Guile-SDL 0.4.0 available

[Thien-Thi Nguyen <ttn@gnuvola.org>]

release notes:

  The price of portability is some increased ugliness;
  hackers please see ‘%%Guile-SDL-obtw’ and suggest alternatives.

  Special thanks go to Brian Templeton for the patch that inspired my
  change of heart wrt Guile 1.8.  (That code is not in Guile-SDL, but
  the ideas it expresses are.)  It has been a long while, but here we
  are at last.

  Please look carefully at the NEWS entries marked [I].  Most likely
  further 0.4.x releases will have more; perhaps they can be avoided for
  0.5.x and later.

  thi

README excerpt:

  This directory contains Guile-SDL, a set of modules that provide
  bindings for SDL (Simple DirectMedia Layer) to enable Guile
  programmers to do all the nice things you can do with SDL.

  Most of the SDL functions have been wrapped with the exception of a
  few functions that were too C-centric.  The SDL Threads and the Audio
  functions haven't been included because of the problems with Guile
  code being run from more than one thread.  However audio programming
  can be done with the module ‘(sdl mixer)’ that requires the SDL_mixer
  library.

  Also included is SDL_gfx 2.0.22 (by Andreas Schiffler) and bindings
  for it.

NEWS excerpt:

  - 0.4.0 | 2011-12-15

    - improved portability

        Guile-SDL is now also tested with Guile-1.8.7 on a 64-bit host.
        It should build and run without any deprecation warnings; if use
        under ‘GUILE_WARN_DEPRECATED=detailed’ reveals something, please
        report that as a bug.

        Notably, for Guile versions that do not support "direct shared-
        object loading" (i.e., Guile 1.8+), configuration arranges for a
        module's sofile to be installed in $(libdir)/guile-sdl -- aka
        $(pkglibdir) -- and its Scheme wrapper (that dynamically loads
        the sofile and handles ‘define-module’ duties) to be installed in
        the "normal" site directory.

        See new file HACKING for details.

    - validation

      - use Free{SansBold,SerifBoldItalic}.ttf for testing

        The testing (distributed but not installed) font files are now
        FreeSansBold.ttf and FreeSerifBoldItalic.ttf (see AUTHORS).  The
        file crystal.ttf is no longer included due to doubts about its
        lineage and licensing.

      - support for ‘VERBOSE=1’

        By default, "make check" is relatively silent.  You can use
        "make check VERBOSE=1" to see more information.

      - support for ‘NOMMX=1’

        You can disable MMX usage dynamically with "make check NOMMX=1".
        This might be necessary if test/gfx.scm segfaults.  :-/

      - infrastructure upgrade

        Doing "make check" now prepares a "simulated installation"
        (under test/.simu) for better approximation to a real-world,
        post-install environment.  It is an approximation only because
        documentation is not included.

        Tests are now run by a driver program which DTRT, we hope.
        Also, they have full access to Guile-BAUX modules.

      - support for "make installcheck"

        (You need to "make install" first!)  Since ‘installcheck’
        piggybacks on ‘check’, it honors ‘INTERACTIVE=1’, etc.

      - tests delay less

        This makes running "make check" less weary for the impatient.

      - tests explicitly ‘exit’

        Moreover, the exit value is taken from the ‘(sdl sdl) quit’
        return value.  This helps detect problems earlier.

      - different ‘#:prefix’ for each module

        Previously, some tests would use the same import-prefix for
        different modules.  Now they are distinct, for clarity.

      - cursor test ignores ‘INTERACTIVE=1’

        There was really no good reason for it to honor ‘INTERACTIVE=1’
        in the first place.

      - better coverage, though still incomplete

        The test suite now includes ‘t99-cov’, marked as ‘xfail’.
        There are still 189 "slighted" procedures.

    - planned retirement: integer arg for ‘grab-input’

        Support for integer ‘mode’ to proc ‘(sdl sdl) grab-input’ WILL BE
        REMOVED after 2012-05-31, such that an integer ‘mode’ will result
        in a ‘wrong-type-arg’ error.

        Migration is straightforward; use a symbol instead of an integer.
        was: -1     now: query
              0          off
              1          on

    - planned retirement: ordinary vector argument(s)

        Some procedures now take uniform vector arguments (see below).
        For backward compatability with Guile-SDL 0.3.10 and earlier,
        you can also use an ordinary vector.  This is expensive and less
        safe, however, so this support WILL BE REMOVED after 2012-12-31.

        Luckily, migration is simple: Import module ‘(srfi srfi-4)’ and
        replace ‘list->vector’ with ‘list->u16vector’, and so on.

    - bugfixes

        Backward-incompatible changes are marked with "[I]".

      - distribute ‘SDL_gfx’ LICENSE

        I.e., the GNU LGPL 2.1.

      - no MMX support if not in "32-bit mode"

        The MMX support (in SDL_gfx) was previously enabled for any
        processor that reported support for it.  However, the code
        itself makes use of instructions (e.g., PUSHA, POPA) that are
        not available in 64-bit mode, so we now don't even bother
        checking for MMX support for 64-bit processors.

      - [I] flags are symbols

        Previously the following procs returned a (possibly empty) list
        of strings.  Now, they return a (possibly empty) list of symbols.

        - (sdl sdl) flagstash-flags
        - (sdl sdl) number->flags
        - (sdl sdl) was-init
        - (sdl sdl) get-mod-state
        - (sdl sdl) event:key:keysym:mod
        - (sdl ttf) font:style

      - exact bitwise match required in ‘ulong2flags’

        This func, a subroutine of the procs listed above in entry
        "flags are symbols" (except for ‘flagstash-flags’), used to
        include a flag in the returned list if any of its constituent
        bits were set in the original number.  Now, it requires all set
        bits to match as a condition for inclusion.

