ckccfg.txt

KERMIT工具 这在办公室下载不了,很多人都没有载不到.

   An alternative to curses is also available at compile time, but should
   be selected if your version of Kermit is to be run in local mode only
   in an ANSI terminal environment, for example on a desktop workstation
   that has an ANSI console driver. To select this option in place of
   curses, define the symbol MYCURSES:

  -DMYCURSES

   instead of CK_CURSES. The MYCURSES option uses built-in ANSI (VT100)
   escape sequences, and depends upon your terminal or console driver to
   interpret them correctly.

   In some C-Kermit builds, we replace printf() via #define printf...
   However, this can cause conflicts with the [n]curses header files.
   Various hacks are required to get around this -- see [51]ckutio.c,
   [52]ckufio.c, [53]ckuusx.c, [54]ckucmd.c, etc.

   [ [55]C-Kermit Home ] [ [56]Kermit Home ]
    ________________________________________________________________________

  4. CHARACTER SETS

   [ [57]Top ] [ [58]Contents ] [ [59]Next ] [ [60]Previous ]

   Since version 5A, C-Kermit has included support for conversion of
   character sets for Western European languages (i.e. languages that
   originated in Western Europe, but are now also spoken in the Western
   Hemisphere and other parts of the world), via ISO 8859-1 Latin
   Alphabet 1, for Eastern European languages (ISO Latin-2), Hebrew (and
   Yiddish), Greek, and Cyrillic-alphabet languages (ISO Latin/Cyrillic).
   Many file (local) character sets are supported: ISO 646 7-bit national
   sets, IBM code pages, Apple, DEC, DG, NeXT, etc.

   To build Kermit with no character-set translation at all, include
   -DNOCSETS in the CFLAGS. To build with no Latin-2, add -DNOLATIN2. To
   build with no Cyrillic, add -DNOCYRIL. To omit Hebrew, add -DNOHEBREW.
   If -DNOCSETS is *not* included, you'll always get LATIN1. To build
   with no KANJI include -DNOKANJI. There is presently no way to include
   Latin-2, Cyrillic, Hebrew, or Kanji without also including Latin-1.

   [61]Unicode support was added in C-Kermit 7.0, and it adds a fair
   amount of tables and code (and this is only a "Level 1" implementation
   -- a higher level would also require building in the entire Unicode
   database). On a PC with RH 5.2 Linux, building C-Kermit 7.0, we get
   the following sizes:

  NOCSETS NOUNICODE NOKANJI   Before    After                  
   [   ]    [   ]    [   ]    1329014   (Full)
   [   ]    [   ]    [ X ]    1325686   (Unicode but no Kanji)
   [   ]    [ X ]    [   ]    1158837   (All charsets except Unicode)
   [ X ]    [ x ]    [ x ]    1090845   (NOCSETS implies the other two)

   Note, by the way, that NOKANJI without NOUNICODE only removes the
   non-Unicode Kanji sets (Shift-JIS, EUC-JP, JIS-7, etc). Kanji is still
   representable in UCS-2 and UTF-8.

   [ [62]C-Kermit Home ] [ [63]Kermit Home ]
    ________________________________________________________________________

  5. APC EXECUTION

   [ [64]Top ] [ [65]Contents ] [ [66]Next ] [ [67]Previous ]

   The Kermit CONNECT and INPUT commands are coded to execute Application
   Program Command escape sequences from the host:

  <ESC>_<text><ESC>\

   where <text> is a C-Kermit command, or a list of C-Kermit commands
   separated by commas, up to about 1K in length.

   To date, this feature has been included in the OS/2, Windows, VMS,
   OS-9, and Unix versions, for which the symbol:

  CK_APC

   is defined automatically in [68]ckuusr.h. For OS/2, APC is enabled at
   runtime by default, for UNIX it is disabled. It is controlled by the
   SET TERMINAL APC command. Configuring APC capability into a version
   that gets it by default (because CK_APC is defined in [69]ckuusr.h)
   can be overridden by including:

  -DNOAPC

   on the CC command line.

   C-Kermit's autodownload feature depends on the APC feature, so
   deconfiguring APC also disables autodownload (it doesn't use APC
   escape sequences, but uses the APC switching mechanism internally).

