Memory leaks in outer

diff --git a/THANKS b/THANKS
index 39099d0a29b2ea5db716b4e419a3d2f64311d059..deddce4479856883ade2170e8ea409674f332446 100644 (file)
--- a/THANKS
+++ b/THANKS
@@ -88,6 +88,7 @@ Frère Roy            roy@taize.fr
                        http://www.taize.fr
 Gabriel P. Silva       gpsilva@geocities.com
                        http://www.nce.ufrj.br/~gabriel
+Gaël Le Mignot                kilobug@freesurf.fr
 Ghislain Plamondon
 Georg Haefele          haefele@atlas.gis.univie.ac.at
 Greg McGary

diff --git a/doc/recode.info b/doc/recode.info
index fd9b2f578519cfacd46ed45ccc0a304f391add5c..cc494143040f21e40e8ff44802a14cfbd78dc5f6 100644 (file)
--- a/doc/recode.info
+++ b/doc/recode.info
@@ -1487,6 +1487,19 @@ his program (let's assume the programmer is a male, here, no prejudice
 intended).  This `outer' variable is given as a first argument to all
 outer level functions.
 
+   The `RECODE_OUTER' structure is really meant to be initialised only
+once in the life of a program, and terminated with the program itself.
+Program interfaces should pay attention to initialise it only once,
+would it be only for speed considerations.  A good deal of overhead goes
+to outer level initialization, and if the outer level was initialized
+afresh for each and every string translated, say, the Recode library
+would appear immensely much slower that it was meant to be!
+
+   Because outer level initialization is meant to be done only once,
+not so much attention has been paid to avoid memory leaks at this level
+within Recode.  This is hardly a reason for not plugging such leaks at
+any level: in the long run, they should all be chased and repaired.
+
    The `<recode.h>' header file uses the Boolean type setup by the
 system header file `<stdbool.h>'.  But this header file is still fairly
 new in C standards, and likely does not exist everywhere.  If you
@@ -1618,10 +1631,11 @@ File: recode.info,  Node: Request level,  Next: Task level,  Prev: Outer level,
 
 The request level functions are meant to cover most recoding needs
 programmers may have; they should provide all usual functionality.
-Their API is almost stable by now.  To get started with request level
-functions, here is a full example of a program which sole job is to
-filter `ibmpc' code on its standard input into `latin1' code on its
-standard output.
+Their API is almost stable by now.
+
+   To get started with request level functions, here is a full example
+of a program which sole job is to filter `ibmpc' code on its standard
+input into `latin1' code on its standard output.
 
      #include <stdio.h>
      #include <stdbool.h>
@@ -1652,6 +1666,14 @@ which the programmer should use for allocating a variable in his
 program.  This REQUEST variable is given as a first argument to all
 request level functions, and in most cases, may be considered as opaque.
 
+   Suppose an application is doing a lot of recoding using only a few
+different requests.  For speed considerations, the `RECODE_REQUEST'
+structure should ideally be cached for each kind of request, so the
+request level initialisation is not redone for each and every string
+translated.  The speedup should be more apparent when Recode is able to
+optimize the work by building on the fly, within the structure, new
+specialized recoding steps and their associated data tables.
+
    * Initialisation functions 
 
