Re: [GENERAL] invalid multibyte character for locale - Mailing list pgsql-patches

Tatsuo Ishii wrote:
> Apparently your hack does not kill #define USE_WIDE_UPPER_LOWER.
>
> BTW, the current code for upper/lower etc. seems to be broken. The
> exact problem you have are happening in Japanese encodings too(EUC_JP)
> too. PostgreSQL should not use wide-character method if LC_CTYPE is C.

Here is the patch I made. I will apply it to CVS HEAD and 8.0.X because
it is a bug fix.

---------------------------------------------------------------------------

> --
> Tatsuo Ishii
>
> > L.S.
> >
> > I have a database created on:
> >
> > db=# select version();
> >                                version
> > ---------------------------------------------------------------------
> >  PostgreSQL 8.0.1 on i686-pc-linux-gnu, compiled by GCC egcs-2.91.66
> > (1 row)
> >
> >
> > The initdb was done using no-locale and unicode as default encoding, the
> > particular database itself is indeed encoded as UNICODE.
> >
> >
> > Due to a buggy glibc, the following patch was applied to this install in order
> > to avoid a crash on things like 'upper(<string>)':
> >
> > --- oracle_compat.c_orig        Mon Dec  6 22:14:11 2004
> > +++ oracle_compat.c     Mon Dec  6 22:14:24 2004
> > @@ -43,7 +43,7 @@
> >   * We assume if we have these two functions, we have their friends too, and
> >   * can use the wide-character method.
> >   */
> > -#if defined(HAVE_WCSTOMBS) && defined(HAVE_TOWLOWER)
> > +#if defined(HAVE_WCSTOMBS) && defined(HAVE_TOWLOWER) && FALSE
> >  #define USE_WIDE_UPPER_LOWER
> >  #endif
> >
> >
> > The database on this machine was dumped and then restored on another, which
> > has a more recent installation of Slack on it:
> >
> >
> > db=# select version();
> >                                 version
> > ------------------------------------------------------------------------
> >  PostgreSQL 8.0.1 on i586-pc-linux-gnu, compiled by GCC gcc (GCC) 3.2.3
> > (1 row)
> >
> >
> > Again, the initdb on this machine was done using no-locale and unicode as
> > default encoding, the particular database obviously is also encoded as
> > UNICODE.
> >
> >
> >
> > On the second machine, I'm now getting the following:
> >
> > db=# select 'J?TERBOG';
> >  ?column?
> > ----------
> >  J?TERBOG
> > (1 row)
> >
> > db=# select lower('J?TERBOG');
> > ERROR:  invalid multibyte character for locale
> > HINT:  The server's LC_CTYPE locale is probably incompatible with the database
> > encoding.
> >
> >
> >
> > As far as I can tell, this didn't happen with v8.0.0, but I'm afraid I can't
> > be totally sure about that. Obviously, the error doesn't occur on the first
> > machine due to the hack needed for the buggy glibc.
> >
> >
> > I'd appreciate a pointer as to what is causing this. It 'shouldn't' be the
> > hack nor the dump/restore cycle, but.......?
> >
> >
> > TIA.
> >
> >
> >
> > --
> > Best,
> >
> >
> >
> >
> > Frank.
> >
> > ---------------------------(end of broadcast)---------------------------
> > TIP 7: don't forget to increase your free space map settings
> >
>
> ---------------------------(end of broadcast)---------------------------
> TIP 2: you can get off all lists at once with the unregister command
>     (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)
>

