1 2005-12-05 Matthias Clasen <mclasen@redhat.com>
3 * glib/gatomic.c: In the ia64 implementation, use
4 __sync builtin without _si or _di suffix. (#321229,
5 Stanislav Brabec, patch by Andreas Schwab)
7 2005-12-04 Behdad Esfahbod <behdad@gnome.org>
9 * glib/gslice.h: Remove comma at the end of enum.
11 2005-12-04 Matthias Clasen <mclasen@redhat.com>
13 Handle multiple user names with the same UID better.
14 (#319535, Laszlo Peter)
16 * glib/gutils.c (g_get_any_init_do): When determining user
17 data, first look up $LOGNAME. If the UID doesn't match
18 getuid(), fall back to the current behaviour of looking
19 up the user data based on getuid().
21 2005-12-04 Matthias Clasen <mclasen@redhat.com>
23 * glib/gmacros.h (G_GNUC_WARN_UNUSED_RESULT): Add a macro
24 to make gcc warn if a function result is ignored. (#145466,
25 Arjan van de Ven, Alex Larsson)
27 * glib/gmem.h: Add the new attribute to g_realloc and
30 2005-12-03 Matthias Clasen <mclasen@redhat.com>
33 * glib/gcache.h: Deprecate g_cache_value_foreach. (#322956,
36 * glib/gmappedfile.c: Make mapping of empty files
39 * glib/gfileutils.c: Don't fork a new process just to
40 fix the permissions of the created temp file. (#321318,
41 Alexis S. L. Carvalho)
43 2005-12-02 Matthias Clasen <mclasen@redhat.com>
45 * README.in: Add a note about Solaris threads.
48 * configure.in: Remove support for Solaris threads.
49 (#136971, Sebastian Wilhelmi, patch by Andrew Paprocki)
51 2005-12-02 Matthias Clasen <mclasen@redhat.com>
53 * configure.in: Define G_GUINT64_CONSTANT in analogy to
54 G_GINT64_CONSTANT. (#322568, Andrew Paprocki)
56 2005-12-02 Matthias Clasen <mclasen@redhat.com>
58 * glib/gslice.c: Win32 portability fixes and C99ism removal,
59 pointed out by Kazuki Iwamoto. (#323052)
61 2005-12-02 Matthias Clasen <mclasen@redhat.com>
63 * glib/gatomic.c (g_atomic_int_add): Add a missing volatile for
64 the IA64 implementation.
66 Fri Dec 2 16:18:09 2005 Tim Janik <timj@imendio.com>
68 * glib/gslice.c: some naming and type size fixups.
70 Fri Dec 2 13:08:58 2005 Tim Janik <timj@gtk.org>
72 * glib/gslice.h (g_slice_free): reworked GCC-specific type-safe macro
73 variant into something less verbose, but digestible for gcc-3.4.
75 Fri Dec 2 10:55:07 2005 Tim Janik <timj@imendio.com>
77 * tests/slice-test.c: extended to perform the benchmarking on the old
78 memchunk code if 'O' is selected.
80 * tests/memchunks.c: new file which contains the old GLib mem chunks
81 implementation with prefix old_mem_chunk_.
83 * tests/Makefile.am: added memchunks.c
85 Fri Dec 2 00:16:59 2005 Tim Janik <timj@imendio.com>
87 * glib/gslice.c: improved gettimeofday() timer resolution by maintaining
88 acceptable load distribution of the syscall.
89 (allocator_get_magazine_threshold): reduce minimum/single-thread magazine
90 sizes to approximately page_size/2.5 to avoid excessive startup allocations.
92 2005-12-01 Matthias Clasen <mclasen@redhat.com>
94 * tests/gobject/ifacecheck.c (test_iface_base_init): Don't
95 double-free base interface members.
97 * tests/Makefile.am: Remove duplicate variable.
99 Thu Dec 1 17:32:46 2005 Tim Janik <timj@imendio.com>
101 * glib/gslice.[hc]: new slice allocator implementation.
103 * tests/slice-test.c: added random slice allocation test.
105 * glib/gthread.[hc]: removed newly added private thread mem API.
107 * glib/gthreadinit.h:
110 * glib/gmem.c: divided glib threading initialisation into three phases,
111 initialisation where private keys and messaging are not available (only
112 needed by gmem.c), initialisation without messaging but private keys
113 available (gslice.c, gmessage.c), and full fledged initialisers that
114 server the rest of glib. initialisation functions got renamed to reflect
115 the limitations of their corresponding phases.
