gcl-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gcl-devel] 2.6.1-19 is out -- release candidate for 2.6.2


From: Camm Maguire
Subject: [Gcl-devel] 2.6.1-19 is out -- release candidate for 2.6.2
Date: 30 Dec 2003 16:25:49 -0500
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2

Greetings!  Just a quick note that I've just uploaded gcl 2.6.1-19 and
gclcvs 2.7.0-16 to debian.org.  Please check 2.6.1-19 out.  I'd like
to have it thoroughly tested, especially by the people recently
working on the mingw port, tag and release it as 2.6.2, and then turn
attention almost exclusively toward the unstable branch.

Also would appreciate feedback on the gprof profiling support.

>From the changelog:

gcl (2.6.1-19) unstable; urgency=low

  * Fix bug in compiler::c2labels in which *ccb-vs* was missing a ocal
    rebind
  * Remove duplicate tags from compiled C switch statements
  * Minor merges for DARWIN support
  * Path to configure to make --enable-emacsdir work
  * Check for readline/readline.h header before configuring for readline
  * Improve system bfd library location detection
  * Make sure external gmp lib is compatible via __GNU_MP_VERSION, else
    backoff to local gmp build;  prepend externally defined CFLAGS into
    output CFLAGS, FINAL_CFLAGS, and NIFLAGS
  * Remove --enable-gmp configure option; gmp is required for GCL
  * Use --enable-emacsdir in debian/rules, make sure --enable-emacsdir
    and --enable-infodir work when arg contains ${prefix}
  * Fix typo in chap-6.texi
  * Make sure to export SGC define from config.h to cmpinclude.h -- Now
    that we used optimized structures in the compiler, we need at least
    the definition of SGC_TOUCH there to prevent GBC errors.  FIXME --
    handle header dependencies more robustly.  Thanks to Robert Boyer
    for the report
  * Improve SGC define extraction for cmpinclude.h
  * Fix variable reference errors which were occurring for compiled
    local functions defined within closure-generating or other
    environment stack pushing functions when safety is set to 3 (thanks
    Paul Dietz for the report.).  When constructing local functions and
    closures within a 'mother' function, *ccb-vs* will hold the number
    of closure environments stacked at the point of each closure
    creation or call to a local function.  This value is stored as the
    cadr of a list pushed onto *local-funs*, and is read when writing
    out the C code for the local function or closure, where it is used
    to initialize *ccb-vs* and *initial-ccb-vs* for subsequent
    processing.  The latter is used as the reference point when
    addressing variables in wt-ccb-vs, as the former could be still
    further incremented within the closure or local function itself.
    Local functions as opposed to closures do not increment *ccb-vs* and
    do not push the environment.  When a local function is defined
    within a closure-generating flet/labels, or a tagbody or block which
    pushes the environment, the value of *ccb-vs* written to the list
    corresponding to the local function can be erroneously incremented
    beyond the *initial-ccb-vs* value established before any environment
    pushing operations were processed.  It is this latter value which is
    appropriate for use in wt-ccb-vs, as the local functions, unlike the
    closures, receive an environment level with the mother generating
    function.  We therefore push *initial-ccb-vs* onto the end the list
    pushed onto *local-funs* only when defining a local function, and use
    it to initialize an added optional variable initialize-ccb-vs in
    t3local-fun and t3local-dcfun, which default to the original ccb-vs.
    We then bind *initial-ccb-vs* to this new optional parameter instead
    of the former *ccb-vs, which was only appropriate for closures.
  * Put in rudimentary logic for the selection of stack vs. heap storage
    for bignums depending on the frame context.  FIXME, this logic is
    too conservative at present.  SETQ_II and SETQ_IO take an additional
    parameter which is malloc when *unwind-exit* is bound and contains
    'frame and  alloca otherwise.  New macro bignum-expansion-storage.
    FIXME, ensure that IDECL does not need similar modification.
  * Cleanup a few compiler warnings in the compiler
  * Cleanup compiler warning in alloc.c
  * Eliminate unneeded transformatio of contniguous pages to other pages
    on save-system.
  * malloc -> gcl_gmp_alloc in recent setjmp frame protected bignum
    allocation
  * Add -Wa,--execstack if on an exec-shield enabled system, can be
    explicitly added otherwise by setting the CFLAGS variable before the
    configure step
  * Better execstack flag handling in configure
  * Allow for commas in CFLAGS in sed command writing *cc*
  * Preliminary gprof profiling support
  * Rework html documentation generation and installation, Closes:
    #221774
  * Remove parentheses from setf class-name info node in chap-7.texi

 -- Camm Maguire <address@hidden>  Tue, 30 Dec 2003 16:26:45 +0000



Take care,
-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah




reply via email to

[Prev in Thread] Current Thread [Next in Thread]