Skip to content
Snippets Groups Projects
  1. Dec 27, 2010
  2. Dec 26, 2010
  3. Dec 22, 2010
  4. Dec 19, 2010
  5. Dec 13, 2010
  6. Dec 11, 2010
    • rtoy's avatar
      Speed up building on sparc. Time taken is now almost half! This was · 1f888eec
      rtoy authored
      caused by all the calls to stat in PROBE-FILE in LOCATE-DOMAIN-FILE
      for files that did not exist.  The default locale was C, so every
      message lookup was causing many stat's to non-exist files.  (There
      were over 1000 calls/sec on a 750 MHz sparc!)
      
      So we cache all the calls to PROBE-FILE in LOCATE-DOMAIN-FILE.  But
      just in case, we also allow the user to get at the hash table to
      examine it (GET-DOMAIN-FILE-CACHE) and also allow the user to clear it
      (CLEAR-DOMAIN-FILE-CACHE) in case new translations are added without
      restarting lisp.
      1f888eec
  7. Dec 09, 2010
    • rtoy's avatar
    • rtoy's avatar
      SUBSEQ was sometimes crashing lisp when the end index was less than · 874f052d
      rtoy authored
      the start.  This was due to one of two things:  The result sequence
      was created with a negative length, creating invalid objects, or
      accessing the invalid object would cause a segfault.
      
      code/seq.lisp:
      o Declare the type of LENGTH in MAKE-SEQUENCE-OF-TYPE better.  It's
        not a fixnum, but an index (non-negative fixnum).  This should catch
        any mistakes where we try to create sequences of negative length.
      o Explicitly catch invalid START and END indices in VECTOR-SUBSEQ* and
        LIST-SUBSEQ* and signal an error
      
      general-info/release-20c.txt:
      o Document bugfix.
      874f052d
  8. Dec 05, 2010
  9. Nov 30, 2010
  10. Nov 16, 2010
  11. Nov 12, 2010
  12. Nov 10, 2010
    • rtoy's avatar
      Add new slots to the backend to hold the foreign linkage space start · e4befc68
      rtoy authored
      and entry size.  Not yet used anywhere.
      
      This change needs a cross-compile; use boot-2010-11-1-cross.lisp as
      the cross-compile script.
      
      compiler/backend.lisp:
      o Add the two new slots to the backend.
      
      compiler/dump.lisp:
      o DUMP-DATA-MAYBE-BYTE-SWAPPING needs to handle (unicode) strings
      o DUMP-DATA-MAYBE-BYTE-SWAPPING should not swap bytes of a string.
        Genesis will make that happen.
      
      code/exports.lisp:
      o Export BACKEND-FOREIGN-LINKAGE-SPACE-START and
        BACKEND-FOREIGN-LINKAGE-ENTRY-SIZE.
      
      bootfiles/20b/boot-2010-11-1-cross.lisp:
      o Cross-compile script for this change.
      e4befc68
  13. Nov 08, 2010
  14. Nov 04, 2010
  15. Nov 02, 2010
  16. Oct 26, 2010
  17. Oct 13, 2010
    • rtoy's avatar
      Some changes to replace calls to gettext with _"" or _N"" for things · b22644d4
      rtoy authored
      compiled with and without Unicode.  This is needed so that the pot
      files have the same content for both unicode and non-unicode builds.
      (The _"" and _N"" are handled by the reader, so things that are
      conditionalized out still get processed, unlike using gettext.)
      b22644d4
  18. Oct 12, 2010
    • rtoy's avatar
      Make read-char and read-byte signal errors when given the wrong kind · 35aebeba
      rtoy authored
      of streams.  This is a change from current 20a and 20b behavior which
      didn't signal errors, but matches the behavior for releases 19f and
      earlier.
      
      But allow them to work on binary-text-streams.  This is the same
      behavior as before for binary-text-stream streams.
      
      However, read-sequence no longer allows reading from streams into
      arbitrary objects, unless the stream is a binary-text-stream stream.
      
      code/fd-stream-extfmt.lisp:
      o In %SET-FD-STREAM-EXTERNAL-FORMAT, only update
        fd-stream-in/fd-stream-out if we have a character or
        binary-text-stream stream.
      o Don't update the fd-stream-string-buffer or lisp-stream-in-buffer if
        we have a binary-text-stream because that will mess up how
        fast-read-char and fast-read-byte dispatch to do the right thing for
        binary-text-stream streams.
      
      code/fd-stream.lisp:
      o Set the fd-stream-in and fd-stream-bin slots appropriately depending
        on whether we have a character, binary, or binary-text-stream
        stream.
      o Only create the lisp-stream-in-buffer if we do NOT have a
        binary-text-stream.  (Binary streams didn't use the
        lisp-stream-buffer previously, so no change there.  Character
        streams use the lisp-stream-buffer and/or lisp-string-buffer.)
      o Set the fd-stream-flags appropriately for the kind of stream this
        is.  Checking a fixnum is faster than checking the type of a stream.
      
      code/struct.lisp:
      o Add FLAGS slot to LISP-STREAM so we can tell what kind of stream
        (character, binary, binary-text-stream) we have.
      
      code/sysmacs.lisp:
      o Change FAST-READ-CHAR so that if we have a have a binary or
        binary-text-stream stream, we dispatch to the fast-read-char methods
        to do the right thing, including signaling an error for the wrong
        kind of stream.
      o Change FAST-READ-BYTE so that if we do not have a binary stream, we
        dispatch to the fast-read-char method to do the right thing.
      
