diff options
Diffstat (limited to 'ldap/c-sdk/libraries/liblber/decode.c')
-rw-r--r-- | ldap/c-sdk/libraries/liblber/decode.c | 833 |
1 files changed, 833 insertions, 0 deletions
diff --git a/ldap/c-sdk/libraries/liblber/decode.c b/ldap/c-sdk/libraries/liblber/decode.c new file mode 100644 index 000000000..732dd6e64 --- /dev/null +++ b/ldap/c-sdk/libraries/liblber/decode.c @@ -0,0 +1,833 @@ +/* ***** BEGIN LICENSE BLOCK ***** + * Version: MPL 1.1/GPL 2.0/LGPL 2.1 + * + * The contents of this file are subject to the Mozilla Public License Version + * 1.1 (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * http://www.mozilla.org/MPL/ + * + * Software distributed under the License is distributed on an "AS IS" basis, + * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License + * for the specific language governing rights and limitations under the + * License. + * + * The Original Code is Mozilla Communicator client code, released + * March 31, 1998. + * + * The Initial Developer of the Original Code is + * Netscape Communications Corporation. + * Portions created by the Initial Developer are Copyright (C) 1998-1999 + * the Initial Developer. All Rights Reserved. + * + * Contributor(s): + * + * Alternatively, the contents of this file may be used under the terms of + * either of the GNU General Public License Version 2 or later (the "GPL"), + * or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"), + * in which case the provisions of the GPL or the LGPL are applicable instead + * of those above. If you wish to allow use of your version of this file only + * under the terms of either the GPL or the LGPL, and not to allow others to + * use your version of this file under the terms of the MPL, indicate your + * decision by deleting the provisions above and replace them with the notice + * and other provisions required by the GPL or the LGPL. If you do not delete + * the provisions above, a recipient may use your version of this file under + * the terms of any one of the MPL, the GPL or the LGPL. + * + * ***** END LICENSE BLOCK ***** */ + +/* + * Copyright (c) 1990 Regents of the University of Michigan. + * All rights reserved. + * + * Redistribution and use in source and binary forms are permitted + * provided that this notice is preserved and that due credit is given + * to the University of Michigan at Ann Arbor. The name of the University + * may not be used to endorse or promote products derived from this + * software without specific prior written permission. This software + * is provided ``as is'' without express or implied warranty. + */ + +/* decode.c - ber input decoding routines */ + +#include "lber-int.h" + +/* + * Note: ber_get_tag() only uses the ber_end and ber_ptr elements of ber. + * If that changes, the ber_peek_tag() and/or ber_skip_tag() implementations + * will need to be changed. + */ +/* return the tag - LBER_DEFAULT returned means trouble */ +ber_tag_t +LDAP_CALL +ber_get_tag( BerElement *ber ) +{ + unsigned char xbyte; + ber_tag_t tag; + char *tagp; + int i; + + if ( ber_read( ber, (char *) &xbyte, 1 ) != 1 ) + return( LBER_DEFAULT ); + + if ( (xbyte & LBER_BIG_TAG_MASK) != LBER_BIG_TAG_MASK ) + return( (ber_uint_t) xbyte ); + + tagp = (char *) &tag; + tagp[0] = xbyte; + for ( i = 1; i < sizeof(ber_int_t); i++ ) { + if ( ber_read( ber, (char *) &xbyte, 