   [ [70]C-Kermit Home ] [ [71]Kermit Home ]
    ________________________________________________________________________

  6. PROGRAM SIZE

   [ [72]Top ] [ [73]Contents ] [ [74]Next ] [ [75]Previous ]

   SECTION CONTENTS

  6.1. [76]Feature Selection
  6.2. [77]Changing Buffer Sizes
  6.3. [78]Other Size-Related Items
  6.4. [79]Space/Time Tradeoffs

   (Also see [80]Section 4)

   Each release of C-Kermit is larger than the last. On some computers
   (usually old ones) the size of the program prevents it from being
   successfully linked and loaded. On some others (also usually old
   ones), it occupies so much memory that it is constantly swapping or
   paging. In such cases, you can reduce C-Kermit's size in various ways,
   outlined in this section. The following options can cut down on the
   program's size at compile time by removing features or changing the
   size of storage areas.

   If you are reading this section because all you want is a small, fast,
   quick-to-load Kermit file-transfer application for the remote end of
   your connection, and the remote end is Unix based, take a look at
   G-Kermit:

  [81]http://www.columbia.edu/kermit/gkermit.html

  6.1. Feature Selection

   Features can be added or removed by defining symbols on the CC (C
   compiler) command line. "-D" is the normal CC directive to define a
   symbol so, for example, "-DNODEBUG" defines the symbol NODEBUG. Some C
   compilers might use different syntax, e.g. "-d NODEBUG" or
   "/DEFINE=NODEBUG". For C compilers that do not accept command-line
   definitions, you can put the corresponding #define statements in the
   file ckcsym.h, for example:

  #define NODEBUG

   The following table shows the savings achieved when building C-Kermit
   8.0 (Beta.04) with selected feature-deselection switches on an
   Intel-based PC with Red Hat Linux 7.0 and gcc 2.96. The sizes are for
   non-security builds. The fully configured non-security build is
   2127408 bytes.

  Option      Size    Savings Effect
  NOICP        545330   74.4% No Interactive Command Parser (command-line only)
  NOLOCAL     1539994   27.6% No making connections.
  NOXFER      1551108   27.1% No file transfer.
  IKSDONLY    1566608   26.4% Internet Kermit Server only.
  NOCSETS     1750097   17.7% No character-set conversion.
  NOSPL       1800293   15.4% No Script Programming Language.
  NONET       1808575   15.0% No making network connections.
  NOUNICODE   1834426   13.8% No Unicode character-set conversion.
  NOHELP      1837877   13.6% No built-in help text.
  NODEBUG     1891669   11.1% No debug log.
  NOFRILLS    1918966    9.8% No "frills".
  NOFTP       1972496    7.3% No FTP client.
  NODIAL      1984488    6.7% No automatic modem dialing.
  NOPUSH      2070184    2.7% No shell access, running external programs, etc.
  NOIKSD      2074129    2.5% No Internet Kermit Server capability.
  NOHTTP      2082610    2.1% No HTTP client.
  NOFLOAT     2091332    1.7% No floating-point arithmetic.
  NOCHANNELIO 2095978    1.5% No FOPEN/FREAD/FWRITE/FCLOSE, etc.
  MINIDIAL    2098035    1.4% No built-in support for many kinds of modems.
  NOSERVER    2098987    1.3% No server mode.
  NOSEXP      2105898    1.0% No S-Expressions.
  NOPTY       2117743    0.5% No pseudoterminal support.
  NORLOGIN    2121089    0.3% No RLOGIN connections.
  NOOLDMODEMS 2124038    0.2% No built-in support for old kinds of modems.
  NOSSH       2125696    0.1% No SSH command.

   And here are a few combinations

   Options Size Savings Effect
   NODEBUG NOICP NOCSETS NOLOCAL 281641 86.7% No debug log, parser,
   character sets, or making connections.
   NOICP NOCSETS NOLOCAL 376468 82.3% No parser, character sets, or
   making connections.
   NOICP NOCSETS NONET 427510 79.9% No parser, character sets, or network
   connections.
   NOSPL NOCSETS 1423784 33.1% No script language, or character sets.

   -DNOFRILLS removes various command synonyms; the following top-level
   commands: CLEAR, DELETE, DISABLE, ENABLE, GETOK, MAIL, RENAME, TYPE,
   WHO; and the following REMOTE commands: KERMIT, LOGIN, LOGOUT, PRINT,
   TYPE, WHO.

  6.2. Changing Buffer Sizes

   Most modern computers have so much memory that (a) there is no need to
   scrimp and save, and (b) C-Kermit, even when fully configured, is
   relatively small by today's standards.

