. Similarly, the --re-interval option is now gone. It had been a no-op
   for a long time.

4. The configure-time option --enable-builtin-intdiv0 and the related
   code and documentation have been removed.  Use the intdiv loadable
   extension if you need that functionality.

Changes from 5.4.0 to 5.4.1
---------------------------

1. A performance problem in PMA with backing store files 1G or more in
   size has been fixed, by disabling assertions in support/pma.c.

2. A bug in gensub() when using MinRX has been fixed.

3. All files that referred to the FSF's old address have been revised.

4. After careful review of the POSIX standard and how MinRX handles
   matching when ignoring case, it was decided that the way it worked
   was already correct, even for cases where one character can have
   multiple alternative case variants. README_d/README.matchers has
   been updated, and several new tests for this issue have been added
   to the test suite.

5. Unary plus and minus now force their operands to be numeric.

6. The intdiv extension has been in the distribution for several years,
   but was lacking both a man page and a description in the manual.
   These flaws have been corrected.

7. The Georgian translation (ka.po) was accidentally disabled. It's now
   enabled, and also updated.

8. Parse time division by zero is no longer a fatal error, but rather
   a lint warning.  Division by zero at runtime remains fatal.

NN. As usual, a number of small bugs have been fixed; see the ChangeLog
    for the details.

Changes from 5.3.x to 5.4.0
---------------------------

1. This release now uses Mike Haertel's MinRX regular expression matcher
   as the default regexp engine. The old regex and dfa engines are still
   available. More detail is available in the manual, and in the file
   README_d/README.matchers. At the very least, read that file!

2. The manual, in the Bugs section, now makes it explicit that
   (a) Ad hominem attacks on the lists will not be tolerated, and
   (b) Discussion of proprietary software is strongly discouraged.
   Repeated offenses are grounds for being banned from the lists.

3. There is now a new directive, @nsinclude, which works like @include
   but does not reset the namespace for the included file to "awk". See
   the manual for details.

4. When using lshift() or rshift() and attempting to shift by as many
   or more bits than in a uintmax_t, gawk returns zero, instead of
   whatever the C compiler and hardware might have done.

5. Gawk's use of persistent memory has changed somewhat:
   A. Gawk now stores additional meta-information in the backing file.
      This means that if you have a backing file with important data
      in it, you should dump the data to a text file using the old version,
      create a new backing file, and then read your data back in with
      the new version, to a *brand new* backing file.
   B. Gawk generates a warning if the version of gawk saved in the backing
      file doesn't match that of the current running gawk.
   C. It's now possible to use persistent memory and dynamic extensions
      without problems. Gawk notices if an extension is being loaded from
      a different path than what was first used and produces a fatal error
      in this case.

6. The ordchr extension now supports multibyte / wide characters.

7. Per the 2024 POSIX standard, `length(array)' is no longer an extension,
   but a regular feature.  Thus --posix no longer rejects it and --lint
   no longer warns about it.

8. The --traditional option has been rationalized to bring gawk into
   sync with BWK awk. It no longer affects the return code from system(),
   and it no longer prevents using a regexp for RS. Internally, the
   code was cleaned up some as well.

9. Assertions in the C code are now enabled.  To disable them, manually
   edit the various Makefiles after running configure and before
   running make. You will need to add -DNDEBUG to the CFLAGS variable.

10. PMA should now work on OpenBSD 7, FreeBSD 12 - 16, NetBSD 10 and 11,
    and MidnightBSD 3 and 4.

11. Hexadecimal floating-point values may now be used in program source code,
    with strtonum(), and with the -n/--non-decimal-data option.  See the
    manual for details.

12. A large number of small "replacement" files for standard functions
    have been removed. These functions are now so standard that we
    simply expect them to always be available. This simplifies the
    distribution and the code maintenance.

13. Support for UDP in gawk's networking support is now obsolete.
    It never worked very well. It will be removed in version 6.0.
    Gawk issues a warning when attempting to use it.

14. Reading regular disk input files should be somewhat faster now,
    since gawk no longer checks for timeouts on such files.  On one
    very large file, gawk '{ print }' saw approximately a 9% speedup.

15. The MinGW port of gawk for MS-Windows now supports UTF-8 encoded
    non-ASCII text when the console window where gawk runs uses the
    Windows codepage 65001 for output, even if the system-wide locale
    specifies another codepage.

    Similarly, the Cygwin port now also fully supports UTF-8.

16. There is a new option to configure: --enable-O3.  This causes gcc to
    use -O3 instead of -O2 when compiling gawk.  This is not the default
    because experience in some projects has shown (sadly) that -O3 can cause
    bugs.

