Re: lynx-dev [PATCH][dev21] the binary size battle: disabling charsets

[Leonid Pauzner]

30-Mar-99 23:38 John Bley wrote:
> Here's an idea for allowing extra rope to hang yourself on binary size:
> disabling all the character sets except ISO-8859-1 and 7-bit ASCII
> approximations.  It makes for a roughly 79KB decrease (here).

Not so effective as you think:

1) just defining ENTITIES_HTML40_ONLY will gain you another 30K decrease.
2) you are disabling charsets from UCDomap.h which have no chartrans tables,
so you gain probably a hundred of bytes but the actual code for handling
of CJK and UTF-8 is spreaded across the binary and can easily weight
for 50K decrease if would #ifndef'ed properly (but this will be
a maintainance nightmare until the code will be rewritten entirely).
3) chartrans handling code (without the tables) weight > 150K of binary
so this is an obvious loss to live with it but without the tables.


> Perhaps an extra comment should be added on the configure
> line like "(not recommended)", since this is really only meant for those
> tight on space and would be a bad choice for general users.

> The wording of the comment, the character sets to in/exclude, heck, the
> whole idea are up for comments.

> * First try at --disable-most-charsets, keeps ISO-8859-1 and 7-bit ASCII
>   approximations in, leaves every other set out. (John Bley)

> --
> John Bley - address@hidden
> Duke '99 - English/Computer Science
>   Since English is a mess, it maps well onto the problem space,
>   which is also a mess, which we call reality.     - Larry Wall

> diff -Burp lynx2-8-2/config.hin lynx2-8-2-patched/config.hin
> --- lynx2-8-2/config.hin      Tue Mar 30 12:10:37 1999
> +++ lynx2-8-2-patched/config.hin      Tue Mar 30 23:13:09 1999
> @@ -20,6 +20,7 @@
>  #undef DISABLE_FINGER                /* CF_ARG_DISABLE(finger) */
>  #undef DISABLE_GOPHER                /* CF_ARG_DISABLE(gopher) */
>  #undef DISABLE_NEWS          /* CF_ARG_DISABLE(news) */
> +#undef DISABLE_MOST_CHARSETS /* CF_ARG_DISABLE(most-charsets) */
>  #undef DISP_PARTIAL          /* CF_ARG_ENABLE(partial) */
>  #undef DONT_TRACK_INTERNAL_LINKS /* CF_ARG_DISABLE(internal-links) */
>  #undef EBCDIC                        /* CF_EBCDIC */
> diff -Burp lynx2-8-2/configure.in lynx2-8-2-patched/configure.in
> --- lynx2-8-2/configure.in    Tue Mar 30 12:10:37 1999
> +++ lynx2-8-2-patched/configure.in    Tue Mar 30 23:12:52 1999
> @@ -792,6 +792,14 @@ CF_ARG_DISABLE(ftp,
>  AC_MSG_RESULT($use_ftp)
>  test $use_ftp != "no" && AC_DEFINE(DISABLE_FTP)

> +AC_MSG_CHECKING(if you want to exclude most charsets)
> +CF_ARG_DISABLE(most-charsets,
> +[  --disable-most-charsets disable most charsets],
> +      [use_most_charsets=yes],
> +      [use_most_charsets=no])
> +AC_MSG_RESULT($use_most_charsets)
> +test $use_ftp != "no" && AC_DEFINE(DISABLE_MOST_CHARSETS)
> +
>  dnl 
> --------------------------------------------------------------------------
>  dnl DirEd (directory-editor) options
>  dnl 
> --------------------------------------------------------------------------
> diff -Burp lynx2-8-2/src/UCdomap.c lynx2-8-2-patched/src/UCdomap.c
> --- lynx2-8-2/src/UCdomap.c   Tue Mar 30 12:10:37 1999
> +++ lynx2-8-2-patched/src/UCdomap.c   Tue Mar 30 23:00:07 1999
> @@ -32,6 +32,9 @@
>  /*
>   *  Include chartrans tables:
>   */
> +#include <iso01_uni.h>               /* ISO Latin 1          */
> +#include <def7_uni.h>                /* 7 bit approximations */
> +#ifndef DISABLE_MOST_CHARSETS
>  #include <cp1250_uni.h>      /* WinLatin2 (cp1250)   */
>  #include <cp1251_uni.h>      /* WinCyrillic (cp1251) */
>  #include <cp1252_uni.h>      /* WinLatin1 (cp1252)   */
> @@ -48,10 +51,8 @@
>  #include <cp864_uni.h>               /* DosArabic (cp864)    */
>  #include <cp866_uni.h>               /* DosCyrillic (cp866)  */
>  #include <cp869_uni.h>               /* DosGreek2 (cp869)    */
> -#include <def7_uni.h>                /* 7 bit approximations */
>  #include <dmcs_uni.h>                /* DEC Multinational    */
>  #include <hp_uni.h>          /* HP Roman8            */
> -#include <iso01_uni.h>               /* ISO Latin 1          */
>  #include <iso02_uni.h>               /* ISO Latin 2          */
>  #include <iso03_uni.h>               /* ISO Latin 3          */
>  #include <iso04_uni.h>               /* ISO Latin 4          */
> @@ -74,6 +75,7 @@
>  #ifdef NOTDEFINED
>  #include <mnem_suni.h>
>  #endif /* NOTDEFINED */
> +#endif /* ifndef DISABLE_MOST_CHARSETS */