117 * glib/gmem.c: removed memchunk code, defer allocations to
120 * glib/gmem.[hc]: removed g_slice_* skeletons.
122 * glib/glib.symbols: added g_slice_* symbols.
124 * configure.in: check for availability of posix_memalign(3), memalign(3)
127 * glib/Makefile.am: added gslice.[hc].
129 2005-12-01 Tor Lillqvist <tml@novell.com>
131 * glib/gstdio.c (g_stat): In the Win32 implementation, strip
132 trailing slash(es) for non-root folders. stat() fails if non-root
133 folders are specified with trailing slashes. It's too much hassle
134 to demand that callers strip such slashes themselves, especially
135 as it is easy to get it wrong and strip the slash of a root
137 (g_rename): On NT-based Windows, use MoveFileEx() with
138 MOVEFILE_REPLACE_EXISTING to better match Unix behaviour.
140 2005-11-28 Matthias Clasen <mclasen@redhat.com>
142 Fix G_STMT_START / G_STMT_END on Solaris. (#321972,
145 * configure.in: Check whether do { } while (0) works.
147 * glib/gmacros.h: Use do { } while (0) for G_STMT_START /
148 G_STMT_END if it works.
150 2005-11-28 Matthias Clasen <mclasen@redhat.com>
152 * glib/gthread.c (g_static_rw_lock_wait, g_static_rw_lock_signal):
153 * glib/gnode.c (g_node_depth_traverse_level):
154 * glib/gmem.c (g_allocator_new):
155 * glib/ghash.c (g_hash_table_unref, g_hash_table_destroy)
156 (g_hash_table_foreach_remove_or_steal): Silence compiler
159 2005-11-27 Matthias Clasen <mclasen@redhat.com>
161 * glib/gunicollate.c (g_utf8_collate_key_for_filename):
162 Don't read beyond len. (#322520, Christian Persch)
164 Wed Nov 23 17:34:01 2005 Tim Janik <timj@imendio.com>
166 * glib/gdataset.c: access datalist flags via atomic pointer access
167 functions, instead of acquiring the dataset lock. this is faster and
168 also matches the atomic pointer readouts in gdatalistprivate.h and
169 direct pointer modifications required by gobject.c.
171 Wed Nov 23 13:35:31 2005 Tim Janik <timj@gtk.org>
173 * glib/gdataset.c: streamlined initialization code somewhat,
174 removed GData node cache.
176 2005-11-22 Matthias Clasen <mclasen@redhat.com>
178 * glib/ghash.c (g_hash_table_ref):
179 (g_hash_table_unref): Mark these as new API.
181 Avoid double locking in g_intern_string (#322133,
184 * glib/gdataset.c (g_quark_from_string_internal): New
185 internal function which factors out common parts of
186 g_quark_from[_static]_string.
187 (g_quark_from_string, g_quark_from_static_string):
188 Use g_quark_from_string_internal.
189 (g_intern_string, g_intern_static_string): Use
190 g_quark_from_string_internal, and only take the
192 (g_quark_new): Don't store the strings shifted by -1
193 in the g_quarks array.
194 (g_quark_to_string): Adapt to the previous change.
196 Tue Nov 22 14:04:26 2005 Tim Janik <timj@imendio.com>
200 g_hash_table_new_full(): create hash tables with a ref count of 1.
201 g_hash_table_ref(): atomically ref_count+=1
202 g_hash_table_unref(): atomically ref_count-=1, destroys hash table
203 when refcount reaches 0.
204 g_hash_table_destroy(): just destroy keys and values, unref by 1.
205 g_hash_table_insert():
206 g_hash_table_replace(): assert ref_count>0.
209 * glib/gatomic.c: added 'volatile' qualifier to all atomic pointer and
212 2005-11-20 Behdad Esfahbod <behdad@gnome.org>
214 * glib/guniprop.c (g_unichar_get_mirror_char): Remove unused
215 variables. (#321984, Andrew Paprocki)
217 2005-11-18 Matthias Clasen <mclasen@redhat.com>
219 * configure.in: Bump version
221 2005-11-17 Matthias Clasen <mclasen@redhat.com>
223 * === Released 2.9.0 ===
225 * glib/glib.symbols: Updates
227 * glib/gutf8.c (g_utf8_offset_to_pointer): Add some performance
228 hints to the docs. (Paolo Borelli)
232 * glib/gthread.c (g_thread_foreach): Mark as new api.