17. There is a new translation: Arabic. The .gmo files for the ca, da, fi,
    ja, ka, ms, and vi translations are no longer built or included in
    the distribution, as those translations have gone too long without
    being updated.  The .po files remain in the distribution, should
    any volunteers wish to come forward to update them.

18. OpenVMS support has been updated. This release builds on
    Alpha, Itanium and x86_64.

19. As usual, a number of small bugs have been fixed; see the ChangeLog
    for the details.

Changes from 5.3.2 to 5.3.x
---------------------------

1. The Hebrew translation has been revived.

2. All non-standard variables are now not installed for --traditional
   and --posix.

3. It's been discovered that persistent memory and dynamic extensions don't mix.
   For now, trying this combination produces a fatal error. It may one day
   get fixed. Or, it may not.

4. A bug in the API has been fixed whereby using a numeric index to set an
   array element will work. As a result, the API minor version was increased to 1.

Changes from 5.3.1 to 5.3.2
---------------------------

1. The pretty printer now produces fewer spurious newlines; at the
   outermost level it now adds newlines between block comments and
   the block or function that follows them. The extra final newline
   is no longer produced.

2. OpenVMS 9.2-2 x86_64 is now supported.

3. On Linux and macos systems, the -no-pie linker flag is no longer required.
   PMA now works on macos systems with Apple silicon, and not just
   Intel systems.

4. Still more subtle issues related to uninitialized array elements have
   been fixed.

5. Associative arrays should now not grow quite as fast as they used to.

6. The code and documentation are now consistent with each other with
   respect to path searching and adding .awk to the filename. Both
   are always done, even with --posix and --traditional.

7. As usual, there have been several minor code cleanups and bug fixes.
   See the ChangeLog for details.

Changes from 5.3.0 to 5.3.1
---------------------------

1. More subtle issues related to uninitialized array elements have
   been fixed.

2. A number of bugs in the debugger related to handling of arrays
   have been fixed.

3. Some subtle bugs in the API have been fixed.

4. Use of MPFR is now possible again on 32-bit Power PC Mac systems.

5. Race conditions around broken pipes for system() and read and write
   pipes should now be closed off.

6. Support for OSF/1 has been removed.

7. The never-documented --nostalgia option has been removed. It was
   causing bug reports.

8. The implementation of printf/sprintf has been thoroughly reworked
   in order to make the code more maintainable and to fix a goodly
   number of corner cases.

9. As usual, there have been several minor code cleanups and bug fixes.
   See the ChangeLog for details.

Changes from 5.2.x to 5.3.0
---------------------------

1. Infrastructure changes: Removed the use of libsigsegv. The
   value-add was never very much and it caused problems in some
   environments.

2. In keeping with new features in BWK awk, gawk now has built-in
   CSV file parsing. The behavior is intended to be identical to
   that of the "One True AWK" when --csv is applied. See the
   manual for details.

3. Also in keeping with BWK awk, gawk now supports a new \u escape
   sequence. This should be followed by 1-8 hexadecimal digits. The
   given code point is converted to its corresponding multibyte encoding
   for storage inside gawk. See the manual.

4. If PROCINFO["BUFFERPIPE"] exists, then pipe output is buffered.
   You can also use PROCINFO["command", "BUFFERPIPE"]. See the manual
   for details.

5. Because of the additional `do_csv' variable in the API, which breaks
   binary compatibility, the API major version was updated to 4 and
   the minor version was reset to zero.  The API remains source code
   compatible; that is, existing extensions should only require recompilation.

6. The manual now requires Texinfo 7.1 and its texinfo.tex for formatting.
   As a result, we no longer need to pre-process it, removing the need
   for gawktexi.in and leaving just gawk.texi.

7. And of course, there have been several minor code cleanups and bug fixes.
   See the ChangeLog for details.

Changes from 5.2.2 to 5.2.x
---------------------------

1. The readdir extension has been updated with additonal code and
   features, see the manual or its man page. As a result, the
   readdir_test.c extension has been removed.

2. We have a new translation: Ukrainian.

3. Several subtle issues related to null regexp matches around
   multibyte characters have been fixed.

Changes from 5.2.1 to 5.2.2
---------------------------

1. Infrastructure upgrades: makeinfo 7.0.1 must be used to format
   the manual.  As a result, the manual can also now be formatted
   with LaTeX by running it through `makeinfo --latex'.

2. Gawk no longer builds an x86_64 executable on M1 macOS systems.
   This means that PMA is unavailable on those systems.

3. Gawk will now diagnose if a heap file was created with a different
   setting of -M/--bignum than in the current invocation and exit with
   a fatal message if so.