      - correct proc name for wrong-type-arg error message

        The following procs used to mis-identify themselves (as another)
        in the ‘wrong-type-arg’ error message:

        - (sdl mixer) paused?
        - (sdl sdl) load-image
        - (sdl sdl) set-gamma-ramp

      - allocate enough memory for temp ‘SDL_Color’ array

        Procs ‘set-colors!’ and ‘set-palette’ used to corrupt memory by
        underallocating a temporary ‘SDL_Color’ array, then overfilling
        it.  Now, they allocate the right amount.

      - avoid ‘SDL_Surface’ double-free

        Previously, both ‘get-video-surface’ and ‘set-video-mode’ would
        return a smob whose free func would call ‘SDL_FreeSurface’ on
        the unpacked ‘SDL_Surface’ object.  This is incorrect because
        those objects are internal (and moreover currently in use),
        rather than freshly constructed for hand-off to the user; the
        smob free func should not call ‘SDL_FreeSurface’ on them.  That
        is the job for the internal cleanup invoked by ‘(sdl sdl) quit’.

        Now, the ‘SDL-Surface’ smob distinguishes between user and
        internal, and its free func calls ‘SDL_FreeSurface’ only for
        user objects.

      - ‘(sdl sdl) get-caption’ handles ‘NULL’ title, icon

        This used to segfault on ‘NULL’ title or icon.

    - other changes

        Backward-incompatible changes are marked with "[I]".  For procs
        that now take uniform vectors, see also the "planned retirement"
        NEWS entry above.

      - [I] arg type-check tightened from ‘exact?’ to ‘integer?’

        Procs that used to accept exact numbers now accept integers
        only.  This basically means rational numbers are now invalid.
        Observant tarball-perusers will note that the modules and tests
        no longer use ‘quotient’.

      - ‘(sdl gfx) imfi-mmx?’ now able to set MMX usage

        This now takes an optional arg ‘setting’ for en-/dis-abling MMX
        usage at runtime.

      - embedded ‘SDL_gfx’ upgraded to 2.0.22

        Unfortunately, some MMX-backed procs now routinely segfault.
        (More unfortunately, the segfault has not been investigated, and
        was in fact the motivation for extending ‘imfi-mmx?’, as a "make
        check" workaround.  Masking bugs via feature creep: suboptimal.)

      - new proc: ‘(sdl gfx) draw-rounded-rectangle’

        This draws a rectangle with rounded corners of specified radius.

      - new proc: ‘(sdl gfx) draw-thick-line’

        This draws a line of specified thickness (in pixels).

      - new proc: ‘(sdl gfx) font-rotation!’

        This causes subsequent ‘draw-character’ and ‘draw-string’ calls
        to render rotated images.  The symbolic specified rotation is
        one of ‘none’, ‘clockwise’, ‘counter-clockwise’, ‘upside-down’.

      - new enumstash: ‘(sdl gfx) font-rotations’

        This is for ‘font-rotation!’ support.

      - new proc: ‘(sdl misc-utils) exact-truncate’

        This is nothing but ‘(inexact->exact (truncate n))’ plus
        documentation, support and the special glow from association
        with other Guile-SDL procs. :-D

      - ‘#:prefix’ used for import renaming

        All ‘#:renamer’ uses were coincident with ‘symbol-prefix-proc’,
        so these have been replaced with ‘#:prefix’.  [Insert blurb
        about ‘symbol-prefix-proc’ regret, here.]

      - some ‘(sdl gfx)’ procs take s16 uniform vectors

        Coordinate arguments for these procs can now be s16 uniform
        vectors.  You can even mix normal w/ s16 (but that's crazy).
        - draw-polygon
        - draw-aa-polygon
        - draw-textured-polygon
        - draw-bezier

      - ‘(sdl sdl) create-cursor’ takes u8 uniform vectors

        The ‘data’ and ‘mask’ arguments can now be u8 uniform vectors.

      - ‘(sdl sdl) set-gamma-ramp’ takes u16 uniform vectors

        The red, green and blue table arguments can now be u16 uniform
        vectors.

      - [I] ‘(sdl sdl) get-gamma-ramp’ now returns list of uvec

        Previously, this proc returned an alist.  Now, it returns a list
        of three elements, each of which is a u16 uniform vector.

      - [I] ‘(sdl sdl) show-cursor’ redesigned

        The optional arg is no longer a "toggle enabler" (which didn't
        seem to be implemented correctly, anyway) but instead a direct
        "setting".  In other words:

        was: show: (or  (show-cursor) (show-cursor #t)) ; buggy?
             hide: (and (show-cursor) (show-cursor #t)) ; buggy?
             flip: (show-cursor #t)                     ; buggy?

        now: show: (show-cursor #t)
             hide: (show-cursor #f)
             flip: (show-cursor (not (show-cursor)))

    - documentation improvements
      - manual license now GNU FDL 1.3
      - table of contents now after title page
      - getter/setter procs documented
      - SDL type no longer mentioned in access procs
      - todo list dropped

    - tools for bootstrap upgraded
      - no longer need gperf(1) to bootstrap
      - GNU Automake 1.11.1
      - GNU Autoconf 2.68
      - GNU Libtool 2.4.2
      - GNU texinfo 4.13
      - Guile-BAUX 20111207.1126.1d610b3
      - snuggle.m4 serial 5

tarball and its detached signature in dir:

  http://download.savannah.gnu.org/releases/guile-sdl/

homepage:

  http://www.nongnu.org/guile-sdl/