           RECODE_REQUEST recode_new_request (OUTER);
@@ -5001,7 +5023,7 @@ Concept Index
 * ambiguous output, error message:       Errors.              (line  31)
 * ASCII table, recreating with Recode:   ASCII.               (line  12)
 * average number of recoding steps:      Main flow.           (line  40)
-* bool data type:                        Outer level.         (line  31)
+* bool data type:                        Outer level.         (line  44)
 * box-drawing characters:                Recoding.            (line  16)
 * bug reports, where to send:            Contributing.        (line  37)
 * byte order mark:                       UCS-2.               (line  12)
@@ -5080,8 +5102,8 @@ Concept Index
 * implied surfaces:                      Requests.            (line  69)
 * impossible conversions:                Charset overview.    (line  33)
 * information about charsets:            Listings.            (line 153)
-* initialisation functions, outer:       Outer level.         (line  84)
-* initialisation functions, request:     Request level.       (line  42)
+* initialisation functions, outer:       Outer level.         (line  97)
+* initialisation functions, request:     Request level.       (line  51)
 * initialisation functions, task:        Task level.          (line  52)
 * interface, with iconv library:         iconv.               (line   6)
 * intermediate charsets:                 Requests.            (line  23)
@@ -5096,6 +5118,7 @@ Concept Index
 * LaTeX files:                           LaTeX.               (line   6)
 * Latin charsets:                        ISO 8859.            (line   6)
 * Latin-1 table, recreating with Recode: ISO 8859.            (line  45)
+* leaks, memory:                         Outer level.         (line  39)
 * letter case, in charset and surface names: Requests.        (line  93)
 * libiconv:                              iconv.               (line   6)
 * library, iconv:                        iconv.               (line   6)
@@ -5104,6 +5127,7 @@ Concept Index
 * map filling:                           Reversibility.       (line  98)
 * map filling, disable:                  Reversibility.       (line  48)
 * markup language:                       HTML.                (line   6)
+* memory leaks:                          Outer level.         (line  39)
 * memory sequencing:                     Sequencing.          (line  23)
 * MIME encodings:                        MIME.                (line   6)
 * misuse of recoding library, error message: Errors.          (line  76)
@@ -5122,7 +5146,7 @@ Concept Index
 * partial conversion:                    Mixed.               (line  20)
 * permutations of groups of bytes:       Permutations.        (line   6)
 * pipe sequencing:                       Sequencing.          (line  40)
-* program_name variable:                 Outer level.         (line 137)
+* program_name variable:                 Outer level.         (line 150)
 * programming language support:          Listings.            (line  26)
 * pseudo-charsets:                       Charset overview.    (line  33)
 * pure charset:                          Surface overview.    (line  17)
@@ -5153,7 +5177,9 @@ Concept Index
 * silent operation:                      Reversibility.       (line  36)
 * single step:                           Main flow.           (line  17)
 * source file generation:                Listings.            (line  26)
-* stdbool.h header:                      Outer level.         (line  31)
+* speed considerations <1>:              Request level.       (line  43)
+* speed considerations:                  Outer level.         (line  31)
+* stdbool.h header:                      Outer level.         (line  44)
 * strict operation:                      Reversibility.       (line  48)
 * string and comments conversion:        Mixed.               (line  39)
 * structural surfaces:                   Surfaces.            (line  44)
@@ -5255,12 +5281,12 @@ and variables in the Recode library.
 * Menu:
 