1 ) != 1 ) + return( LBER_DEFAULT ); + + tagp[i] = xbyte; + + if ( ! (xbyte & LBER_MORE_TAG_MASK) ) + break; + } + + /* tag too big! */ + if ( i == sizeof(ber_int_t) ) + return( LBER_DEFAULT ); + + /* want leading, not trailing 0's */ + return( tag >> (sizeof(ber_int_t) - i - 1) ); +} + +/* + * Note: ber_skip_tag() only uses the ber_end and ber_ptr elements of ber. + * If that changes, the implementation of ber_peek_tag() will need to + * be changed. + */ +ber_tag_t +LDAP_CALL +ber_skip_tag( BerElement *ber, ber_len_t *len ) +{ + ber_tag_t tag; + unsigned char lc; + int noctets, diff; + ber_len_t netlen; + + /* + * Any ber element looks like this: tag length contents. + * Assuming everything's ok, we return the tag byte (we + * can assume a single byte), and return the length in len. + * + * Assumptions: + * 1) definite lengths + * 2) primitive encodings used whenever possible + */ + + /* + * First, we read the tag. + */ + + if ( (tag = ber_get_tag( ber )) == LBER_DEFAULT ) + return( LBER_DEFAULT ); + + /* + * Next, read the length. The first byte contains the length of + * the length. If bit 8 is set, the length is the long form, + * otherwise it's the short form. We don't allow a length that's + * greater than what we can hold in an unsigned long. + */ + + *len = netlen = 0; + if ( ber_read( ber, (char *) &lc, 1 ) != 1 ) + return( LBER_DEFAULT ); + if ( lc & 0x80 ) { + noctets = (lc & 0x7f); + if ( noctets > sizeof(ber_uint_t) ) + return( LBER_DEFAULT ); + diff = sizeof(ber_int_t) - noctets; + if ( ber_read( ber, (char *) &netlen + diff, noctets ) + != noctets ) + return( LBER_DEFAULT ); + *len = LBER_NTOHL( netlen ); + } else { + *len = lc; + } + + return( tag ); +} + + +/* + * Note: Previously, we passed the "ber" parameter directly to ber_skip_tag(), + * saving and restoring the ber_ptr element only. We now take advantage + * of the fact that the only ber structure elements touched by ber_skip_tag() + * are ber_end and ber_ptr. If that changes, this code must change too. + */ +ber_tag_t +LDAP_CALL +ber_peek_tag( BerElement *ber, ber_len_t *len ) +{ + BerElement bercopy; + + bercopy.ber_end = ber->ber_end; + bercopy.ber_ptr = ber->ber_ptr; + return( ber_skip_tag( &bercopy, len )); +} + +static int +ber_getnint( BerElement *ber, ber_int_t *num, ber_slen_t len ) +{ + int i; + ber_int_t value; + unsigned char buffer[sizeof(ber_int_t)]; + /* + * The tag and length have already been stripped off. We should + * be sitting right before len bytes of 2's complement integer, + * ready to be read straight into an int. We may have to sign + * extend after we read it in. + */ + + if ( len > sizeof(ber_slen_t) ) + return( -1 ); + + /* read into the low-order bytes of netnum */ + if ( ber_read( ber, (char *) buffer, len ) != len ) + return( -1 ); + + /* This sets the required sign extension */ + if ( len != 0) { + value = 0x80 & buffer[0] ? (-1) : 0; + } else { + value = 0; + } + + for ( i = 0; i < len; i++ ) + value = (value << 8) | buffer[i]; + + *num = value; + + return( len ); +} + +ber_tag_t +LDAP_CALL +ber_get_int( BerElement *ber, ber_int_t *num ) +{ + ber_tag_t tag; + ber_len_t len; + + if ( (tag = ber_skip_tag( ber, &len )) == LBER_DEFAULT ) + return( LBER_DEFAULT ); + + /* + * len is being demoted to a long here -- possible conversion error + */ + + if ( ber_getnint( ber, num, (int)len ) != (ber_slen_t)len ) + return( LBER_DEFAULT ); + else + return( tag ); +} + +ber_tag_t +LDAP_CALL +ber_get_stringb( BerElement *ber, char *buf, ber_len_t *len ) +{ + ber_len_t datalen; + ber_tag_t tag; +#ifdef STR_TRANSLATION + char *transbuf; +#endif /* STR_TRANSLATION */ + + if ( (tag = ber_skip_tag( ber, &datalen )) == LBER_DEFAULT ) + return( LBER_DEFAULT ); + if ( datalen > (*len - 1) ) + return( LBER_DEFAULT ); + + /* + * datalen is being demoted to a long here -- possible conversion error + */ + + if ( ber_read( ber, buf, datalen ) != (ber_slen_t) datalen ) + return( LBER_DEFAULT ); + + buf[datalen] = '\0'; + +#ifdef STR_TRANSLATION + if ( datalen > 0 && ( ber->ber_options & LBER_OPT_TRANSLATE_STRINGS ) + != 0 && ber->ber_decode_translate_proc != NULL ) { + transbuf = buf; + ++datalen; + if ( (*(ber->ber_decode_translate_proc))( &transbuf, &datalen, + 0 ) != 0 ) { + return( LBER_DEFAULT ); + } + if ( datalen > *len ) { + NSLBERI_FREE( transbuf ); + return( LBER_DEFAULT ); + } + SAFEMEMCPY( buf, transbuf, datalen ); + NSLBERI_FREE( transbuf ); + --datalen; + } +#endif /* STR_TRANSLATION */ + + *len = datalen; + return( tag ); +} + +ber_tag_t +LDAP_CALL +ber_get_stringa( BerElement *ber, char **buf ) +{ + ber_len_t datalen, ndatalen; + ber_tag_t tag; + + if ( (tag = ber_skip_tag( ber, &datalen )) == LBER_DEFAULT ) + return( LBER_DEFAULT ); + + if ( ((ndatalen = (size_t)datalen + 1) < (size_t) datalen) || + ( datalen > (ber->ber_end - ber->ber_ptr) ) || + ( (*buf = (char *)NSLBERI_MALLOC( (size_t)ndatalen )) == NULL )) + return( LBER_DEFAULT ); + + /* + * datalen is being demoted to a long here -- possible conversion error + */ + if ( ber_read( ber, *buf, datalen ) != (ber_slen_t) datalen ) { + NSLBERI_FREE( *buf ); + *buf = NULL; + return( LBER_DEFAULT ); + } + + (*buf)[datalen] = '\0'; + +#ifdef STR_TRANSLATION + if ( datalen > 0 && ( ber->ber_options & LBER_OPT_TRANSLATE_STRINGS ) + != 0 && ber->ber_decode_translate_proc != NULL ) { + ++datalen; + if ( (*(ber->ber_decode_translate_proc))( buf, &datalen, 1 ) + != 0 ) { + NSLBERI_FREE( *buf ); + *buf = NULL; + return( LBER_DEFAULT ); + } + } +#endif /* STR_TRANSLATION */ + + return( tag ); +} + +ber_tag_t +LDAP_CALL +ber_get_stringal( BerElement *ber, struct berval **bv ) +{ + ber_len_t len, nlen; + ber_tag_t tag; + + if ( (*bv = (struct berval *)NSLBERI_MALLOC( sizeof(struct berval) )) + == NULL ) { + return( LBER_DEFAULT ); + } + + (*bv)->bv_val = NULL; + (*bv)->bv_len = 0; + + if ( (tag = ber_skip_tag( ber, &len )) == LBER_DEFAULT ) { + NSLBERI_FREE( *bv ); + *bv = NULL; + return( LBER_DEFAULT ); + } + + if ( ((nlen = (size_t) len + 1) < (size_t)len) || + ( len > (ber->ber_end - ber->ber_ptr) ) || + (((*bv)->bv_val = (char *)NSLBERI_MALLOC( (size_t)nlen )) + == NULL )) { + NSLBERI_FREE( *bv ); + *bv = NULL; + return( LBER_DEFAULT ); + } + + /* + * len is being demoted to a long here -- possible conversion error + */ + if ( ber_read( ber, (*bv)->bv_val, len ) != (ber_slen_t) len ) { + NSLBERI_FREE( (*bv)->bv_val ); + (*bv)->bv_val = NULL; + NSLBERI_FREE( *bv ); + *bv = NULL; + return( LBER_DEFAULT ); + } + + ((*bv)->bv_val)[len] = '\0'; + (*bv)->bv_len = len; + +#ifdef STR_TRANSLATION + if ( len > 0 && ( ber->ber_options & LBER_OPT_TRANSLATE_STRINGS ) != 0 + && ber->ber_decode_translate_proc != NULL ) { + ++len; + if ( (*(ber->ber_decode_translate_proc))( &((*bv)->bv_val), + &len, 1 ) != 0 ) { + NSLBERI_FREE( (*bv)->bv_val ); + (*bv)->bv_val = NULL; + NSLBERI_FREE( *bv ); + *bv = NULL; + return( LBER_DEFAULT ); + } + (*bv)->bv_len = len - 1; + } +#endif /* STR_TRANSLATION */ + + return( tag ); +} + +ber_tag_t +LDAP_CALL +ber_get_bitstringa( BerElement *ber, char **buf, ber_len_t *blen ) +{ + ber_len_t datalen; + ber_tag_t tag; + unsigned char unusedbits; + + if ( (tag = ber_skip_tag( ber, &datalen )) == LBER_DEFAULT ) + return( LBER_DEFAULT ); + --datalen; + + if ( (datalen > (ber->ber_end - ber->ber_ptr)) || + ( (*buf = (char *)NSLBERI_MALLOC((size_t)datalen )) == NULL ) ) + return( LBER_DEFAULT ); + + if ( ber_read( ber, (char *)&unusedbits, 1 ) != 1 ) { + NSLBERI_FREE( *buf ); + *buf = NULL; + return( LBER_DEFAULT ); + } + + /* + * datalen is being demoted to a long here -- possible conversion error + */ + if ( ber_read( ber, *buf, datalen ) != (ber_slen_t) datalen ) { + NSLBERI_FREE( *buf ); + *buf = NULL; + return( LBER_DEFAULT ); + } + + *blen = datalen * 8 - unusedbits; + return( tag ); +} + +ber_tag_t +LDAP_CALL +ber_get_null( BerElement *ber ) +{ + ber_len_t len; + ber_tag_t tag; + + if ( (tag = ber_skip_tag( ber, &len )) == LBER_DEFAULT ) + return( LBER_DEFAULT ); + + if ( len != 0 ) + return( LBER_DEFAULT ); + + return( tag ); +} + +ber_tag_t +LDAP_CALL +ber_get_boolean( BerElement *ber, ber_int_t *boolval ) +{ + int rc; + + rc = ber_get_int( ber, boolval ); + + return( rc ); +} + +ber_tag_t +LDAP_CALL +ber_first_element( BerElement *ber, ber_len_t *len, char **last ) +{ + /* skip the sequence header, use the len to mark where to stop */ + if ( ber_skip_tag( ber, len ) == LBER_DEFAULT ) { + return( LBER_ERROR ); + } + + *last = ber->ber_ptr + *len; + + if ( *last == ber->ber_ptr ) { + return( LBER_END_OF_SEQORSET ); + } + + return( ber_peek_tag( ber, len ) ); +} + +ber_tag_t +LDAP_CALL +ber_next_element( BerElement *ber, ber_len_t *len, char *last ) +{ + if ( ber->ber_ptr == last ) { + return( LBER_END_OF_SEQORSET ); + } + + return( ber_peek_tag( ber, len ) ); +} + +/* VARARGS */ +ber_tag_t +LDAP_C +ber_scanf( BerElement *ber, const char *fmt, ... ) +{ + va_list ap; + char *last, *p; + char *s, **ss, ***sss; + struct berval ***bv, **bvp, *bval; + int *i, j; + ber_int_t *l, rc, tag; + ber_tag_t *t; + ber_len_t len; + size_t array_size; + + va_start( ap, fmt ); + +#ifdef LDAP_DEBUG + if ( lber_debug & 64 ) { + char msg[80]; + sprintf( msg, "ber_scanf fmt (%s) ber:\n", fmt ); + ber_err_print( msg ); + ber_dump( ber, 1 ); + } +#endif + for ( rc = 0, p = (char *) fmt; *p && rc != LBER_DEFAULT; p++ ) { + switch ( *p ) { + case 'a': /* octet string - allocate storage as needed */ + ss = va_arg( ap, char ** ); + rc = ber_get_stringa( ber, ss ); + break; + + case 'b': /* boolean */ + i = va_arg( ap, int * ); + rc = ber_get_boolean( ber, i ); + break; + + case 'e': /* enumerated */ + case 'i': /* int */ + l = va_arg( ap, ber_slen_t * ); + rc = ber_get_int( ber, l ); + break; + + case 'l': /* length of next item */ + l = va_arg( ap, ber_slen_t * ); + rc = ber_peek_tag( ber, (ber_len_t *)l ); + break; + + case 'n': /* null */ + rc = ber_get_null( ber ); + break; + + case 's': /* octet string - in a buffer */ + s = va_arg( ap, char * ); + l = va_arg( ap, ber_slen_t * ); + rc = ber_get_stringb( ber, s, (ber_len_t *)l ); + break; + + case 'o': /* octet string in a supplied berval */ + bval = va_arg( ap, struct berval * ); + ber_peek_tag( ber, &bval->bv_len ); + rc = ber_get_stringa( ber, &bval->bv_val ); + break; + + case 'O': /* octet string - allocate & include length */ + bvp = va_arg( ap, struct berval ** ); + rc = ber_get_stringal( ber, bvp ); + break; + + case 'B': /* bit string - allocate storage as needed */ + ss = va_arg( ap, char ** ); + l = va_arg( ap, ber_slen_t * ); /* for length, in bits */ + rc = ber_get_bitstringa( ber, ss, (ber_len_t *)l ); + break; + + case 't': /* tag of next item */ + t = va_arg( ap, ber_tag_t * ); + *t = rc = ber_peek_tag( ber, &len ); + break; + + case 'T': /* skip tag of next item */ + t = va_arg( ap, ber_tag_t * ); + *t = rc = ber_skip_tag( ber, &len ); + break; + + case 'v': /* sequence of strings */ + sss = va_arg( ap, char *** ); + *sss = NULL; + j = 0; + array_size = 0; + for ( tag = ber_first_element( ber, &len, &last ); + tag != LBER_DEFAULT && tag != LBER_END_OF_SEQORSET + && rc != LBER_DEFAULT; + tag = ber_next_element( ber, &len, last ) ) { + if ( *sss == NULL ) { + /* Make room for at least 15 strings */ + *sss = (char **)NSLBERI_MALLOC(16 * sizeof(char *) ); + if (!*sss) { + rc = LBER_DEFAULT; + break; /* out of memory - cannot continue */ + } + array_size = 16; + } else { + char **save_sss = *sss; + if ( (size_t)(j+2) > array_size) { + /* We'v overflowed our buffer */ + *sss = (char **)NSLBERI_REALLOC( *sss, (array_size * 2) * sizeof(char *) ); + array_size = array_size * 2; + } + if (!*sss) { + rc = LBER_DEFAULT; + ber_svecfree(save_sss); + break; /* out of memory - cannot continue */ + } + } + (*sss)[j] = NULL; + rc = ber_get_stringa( ber, &((*sss)[j]) ); + j++; + } + if ( rc != LBER_DEFAULT && + tag != LBER_END_OF_SEQORSET ) { + rc = LBER_DEFAULT; + } + if ( *sss && (j > 0) ) { + (*sss)[j] = NULL; + } + break; + + case 'V': /* sequence of strings + lengths */ + bv = va_arg( ap, struct berval *** ); + *bv = NULL; + j = 0; + for ( tag = ber_first_element( ber, &len, &last ); + tag != LBER_DEFAULT && tag != LBER_END_OF_SEQORSET + && rc != LBER_DEFAULT; + tag = ber_next_element( ber, &len, last ) ) { + if ( *bv == NULL ) { + *bv = (struct berval **)NSLBERI_MALLOC( + 2 * sizeof(struct berval *) ); + if (!*bv) { + rc = LBER_DEFAULT; + break; /* out of memory - cannot continue */ + } + } else { + struct berval **save_bv = *bv; + *bv = (struct berval **)NSLBERI_REALLOC( + *bv, + (j + 2) * sizeof(struct berval *) ); + if (!*bv) { + rc = LBER_DEFAULT; + ber_bvecfree(save_bv); + break; /* out of memory - cannot continue */ + } + } + rc = ber_get_stringal( ber, &((*bv)[j]) ); + j++; + } + if ( rc != LBER_DEFAULT && + tag != LBER_END_OF_SEQORSET ) { + rc = LBER_DEFAULT; + } + if ( *bv && (j > 0) ) { + (*bv)[j] = NULL; + } + break; + + case 'x': /* skip the next element - whatever it is */ + if ( (rc = ber_skip_tag( ber, &len )) == LBER_DEFAULT ) + break; + ber->ber_ptr += len; + break; + + case '{': /* begin sequence */ + case '[': /* begin set */ + if ( *(p + 1) != 'v' && *(p + 1) != 'V' ) + rc = ber_skip_tag( ber, &len ); + break; + + case '}': /* end sequence */ + case ']': /* end set */ + break; + + default: + { + char msg[80]; + sprintf( msg, "unknown fmt %c\n", *p ); + ber_err_print( msg ); + } + rc = LBER_DEFAULT; + break; + } + } + + va_end( ap ); + + if (rc == LBER_DEFAULT) { + va_start( ap, fmt ); + for ( p--; fmt < p && *fmt; fmt++ ) { + switch ( *fmt ) { + case 'a': /* octet string - allocate storage as needed */ + ss = va_arg( ap, char ** ); + NSLBERI_FREE(*ss); + *ss = NULL; + break; + + case 'b': /* boolean */ + i = va_arg( ap, int * ); + break; + + case 'e': /* enumerated */ + case 'i': /* int */ + l = va_arg( ap, ber_slen_t * ); + break; + + case 'l': /* length of next item */ + l = va_arg( ap, ber_slen_t * ); + break; + + case 'n': /* null */ + break; + + case 's': /* octet string - in a buffer */ + s = va_arg( ap, char * ); + l = va_arg( ap, ber_slen_t * ); + break; + + case 'o': /* octet string in a supplied berval */ + bval = va_arg( ap, struct berval * ); + if (bval->bv_val) NSLBERI_FREE(bval->bv_val); + memset(bval, 0, sizeof(struct berval)); + break; + + case 'O': /* octet string - allocate & include length */ + bvp = va_arg( ap, struct berval ** ); + ber_bvfree(*bvp); + bvp = NULL; + break; + + case 'B': /* bit string - allocate storage as needed */ + ss = va_arg( ap, char ** ); + l = va_arg( ap, ber_slen_t * ); /* for length, in bits */ + if (*ss) NSLBERI_FREE(*ss); + *ss = NULL; + break; + + case 't': /* tag of next item */ + t = va_arg( ap, ber_tag_t * ); + break; + + case 'T': /* skip tag of next item */ + t = va_arg( ap, ber_tag_t * ); + break; + + case 'v': /* sequence of strings */ + sss = va_arg( ap, char *** ); + ber_svecfree(*sss); + *sss = NULL; + break; + + case 'V': /* sequence of strings + lengths */ + bv = va_arg( ap, struct berval *** ); + ber_bvecfree(*bv); + *bv = NULL; + break; + + case 'x': /* skip the next element - whatever it is */ + break; + + case '{': /* begin sequence */ + case '[': /* begin set */ + break; + + case '}': /* end sequence */ + case ']': /* end set */ + break; + + default: + break; + } + } /* for */ + va_end( ap ); + } /* if */ + + return( rc ); +} + +void +LDAP_CALL +ber_bvfree( struct berval *bv ) +{ + if ( bv != NULL ) { + if ( bv->bv_val != NULL ) { + NSLBERI_FREE( bv->bv_val ); + } + NSLBERI_FREE( (char *) bv ); + } +} + +void +LDAP_CALL +ber_bvecfree( struct berval **bv ) +{ + int i; + + if ( bv != NULL ) { + for ( i = 0; bv[i] != NULL; i++ ) { + ber_bvfree( bv[i] ); + } + NSLBERI_FREE( (char *) bv ); + } +} + +struct berval * +LDAP_CALL +ber_bvdup( const struct berval *bv ) +{ + struct berval *new; + + if ( (new = (struct berval *)NSLBERI_MALLOC( sizeof(struct berval) )) + == NULL ) { + return( NULL ); + } + if ( bv->bv_val == NULL ) { + new->bv_val = NULL; + new->bv_len = 0; + } else { + if ( (new->bv_val = (char *)NSLBERI_MALLOC( bv->bv_len + 1 )) + == NULL ) { + NSLBERI_FREE( new ); + new = NULL; + return( NULL ); + } + SAFEMEMCPY( new->bv_val, bv->bv_val, (size_t) bv->bv_len ); + new->bv_val[bv->bv_len] = '\0'; + new->bv_len = bv->bv_len; + } + + return( new ); +} + +void +LDAP_CALL +ber_svecfree( char **vals ) +{ + int i; + + if ( vals == NULL ) + return; + for ( i = 0; vals[i] != NULL; i++ ) + NSLBERI_FREE( vals[i] ); + NSLBERI_FREE( (char *) vals ); +} + +#ifdef STR_TRANSLATION +void +LDAP_CALL +ber_set_string_translators( + BerElement *ber, + BERTranslateProc encode_proc, + BERTranslateProc decode_proc +) +{ + ber->ber_encode_translate_proc = encode_proc; + ber->ber_decode_translate_proc = decode_proc; +} +#endif /* STR_TRANSLATION */ |