4. Gawk no longer "leaks" its free list of NODEs in the heap file, resulting
   in much more efficient usage of persistent storage.

5. PROCINFO["pma"] exists if the PMA allocator is compiled into gawk.
   Its value is the PMA version.

6. The time extension is no longer deprecated. The strptime() function
   from gawkextlib's timex extension has been added to it.

7. Better information is passed to input parsers for when they want to
   decide whether or not to take control of a file. In particular, the
   readdir extension is simplified for Windows because of this.

8. The various PNG files are now installed for Info and HTML. The
   images files now have gawk_ prefixed names to avoid any conflicts
   with other installed PNG file names.

9. As usual, there have been several minor code cleanups and bug fixes.
   See the ChangeLog for details.

Changes from 5.2.0 to 5.2.1
---------------------------

1. Infrastructure upgrades: PMA version Avon 8.

2. Issues related to the sign of NaN and Inf values on RiscV have
   been fixed; gawk now gives identical results on that platform as
   it does on others.

3. A few issues with the debugger have been fixed.

4. More subtle issues with untyped array elements being passed to
   functions have been fixed.

5. The rwarray extension's readall() function has had some bugs fixed.

6. The PMA allocator is now supported on FreeBSD, OpenBSD and Linux on S/390x.
   It is now supported also on both Intel and M1 macOS systems.

7. There have been several minor code cleanups and bug fixes. See the
   ChangeLog for details.

Changes from 5.1.x to 5.2.0
---------------------------

*****************************************************************************
* MPFR mode (the -M option) is now ON PAROLE.  This feature is now being    *
* supported by a volunteer in the development team and not by the primary   *
* maintainer.  If this situation changes, then the feature will be removed. *
* For more information see this section in the manual:                      *
* https://www.gnu.org/software/gawk/manual/html_node/MPFR-On-Parole.html    *
*****************************************************************************

1. Infrastructure upgrades: Libtool 2.4.7, Bison 3.8.2.

2. Numeric scalars now compare in the same way as C for the relational
   operators. Comparison order for sorting has not changed.  This only
   makes a difference when comparing Infinity and NaN values with
   regular numbers; it should not be noticeable most of the time.

3. If the AWK_HASH environment variable is set to "fnv1a" gawk will
   use the FNV1-A hash function for associative arrays.

4. The CMake infrastructure has been removed. In the five years it was in
   the tree, nobody used it, and it was not updated.

5. There is now a new function, mkbool(), that creates Boolean-typed
   values.  These values *are* numbers, but they are also tagged as
   Boolean. This is mainly for use with data exchange to/from languages
   or environments that support real Boolean values. See the manual
   for details.

6. As BWK awk has supported interval expressions since 2019, they are
   now enabled even if --traditional is supplied. The -r/--re-interval option
   remains, but it does nothing.

7. The rwarray extension has two new functions, writeall() and readall(),
   for saving / restoring all of gawk's variables and arrays.

8. The new `gawkbug' script should be used for reporting bugs.

9. The manual page (doc/gawk.1) has been considerably reduced in size.
   Wherever possible, details were replaced with references to the online
   copy of the manual.

10. Gawk now supports Terence Kelly's "persistent malloc" (pma),
    allowing gawk to preserve its variables, arrays and user-defined
    functions between runs. THIS IS AN EXPERIMENTAL FEATURE!
   
    For more information, see the manual. A new pm-gawk.1 man page
    is included, as is a separate user manual that focuses on the feature.

11. Support for OS/2 has been removed. It was not being actively
    maintained.

12. Similarly, support for DJGPP has been removed. It also was not
    being actively maintained.

13. VAX/VMS is no longer supported, as it can no longer be tested.
    The files for it remain in the distribution but will be removed
    eventually.

14. Some subtle issues with untyped array elements being passed to
    functions have been fixed.

15. Syntax errors are now immediately fatal. This prevents problems
    with errors from fuzzers and other such things.

16. There have been numerous minor code cleanups and bug fixes. See the
    ChangeLog for details.

Changes from 5.1.1 to 5.1.x
---------------------------

1. Infrastructure upgrades: Automake 1.16.5, Texinfo 6.8.

2. The rwarray extension now supports writing and reading GMP and
   MPFR values. As a result, a bug in the API code was fixed.

Changes from 5.1.0 to 5.1.1
---------------------------

1. Infrastructure upgrades: Bison 3.8, Gettext 0.20.2, Automake 1.16.4,
   and (will wonders never cease) Autoconf 2.71.

2. asort and asorti now allow FUNCTAB and SYMTAB as the first argument if a
   second destination array is supplied. Similarly, using either array as
   the second argument is now a fatal error. Additionally, using either
   array as the destination for split(), match(), etc. also causes a
   fatal error.