 * abort_level:                           Task level.          (line 198)
-* ascii_graphics:                        Request level.       (line 101)
+* ascii_graphics:                        Request level.       (line 110)
 * byte_order_mark:                       Task level.          (line 182)
 * declare_step:                          New surfaces.        (line  13)
 * DEFAULT_CHARSET:                       Requests.            (line 104)
-* diacritics_only:                       Request level.       (line  92)
-* diaeresis_char:                        Request level.       (line  76)
+* diacritics_only:                       Request level.       (line 101)
+* diaeresis_char:                        Request level.       (line  85)
 * error_so_far:                          Task level.          (line 210)
 * fail_level:                            Task level.          (line 188)
 * file_one_to_many:                      New charsets.        (line  70)
@@ -5271,48 +5297,48 @@ and variables in the Recode library.
 * list_all_charsets:                     Charset level.       (line  15)
 * list_concise_charset:                  Charset level.       (line  15)
 * list_full_charset:                     Charset level.       (line  15)
-* make_header_flag:                      Request level.       (line  83)
+* make_header_flag:                      Request level.       (line  92)
 * RECODE_AMBIGUOUS_OUTPUT:               Errors.              (line  31)
-* recode_buffer_to_buffer:               Request level.       (line 147)
-* recode_buffer_to_file:                 Request level.       (line 147)
-* recode_delete_outer:                   Outer level.         (line  89)
-* recode_delete_request:                 Request level.       (line  47)
+* recode_buffer_to_buffer:               Request level.       (line 156)
+* recode_buffer_to_file:                 Request level.       (line 156)
+* recode_delete_outer:                   Outer level.         (line 102)
+* recode_delete_request:                 Request level.       (line  56)
 * recode_delete_task:                    Task level.          (line  54)
-* recode_file_to_buffer:                 Request level.       (line 147)
-* recode_file_to_file:                   Request level.       (line 147)
+* recode_file_to_buffer:                 Request level.       (line 156)
+* recode_file_to_file:                   Request level.       (line 156)
 * recode_filter_close:                   Task level.          (line 217)
-* recode_filter_close, not available:    Request level.       (line 212)
+* recode_filter_close, not available:    Request level.       (line 221)
 * recode_filter_open:                    Task level.          (line 217)
-* recode_filter_open, not available:     Request level.       (line 212)
-* recode_format_table:                   Request level.       (line 227)
+* recode_filter_open, not available:     Request level.       (line 221)
+* recode_format_table:                   Request level.       (line 236)
 * RECODE_INTERNAL_ERROR:                 Errors.              (line  81)
 * RECODE_INVALID_INPUT:                  Errors.              (line  61)
 * RECODE_MAXIMUM_ERROR <1>:              Errors.              (line  88)
 * RECODE_MAXIMUM_ERROR:                  Task level.          (line 198)
-* recode_new_outer:                      Outer level.         (line  89)
-* recode_new_request:                    Request level.       (line  47)
+* recode_new_outer:                      Outer level.         (line 102)
+* recode_new_request:                    Request level.       (line  56)
 * recode_new_task:                       Task level.          (line  54)
 * RECODE_NO_ERROR:                       Errors.              (line  16)
 * RECODE_NOT_CANONICAL:                  Errors.              (line  19)
 * RECODE_OUTER structure:                Outer level.         (line  25)
 * recode_perform_task:                   Task level.          (line 217)
-* recode_request structure:              Request level.       (line  59)
-* RECODE_REQUEST structure:              Request level.       (line  37)
-* recode_scan_request:                   Request level.       (line 111)
+* recode_request structure:              Request level.       (line  68)
+* RECODE_REQUEST structure:              Request level.       (line  38)
+* recode_scan_request:                   Request level.       (line 120)
 * RECODE_SEQUENCE_IN_MEMORY:             Task level.          (line 169)
 * RECODE_SEQUENCE_WITH_FILES:            Task level.          (line 172)
 * RECODE_SEQUENCE_WITH_PIPE:             Task level.          (line 175)
 * RECODE_STRATEGY_UNDECIDED:             Task level.          (line 162)
-* recode_string:                         Request level.       (line 140)
-* recode_string_to_buffer:               Request level.       (line 147)
-* recode_string_to_file:                 Request level.       (line 147)
+* recode_string:                         Request level.       (line 149)
+* recode_string_to_buffer:               Request level.       (line 156)
+* recode_string_to_file:                 Request level.       (line 156)
 * RECODE_SYSTEM_ERROR:                   Errors.              (line  71)
 * RECODE_TASK structure:                 Task level.          (line  46)
 * RECODE_UNTRANSLATABLE:                 Errors.              (line  50)
 * RECODE_USER_ERROR:                     Errors.              (line  76)
 * strategy:                              Task level.          (line 162)
 * task_request structure:                Task level.          (line  81)
-* verbose_flag:                          Request level.       (line  71)
+* verbose_flag:                          Request level.       (line  80)
 