>  /*
>   *  Some of the code below, and some of the comments, are left in for
> @@ -2037,6 +2039,7 @@ PUBLIC void UCInit NOARGS
>   */

>      UC_CHARSET_SETUP_iso_8859_1;       /* ISO Latin 1          */
> +#ifndef DISABLE_MOST_CHARSETS
>      UC_CHARSET_SETUP_iso_8859_15;      /* ISO 8859-15 (Latin 9)*/
>      UC_CHARSET_SETUP_cp850;            /* DosLatin1 (cp850)    */
>      UC_CHARSET_SETUP_windows_1252;     /* WinLatin1 (cp1252)   */
> @@ -2054,7 +2057,10 @@ PUBLIC void UCInit NOARGS
>      UC_CHARSET_SETUP_big5;             /*** Taipei (Big5)        */

>      UC_CHARSET_SETUP_viscii;           /* Vietnamese (VISCII)  */
> +#endif /* ifndef DISABLE_MOST_CHARSETS */
> +    /* This (endif, then ifdef) looks weird.  Think before reacting. */
>      UC_CHARSET_SETUP;        /* us-ascii */    /* 7 bit approximations */
> +#ifndef DISABLE_MOST_CHARSETS

>      UC_CHARSET_SETUP_x_transparent;    /*** Transparent        */

> @@ -2090,6 +2096,7 @@ PUBLIC void UCInit NOARGS
>  #ifdef NOTDEFINED
>      UC_CHARSET_SETUP_mnem;
>  #endif /* NOTDEFINED */
> +#endif /* ifndef DISABLE_MOST_CHARSETS */

>  /*
>   *  To add synonyms for any charset name
> @@ -2099,7 +2106,10 @@ PUBLIC void UCInit NOARGS
>  /* easy to type: */
>      LATIN1   = UCGetLYhndl_byMIME("iso-8859-1");
>      US_ASCII = UCGetLYhndl_byMIME("us-ascii");
> +#ifndef DISABLE_MOST_CHARSETS
>      UTF8     = UCGetLYhndl_byMIME("utf-8");
> +#endif /* ifndef DISABLE_MOST_CHARSETS */
> +
>  }

>  /*
> diff -Burp lynx2-8-2/src/UCdomap.h lynx2-8-2-patched/src/UCdomap.h
> --- lynx2-8-2/src/UCdomap.h   Tue Mar 30 12:10:37 1999
> +++ lynx2-8-2-patched/src/UCdomap.h   Tue Mar 30 23:08:11 1999
> @@ -70,6 +70,8 @@ extern void UCInit NOARGS;
>   *  see also UC_Charset_Setup() above for details.
>   */

> +#ifndef DISABLE_MOST_CHARSETS
> +
>    /*
>     *  There is no strict correlation for the next five, since the transfer
>     *  charset gets decoded into Display Char Set by the CJK code (separate
> @@ -106,5 +108,6 @@ static CONST struct unimapdesc_str dfont
>         NULL,NULL,0,dfont_replacedesc_NO_fallback,\
>         128,UCT_ENC_UTF8,0)

> +#endif /* ifndef DISABLE_MOST_CHARSETS */

>  #endif /* UCDOMAP_H */