   Two major factors affect Kermit's size: feature selection and buffer
   sizes. Buffer sizes affect such things as the maximum length for a
   Kermit packet, the maximum length for a command, for a macro, for the
   name of a macro, etc. Big buffer sizes are used when the following
   symbol is defined:

  BIGBUFOK

   as it is by default for most modern platforms (Linux, AIX 4 and 5,
   HP-UX 10 and 11, Solaris, etc) in [82]ckuusr.h. If your build does not
   get big buffers automatically (SHOW FEATURES tells you), you can
   include them by rebuilding with BIGBUFOK defined; e.g. in Unix:

  make xxxx KFLAGS=-DBIGBUFOK

   where xxxx is the makefile target. On the other hand, if you want to
   build without big buffers when they normally would be selected, use:

  make xxxx KFLAGS=-DNOBIGBUF

   There are options to control Kermit's packet buffer allocations. The
   following symbols are defined in [83]ckcker.h in such a way that you
   can override them by redefining them in CFLAGS:

  -DMAXSP=xxxx - Maximum send-packet length.
  -DMAXRP=xxxx - Maximum receive-packet length.
  -DSBSIZ=xxxx - Total allocation for send-packet buffers.
  -DRBSIZ=xxxx - Total allocation for receive-packet buffers.

   The defaults depend on the platform.

   Using dynamic allocation (-DDYNAMIC) reduces storage requirements for
   the executable program on disk, and allows more and bigger packets at
   runtime. This has proven safe over the years, and now most builds
   (e.g. all Unix, VMS, Windows, and OS/2 ones) use dynamic memory
   allocation by default. If it causes trouble, however, then omit the
   -DDYNAMIC option from CFLAGS, or add -DNODYNAMIC.

  6.3. Other Size-Related Items

   To make Kermit compile and load successfully, you might have to change
   your build procedure to:

    a. Request a larger ("large" or "huge") compilation / code-generation
       model. This is needed for 16-bit PC-based UNIX versions (most or
       all of which fail to build C-Kermit 7.0 and later anyway). This is
       typically done with a -M and/or -F switch (see your cc manual or
       man page for details).
    b. Some development systems support overlays. If the program is too
       big to be built as is, check your loader manual ("man ld") to see
       if an overlay feature is available. See the 2.10/2.11 BSD example
       in the UNIX makefile. (Actually, as of version 7.0, C-Kermit is
       too big to build, period, even with overlays, on 2.xx BSD).
    c. Similarly, some small and/or segment-based architectures support
       "code mapping", which is similar to overlays (PDP11-based VENIX
       1.0, circa 1984, was an example). See the linker documentation on
       the affected platform.

   It is also possible to reduce the size of the executable program file
   in several other ways:

    a. Include the -O (optimize) compiler switch if it isn't already
       included in your "make" entry (and if it works!). If your compiler
       supports higher levels of optimization (e.g. -O2 or higher number,
       -Onolimit (HP-UX), etc), try them; the greater the level of
       optimization, the longer the compilation and more likely the
       compiler will run out of memory. The the latter eventuality, some
       compilers also provide command-line options to allocate more
       memory for the optimizer, like "-Olimit number" in Ultrix.
    b. If your platofrm supports shared libraries, change the make entry
       to take advantage of this feature. The way to do this is, of
       course, platform dependent; see the NeXT makefile target for an
       example. some platforms (like Solaris) do it automatically and
       give you no choice. But watch out: executables linked with shared
       libraries are less portable than statically linked executables.
    c. Strip the program image after building ("man strip" for further
       info), or add -s to the LNKFLAGS (UNIX only). This strips the
       program of its symbol table and relocation information.
    d. Move character strings into a separate file. See the 2.11 BSD
       target for an example.

  6.4. Space/Time Tradeoffs

   There are more than 6000 debug() statements in the program. If you
   want to save both space (program size) and time (program execution
   time), include -DNODEBUG in the compilation. If you want to include
   debugging for tracking down problems, omit -DNODEBUG from the make
   entry. But when you include debugging, you have two choices for how
   it's done. One definition defines debug() to be a function call; this
   is cheap in space but expensive in execution. The other defines debug
   as "if (deblog)" and then the function call, to omit the function call
   overhead when the debug log is not active. But this adds a lot of
   space to the program. Both methods work, take your choice; IFDEBUG is
   preferred if memory is not a constraint but the computer is likely to
   be slow. The first method is the default, i.e. if nothing is done to
   the CFLAGS or in [84]ckcdeb.h (but in some cases, e.g. VMS, it is). To
   select the second method, include -DIFDEBUG in the compilation (and
   don't include -DNODEBUG).

   [ [85]C-Kermit Home ] [ [86]Kermit Home ]
    ________________________________________________________________________

  7. MODEM DIALING

   [ [87]Top ] [ [88]Contents ] [ [89]Next ] [ [90]Previous ]