 \1f
 File: recode.info,  Node: Charset and Surface Index,  Prev: Library Index,  Up: Top
@@ -6081,71 +6107,71 @@ Node: Emacs\7f58661
 Node: Debugging\7f59695
 Node: Library\7f63965
 Node: Outer level\7f65319
-Node: Request level\7f71429
-Node: Task level\7f81896
-Node: Charset level\7f92318
-Node: Errors\7f93160
-Ref: Errors-Footnote-1\7f98006
-Ref: Errors-Footnote-2\7f98120
-Node: Universal\7f98481
-Ref: Universal-Footnote-1\7f101593
-Ref: Universal-Footnote-2\7f101659
-Node: UCS-2\7f101872
-Node: UCS-4\7f104398
-Node: UTF-7\7f104938
-Node: UTF-8\7f105533
-Node: UTF-16\7f109838
-Node: count-characters\7f110986
-Node: dump-with-names\7f111657
-Node: iconv\7f114206
-Node: Tabular\7f117637
-Node: ASCII misc\7f139850
-Node: ASCII\7f140216
-Node: ISO 8859\7f141032
-Node: ASCII-BS\7f143326
-Node: flat\7f145163
-Node: IBM and MS\7f145834
-Node: EBCDIC\7f146378
-Node: IBM-PC\7f148474
-Ref: IBM-PC-Footnote-1\7f150588
-Node: Icon-QNX\7f150747
-Node: CDC\7f151172
-Node: Display Code\7f152853
-Ref: Display Code-Footnote-1\7f155134
-Node: CDC-NOS\7f155339
-Node: Bang-Bang\7f157301
-Node: Micros\7f159230
-Node: Apple-Mac\7f159613
-Node: AtariST\7f161647
-Node: Miscellaneous\7f162633
-Node: HTML\7f163366
-Node: LaTeX\7f169355
-Node: Texinfo\7f170129
-Node: Vietnamese\7f170901
-Node: African\7f171877
-Node: Others\7f173227
-Node: Texte\7f174681
-Ref: Texte-Footnote-1\7f179231
-Ref: Texte-Footnote-2\7f179311
-Ref: Texte-Footnote-3\7f179786
-Node: Mule\7f179883
-Ref: Mule-Footnote-1\7f181664
-Node: Surfaces\7f182183
-Ref: Surfaces-Footnote-1\7f185602
-Node: Permutations\7f185706
-Node: End lines\7f186547
-Node: MIME\7f188748
-Node: Dump\7f189935
-Node: Test\7f194105
-Node: Internals\7f196583
-Node: Main flow\7f197811
-Node: New charsets\7f200914
-Node: New surfaces\7f205452
-Node: Design\7f206178
-Ref: Design-Footnote-1\7f215344
-Node: Concept Index\7f215448
-Node: Option Index\7f230191
-Node: Library Index\7f233044
-Node: Charset and Surface Index\7f237619
+Node: Request level\7f72193
+Node: Task level\7f83140
+Node: Charset level\7f93562
+Node: Errors\7f94404
+Ref: Errors-Footnote-1\7f99250
+Ref: Errors-Footnote-2\7f99364
+Node: Universal\7f99725
+Ref: Universal-Footnote-1\7f102837
+Ref: Universal-Footnote-2\7f102903
+Node: UCS-2\7f103116
+Node: UCS-4\7f105642
+Node: UTF-7\7f106182
+Node: UTF-8\7f106777
+Node: UTF-16\7f111082
+Node: count-characters\7f112230
+Node: dump-with-names\7f112901
+Node: iconv\7f115450
+Node: Tabular\7f118881
+Node: ASCII misc\7f141094
+Node: ASCII\7f141460
+Node: ISO 8859\7f142276
+Node: ASCII-BS\7f144570
+Node: flat\7f146407
+Node: IBM and MS\7f147078
+Node: EBCDIC\7f147622
+Node: IBM-PC\7f149718
+Ref: IBM-PC-Footnote-1\7f151832
+Node: Icon-QNX\7f151991
+Node: CDC\7f152416
+Node: Display Code\7f154097
+Ref: Display Code-Footnote-1\7f156378
+Node: CDC-NOS\7f156583
+Node: Bang-Bang\7f158545
+Node: Micros\7f160474
+Node: Apple-Mac\7f160857
+Node: AtariST\7f162891
+Node: Miscellaneous\7f163877
+Node: HTML\7f164610
+Node: LaTeX\7f170599
+Node: Texinfo\7f171373
+Node: Vietnamese\7f172145
+Node: African\7f173121
+Node: Others\7f174471
+Node: Texte\7f175925
+Ref: Texte-Footnote-1\7f180475
+Ref: Texte-Footnote-2\7f180555
+Ref: Texte-Footnote-3\7f181030
+Node: Mule\7f181127
+Ref: Mule-Footnote-1\7f182908
+Node: Surfaces\7f183427
+Ref: Surfaces-Footnote-1\7f186846
+Node: Permutations\7f186950
+Node: End lines\7f187791
+Node: MIME\7f189992
+Node: Dump\7f191179
+Node: Test\7f195349
+Node: Internals\7f197827
+Node: Main flow\7f199055
+Node: New charsets\7f202158
+Node: New surfaces\7f206696
+Node: Design\7f207422
+Ref: Design-Footnote-1\7f216588
+Node: Concept Index\7f216692
+Node: Option Index\7f231727
+Node: Library Index\7f234580
+Node: Charset and Surface Index\7f239155
 \1f
 End Tag Table