3. The new -I/--trace option prints a trace of the byte codes as they
   are executed.

4. A number of subtle bugs relating to MPFR mode that caused differences
   between regular operation and MPFR mode have been fixed.

5. The API now handles MPFR/GMP values slightly differently, requiring
   different memory management for those values. See the manual for the
   details if you have an extension using those values.  As a result,
   the minor version was incremented.

6. $0 and the fields are now cleared before starting a BEGINFILE rule.

7. The duplication of m4 and build-aux directories between the main
   directory and the extension directory has been removed. This
   simplifies the distribution.

8. The test suite has been improved, making it easier to run the entire
   suite with -M. Use `GAWK_TEST_ARGS=-M make check' to do so.

9. Profiling and pretty-printing output has been modified slightly so
   that functions are presented in a reasonable order with respect
   to the namespaces that contain them.

10. Several example programs in the manual have been updated to their
    modern POSIX equivalents.

11. A number of examples in doc/gawkinet.texi have been updated for
    current times. Thanks to Juergen Kahrs for the work.

12. Handling of Infinity and NaN values has been improved.

13. There has been a general tightening up of the use of const and
    of types.

14. The "no effect" lint warnings have been fixed up and now behave
    more sanely.

15. The manual has been updated with much more information about what is
    and is not a bug, and the changes in the gawk mailing lists.

16. The behavior of strongly-typed regexp constants when passed as the
    third argument to sub() or gsub() has been clarified in the code and
    in the manual.

17. Similar to item #4 above, division by zero is now fatal in MPFR
    mode, as it is in regular mode.

18. There have been numerous minor code cleanups and bug fixes. See the
    ChangeLog for details.

Changes from 5.0.1 to 5.1.0
---------------------------

1. The major version of the API is bumped to 3, something that should
   have happened at the 5.0.0 release but didn't.

2. A number of memory leak issues have been fixed.

3. Infrastructure upgrades: Bison 3.5.4, Texinfo 6.7, Gettext 0.20.1,
   Automake 1.16.2.

4. The indexing in the manual has been thoroughly revised, in particular
   making use of the facilities in Texinfo 6.7.  That version (or newer)
   must be used to format the manual.

5. MSYS2 is now supported out-of-the-box by configure.

6. Several bugs have been fixed.  See the ChangeLog for details.

Changes from 5.0.0 to 5.0.1
---------------------------

1. A number of ChangeLog.1 files that were left out of the distribution
   have been restored.

2. Multiple syntax errors should no longer be able to cause a core dump.

3. Sandbox mode now disallows assigning new filename values in ARGV that
   were not there when gawk was invoked.

4. There are many small documentation improvements in the manual.

5. The new argument "no-ext" to --lint disables ``XXX is a gawk extension''
   lint warnings.

6. Infrastructure upgrades: Bison 3.4.

7. A number of bugs, some of them quite significant, have been fixed.
   See the ChangeLog for details.

Changes from 4.2.1 to 5.0.0
---------------------------

1. Support for the POSIX standard %a and %A printf formats has been added.

2. The test infrastructure has been greatly improved, simplifying the
   contents of test/Makefile.am and making it possible to generate
   pc/Makefile.tst from test/Makefile.in.

3. The regex routines have been replaced with those from GNULIB, allowing
   me to stop carrying forward decades of changes against the original
   ones from GLIBC.

4. Infrastructure upgrades: Bison 3.3, Automake 1.16.1, Gettext 0.19.8.1,
   makeinfo 6.5.

5. The undocumented configure option and code that enabled the use of
   non-English "letters" in identifiers is now gone.

6. The `--with-whiny-user-strftime' configuration option is now gone.

7. The code now makes some stronger assumptions about a C99 environment.

8. PROCINFO["platform"] yields a string indicating the platform for
   which gawk was compiled.

9. Writing to elements of SYMTAB that are not variable names now
   causes a fatal error. THIS CHANGES BEHAVIOR.

10. Comment handling in the pretty-printer has been reworked almost completely
    from scratch. As a result, comments in many corner cases that were
    previously lost are now included in the formatted output.

11. Namespaces have been implemented! See the manual.  One consequence of this
    is that files included with -i, read with -f, and command line program
    segments must all be self-contained syntactic units. E.g., you can no
    longer do something like this:

   	gawk -e 'BEGIN {' -e 'print "hello" }'

12. Gawk now uses the locale settings for ignoring case in single byte
    locales, instead of hardwiring in Latin-1.

13. A number of bugs, some of them quite significant, have been fixed.
    See the ChangeLog for details.