      compiler/dump.lisp:
      o With the above changes, we can no longer write characters to a
        binary stream, like a FASL file.  Make the fasl file a
        binary-text-stream so that we can.  (Alternatively, we could create
        the FASL header as a string, convert to octets and dump the octest
        to the file.  This is easier, and should still be fast for writing
        fasls.)
      35aebeba
  19. Sep 29, 2010
  20. Sep 26, 2010
    • rtoy's avatar
      Actually set the library search list to the saved value. Previously · 5871d292
      rtoy authored
      we weren't because *cmucl-lib* is almost always bound.  Normally
      *old-cmucl-library-search-list* is unbound, but when a core is made,
      it becomes bound.  We use that to indicate that the library search
      list should be set to *old-cmucl-library-search-list*.
      5871d292
  21. Sep 24, 2010
  22. Sep 21, 2010
  23. Sep 20, 2010
  24. Sep 19, 2010
    • rtoy's avatar
      o Move %STR, %STRX and %MATCH around so that we can inline them · 119f21c7
      rtoy authored
        (because they're so simple).
      o Add some comments for %STR.
      o Change implementation of %MATCH to be simpler and add comments on
        why we do what we do and explain what happens if we don't.
      o Handle completion of Hangul syllables better:
        - Match "Hangul_S" instead of "Hangul_Syllable" because there's
          #\Hangul_Single_Dot_Tone_Mark.
        - If we match "Hangul_S", try to complete some Hangul syllables so
          we don't fool slime into thinking "Hangul_Syllable_" is the only
          completion.  There are obviously more.
      o Handle completion of CJK Unified Ideographs better by trying to
        complete more so slime isn't fooled into thinking
        "CJK_Unified_Ideograph-" is the only possible completion.
      119f21c7
    • rtoy's avatar
      o Construction of the Hangul syllable codebook was wrong. To satisfy · dc4cdb68
      rtoy authored
        the constraints on the codebook, we just sort them in descreasing
        order of length.
      o In %MIP, it might happen that MISMATCH returns NIL, which means a
        match.  In this case, don't change the position.
      dc4cdb68
  25. Sep 18, 2010
    • rtoy's avatar
      Some Hangul syllables were left out of the Hangul syllable dictionary. · f2065a91
      rtoy authored
      Redo this by looping over all codepoints and selecting the codepoints
      that are Hangul syllables.
      f2065a91
    • rtoy's avatar
      code/unidata.lisp: · 820f2554
      rtoy authored
      o Update constants to Unicode version 5.2.0.
      
      i18n/unidata.bin:
      o Regenerated using Unicode version 5.2.0.
      820f2554
    • rtoy's avatar
      code/unidata.lisp: · 3d1d8295
      rtoy authored
      o Just add some comments on why we don't put the dictionaries in
        unidata.bin.
      o Print out some messages when building the hangul and cjk
        dictionaries so the user knows what's happening.
      
      tools/build-unidata.lisp:
      o Add some comments on the various parts of unidata.bin.
      3d1d8295
  26. Sep 17, 2010
    • rtoy's avatar
      exports.lisp: · 9563cc0b
      rtoy authored
      o Export STRING-TO-NFC, UNICODE-COMPLETE, and UNICODE-COMPLETE-NAME.
      
      unidata.lisp:
      o Add explicit exports.
      9563cc0b
    • rtoy's avatar
      Optimize the completion of the Hangul syllables and the CJK unified · d4b307df
      rtoy authored
      ideographs by using dictionaries.  (Should these dictionaries be part
      of unidata.bin so they don't have to be built at run time?  One the
      one hand, it makes things simpler, but unnecessarily bloats
      unidata.in.  I suspect the hangul syllables and cjk ideographs
      characters not not used very often.)
      
      o Change NODE-NEXT and CLOSE-NODE to have an optional parameter for
        the dictionary to use.
      o Update UNICODE-COMPLETE-NAME to pass the dictionary to NODE-NEXT and
        CLOSE-NODE.
      o Update UNICODE-COMPLETE to use the hangul syllable dictionary and
        the cjk ideograph dictionary when searching.
      o Fix typo in UNICODE-COMPLETE.
      o Add defvars for dictionaries for hangul syllables and cjk
        ideographs.
      o Add functions to build the hangul and cjk dictionaries.
      o Steal the implementations of BUILD-DICTIONARY, NAME-LOOKUP, and
        ENCODE-NAME from tools/build-unidata.lisp.
      d4b307df
    • rtoy's avatar
      Add support for character completion. This is primarily intended to · d4b888a2
      rtoy authored
      support character completion for slime.  The implementation is from
      Paul Foley, but some slight modifications by Raymond Toy to handle a
      few corner cases.
      
      o Modify SEARCH-DICTIONARY to take optional current and posn
        parameters so that SEARCH-DICTIONARY can be started from a different
        place.
      o Add UNICODE-COMPLETE, which is the main function for character name
        completion.
      o Add other support functions for UNICODE-COMPLETE.
      d4b888a2
    • rtoy's avatar
      o Fix typo in UNICODE-DECOMP. (It's hangul-syllable-p, not · 34af3581
      rtoy authored
        hangule-syllable-p.)
      o Move the computation of *reverse-hangule-choseong*,
        *reverse-hangul-jungseong*, and *reverse-hangul-jongseong* to its
        own routine.  Call it in UNICODE-NAME-TO-CODEPOINT.
      34af3581
Loading