diff --git a/doc/recode.texi b/doc/recode.texi
index 0464332331a1d78593d791a42bb15333ac8d2ff7..0fe7eec691eb5f911fbde9e2227da003dd9283b5 100644 (file)
--- a/doc/recode.texi
+++ b/doc/recode.texi
@@ -1725,6 +1725,7 @@ at outer level, and then, various functions at request level.
 @section Outer level functions
 
 @cindex outer level functions
+
 The outer level functions mainly prepare the whole recoding library for
 use, or do actions which are unrelated to specific recodings.  Here is
 an example of a program which does not really make anything useful.
@@ -1755,6 +1756,22 @@ his program (let's assume the programmer is a male, here, no prejudice
 intended).  This @samp{outer} variable is given as a first argument to
 all outer level functions.
 
+@cindex speed considerations
+The @code{RECODE_OUTER} structure is really meant to be initialised only
+once in the life of a program, and terminated with the program itself.
+Program interfaces should pay attention to initialise it only once,
+would it be only for speed considerations.  A good deal of overhead goes
+to outer level initialization, and if the outer level was initialized
+afresh for each and every string translated, say, the Recode library
+would appear immensely much slower that it was meant to be!
+
+@cindex memory leaks
+@cindex leaks, memory
+Because outer level initialization is meant to be done only once, not so
+much attention has been paid to avoid memory leaks at this level within
+Recode.  This is hardly a reason for not plugging such leaks at any
+level: in the long run, they should all be chased and repaired.
+
 @cindex @code{stdbool.h} header
 @cindex @code{bool} data type
 The @code{<recode.h>} header file uses the Boolean type setup by the
@@ -1899,10 +1916,11 @@ may be used by the library @emph{when} the user sets it to diagnose itself.
 @cindex request level functions
 The request level functions are meant to cover most recoding needs
 programmers may have; they should provide all usual functionality.
-Their API is almost stable by now.  To get started with request level
-functions, here is a full example of a program which sole job is to filter
-@code{ibmpc} code on its standard input into @code{latin1} code on its
-standard output.
+Their API is almost stable by now.
+
+To get started with request level functions, here is a full example of
+a program which sole job is to filter @code{ibmpc} code on its standard
+input into @code{latin1} code on its standard output.
 
 @example
 @group
@@ -1938,6 +1956,15 @@ which the programmer should use for allocating a variable in his program.
 This @var{request} variable is given as a first argument to all request
 level functions, and in most cases, may be considered as opaque.
 