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073
Index: src/backend/utils/adt/oracle_compat.c
===================================================================
RCS file: /cvsroot/pgsql/src/backend/utils/adt/oracle_compat.c,v
retrieving revision 1.57
diff -c -c -r1.57 oracle_compat.c
*** src/backend/utils/adt/oracle_compat.c    31 Dec 2004 22:01:22 -0000    1.57
--- src/backend/utils/adt/oracle_compat.c    15 Mar 2005 13:45:22 -0000
***************
*** 166,173 ****
  lower(PG_FUNCTION_ARGS)
  {
  #ifdef USE_WIDE_UPPER_LOWER
!     /* use wide char code only when max encoding length > one */
!     if (pg_database_encoding_max_length() > 1)
      {
          text       *string = PG_GETARG_TEXT_P(0);
          text       *result;
--- 166,173 ----
  lower(PG_FUNCTION_ARGS)
  {
  #ifdef USE_WIDE_UPPER_LOWER
!     /* use wide char code only when max encoding length > 1 and ctype != C */
!     if (pg_database_encoding_max_length() > 1 && !lc_ctype_is_c())
      {
          text       *string = PG_GETARG_TEXT_P(0);
          text       *result;
***************
*** 228,235 ****
  upper(PG_FUNCTION_ARGS)
  {
  #ifdef USE_WIDE_UPPER_LOWER
!     /* use wide char code only when max encoding length > one */
!     if (pg_database_encoding_max_length() > 1)
      {
          text       *string = PG_GETARG_TEXT_P(0);
          text       *result;
--- 228,235 ----
  upper(PG_FUNCTION_ARGS)
  {
  #ifdef USE_WIDE_UPPER_LOWER
!     /* use wide char code only when max encoding length > 1 and ctype != C */
!     if (pg_database_encoding_max_length() > 1 && !lc_ctype_is_c())
      {
          text       *string = PG_GETARG_TEXT_P(0);
          text       *result;
***************
*** 293,300 ****
  initcap(PG_FUNCTION_ARGS)
  {
  #ifdef USE_WIDE_UPPER_LOWER
!     /* use wide char code only when max encoding length > one */
!     if (pg_database_encoding_max_length() > 1)
      {
          text       *string = PG_GETARG_TEXT_P(0);
          text       *result;
--- 293,300 ----
  initcap(PG_FUNCTION_ARGS)
  {
  #ifdef USE_WIDE_UPPER_LOWER
!     /* use wide char code only when max encoding length > 1 and ctype != C */
!     if (pg_database_encoding_max_length() > 1 && !lc_ctype_is_c())
      {
          text       *string = PG_GETARG_TEXT_P(0);
          text       *result;
Index: src/backend/utils/adt/pg_locale.c
===================================================================
RCS file: /cvsroot/pgsql/src/backend/utils/adt/pg_locale.c,v
retrieving revision 1.30
diff -c -c -r1.30 pg_locale.c
*** src/backend/utils/adt/pg_locale.c    1 Jan 2005 05:43:07 -0000    1.30
--- src/backend/utils/adt/pg_locale.c    15 Mar 2005 13:45:22 -0000
***************
*** 197,202 ****
--- 197,229 ----


  /*
+  * We'd like to cache whether LC_CTYPE is C (or POSIX), so we can
+  * optimize a few code paths in various places.
+  */
+ bool
+ lc_ctype_is_c(void)
+ {
+     /* Cache result so we only have to compute it once */
+     static int    result = -1;
+     char       *localeptr;
+
+     if (result >= 0)
+         return (bool) result;
+     localeptr = setlocale(LC_CTYPE, NULL);
+     if (!localeptr)
+         elog(ERROR, "invalid LC_CTYPE setting");
+
+     if (strcmp(localeptr, "C") == 0)
+         result = true;
+     else if (strcmp(localeptr, "POSIX") == 0)
+         result = true;
+     else
+         result = false;
+     return (bool) result;
+ }
+
+
+ /*
   * Frees the malloced content of a struct lconv.  (But not the struct
   * itself.)
   */
Index: src/include/utils/pg_locale.h
===================================================================
RCS file: /cvsroot/pgsql/src/include/utils/pg_locale.h,v
retrieving revision 1.19
diff -c -c -r1.19 pg_locale.h
*** src/include/utils/pg_locale.h    1 Jan 2005 05:43:09 -0000    1.19
--- src/include/utils/pg_locale.h    15 Mar 2005 13:45:23 -0000
***************
*** 32,37 ****
--- 32,38 ----
                     bool doit, GucSource source);

  extern bool lc_collate_is_c(void);
+ extern bool lc_ctype_is_c(void);

  /*
   * Return the POSIX lconv struct (contains number/money formatting