+@cindex speed considerations
+Suppose an application is doing a lot of recoding using only a few
+different requests.  For speed considerations, the @code{RECODE_REQUEST}
+structure should ideally be cached for each kind of request, so the
+request level initialisation is not redone for each and every string
+translated.  The speedup should be more apparent when Recode is able
+to optimize the work by building on the fly, within the structure, new
+specialized recoding steps and their associated data tables.
+
 @itemize @bullet
 @item Initialisation functions
 @cindex initialisation functions, request

diff --git a/src/ChangeLog b/src/ChangeLog
index 944143e24242bdf65f7aec3e5a66d22cb9c0a36a..b70d2f113104cea2cb921368dbe72fd96db5bfe8 100644 (file)
--- a/src/ChangeLog
+++ b/src/ChangeLog
@@ -1,3 +1,11 @@
+2008-03-12  Andreas Schwab  <schwab@suse.de>
+
+       * names.c (alias_free): New function.
+       (prepare_for_aliases): Pass it to hash_initialize.
+       * outer.c (recode_delete_outer): Free elements of
+       argmatch_charset_array and argmatch_surface_array.
+       Also reported by Gaël Le Mignot and Pawel Krawczyk.
+
 2008-03-11  François Pinard  <pinard@iro.umontreal.ca>
 
        * recodext.h (struct recode_symbol): Add iconv_name.
@@ -267,7 +275,7 @@
        * recode.c (init_ucs2_to_byte): Free hash table after use.
        * testdump.c (produce_count): Free hash table entries and the table
        itself after use.  Fix clean-up code.
-\f
+
 2000-06-28  François Pinard  <pinard@iro.umontreal.ca>
 
        * task.c: Insert union wait portability from GNU make.

diff --git a/src/names.c b/src/names.c
index 4b33a4f352ef49a77e167b8ee53cf75b7f5c082a..6e3e86a368de424fd653d86a8ef6685b4cb05f6a 100644 (file)
--- a/src/names.c
+++ b/src/names.c
@@ -101,6 +101,20 @@ alias_comparator (const void *void_first, const void *void_second)
   return strcmp (first->name, second->name) == 0;
 }
 
+static void
+alias_free (void *void_alias)
+{
+  RECODE_ALIAS alias = void_alias;
+  struct recode_surface_list *list, *next;
+
+  for (list = alias->implied_surfaces; list; list = next)
+    {
+      next = list->next;
+      free (list);
+    }
+  free (alias);
+}
+
 bool
 prepare_for_aliases (RECODE_OUTER outer)
 {
@@ -108,7 +122,7 @@ prepare_for_aliases (RECODE_OUTER outer)
   outer->number_of_symbols = 0;
 
   outer->alias_table
-    = hash_initialize (800, NULL, alias_hasher, alias_comparator, free);
+    = hash_initialize (800, NULL, alias_hasher, alias_comparator, alias_free);
   if (!outer->alias_table)
     return false;
 

diff --git a/src/outer.c b/src/outer.c
index 1854d768ad0c8de12a0160ed6fc8aee485980ae0..1d95d93224e6b5d410755de0c57af97bff3b0c1b 100644 (file)
--- a/src/outer.c
+++ b/src/outer.c
@@ -602,9 +602,6 @@ bool
 recode_delete_outer (RECODE_OUTER outer)
 {
   unregister_all_modules (outer);
-  /* FIXME: Pawel Krawczyk reports that calling new_outer ... delete_outer
-     20000 times in a program has the effect of consuming all virtual memory.
-     So there might be memory leaks should to track down and resolve.  */
   while (outer->number_of_symbols > 0)
     {
       RECODE_SYMBOL symbol = outer->symbol_list;
@@ -626,7 +623,15 @@ recode_delete_outer (RECODE_OUTER outer)
   if (outer->alias_table)
     hash_free (outer->alias_table);
   if (outer->argmatch_charset_array)
-    free (outer->argmatch_charset_array);
+    {
+      char **cursor;
+
+      for (cursor = outer->argmatch_charset_array; *cursor; cursor++)
+       free (*cursor);
+      for (cursor = outer->argmatch_surface_array; *cursor; cursor++)
+       free (*cursor);
+      free (outer->argmatch_charset_array);
+    }
   if (outer->one_to_same)
     free ((void *) outer->one_to_same);
   free (outer);