234 * README.in: Updates.
238 * glib/gdate.c (g_date_set_time_t):
239 (g_date_set_time_val): Add functions to set a date from
240 a time_t and from a GTimeVal.
241 (g_date_set_time): Deprecate. (#314067, Roger Leigh)
244 * tests/utf8-pointer.c: Unit tests for g_utf8_pointer_to_offset
245 and g_utf8_offset_to_pointer.
247 * glib/gutf8.c (g_utf8_pointer_to_offset)
248 (g_utf8_offset_to_pointer): Handle negative offsets, and use
249 "stutter stepping" for going backwards. (#320638, Larry
254 * glib/gthread.c: const correctness fixes, found
255 by Arjan van de Ven and gcc.
257 2005-11-16 Behdad Esfahbod <behdad@gnome.org>
259 * docs/reference/glib/tmpl/unicode.sgml: Correct typo on mentioning
260 @G_UNICODE_BREAK_UNKNOWN that should be #G_UNICODE_BREAK_UNKNOWN.
262 2005-11-09 Behdad Esfahbod <behdad@gnome.org>
264 * glib/guniprop.c: Use bit hacks instead when checking a general
265 category value against multiple values.
267 * glib/gutf8.c: Change ISO10646 to Unicode in docs.
269 2005-11-10 Simos Xenitellis <simos@gnome.org>
271 * configure.in: Added tt (Tatar) to ALL_LINGUAS.
273 2005-11-05 Tor Lillqvist <tml@novell.com>
275 * glib/giowin32.c (g_io_win32_set_flags): Don't call g_warning().
276 (#320688) Instead set error and return failure.
278 2005-11-07 Matthias Clasen <mclasen@redhat.com>
280 * glib/gmain.c (g_main_dispatch): Don't call cb_funcs->unref
281 while holding the context lock. (#320886, Andy Wingo)
283 2005-11-07 Billy Biggs <vektor@dumbterm.net>
285 * tests/refcount/Makefile.am: Add a missing reference to libglib
286 in the LDADD for this test.
288 2005-11-07 Matthias Clasen <mclasen@redhat.com>
290 * glib/gmem.c (g_allocator_new): We need to set n_preallocs to a
291 nonzero value, otherwise GTK+ 2.8 breaks when compiled against
292 GLib 2.9. (#320755, Luca Ferretti)
294 2005-11-04 Matthias Clasen <mclasen@redhat.com>
296 * glib/Makefile.am: Apply a patch from Behdad Esfahbod to
297 use a faster lookup table for g_unichar_get_mirror_char().
298 * glib/gmirroringtable.h: The new table.
299 * glib/gunichartables.h: Remove bidi_mirroring_table.
300 * glib/gen-unicode-tables.pl: Don't generate the mirroring
302 * glib/glib-mirroring-tab/*: Sources for the program
303 which generated gmirroringtable.h.
305 * glib/glist.c: Avoid some code duplication.
307 * glib/gscanner.h: Include gdataset.h. (#320322)
309 * glib/gdebug.h: add new GDebugFlag for fatal_criticals
310 * glib/gmessages.c: (_g_debug_init): handle G_DEBUG=fatal_criticals,
311 to help find critical warnings in applications. (#320017,
314 2005-11-02 Tor Lillqvist <tml@novell.com>
316 * glib/glib.symbols: Remove large amount of trailing whitespace
317 from one line. Remove #ifdef G_UNDEFINED from around
318 g_hash_table_new and g_hash_table_new_full symbols, what was the
319 purpose of that? Add the g_slice functions.
321 Tue Nov 1 16:24:20 2005 Tim Janik <timj@imendio.com>
323 * glib/gmem.[hc]: prepared deprecation of GMemChunk and GAllocator.
324 added g_slice_*() API to allocate and cache small bits of memory.
325 an actuall allocator implementation for g_slice_*() is still pending.
327 * glib/gthread.[hc]: changes from a patch by Matthias Clasen.
328 changed GRealThread list to use in-structure *next; fields instead
329 of GSList, in order for thread iteration to not depenend on g_slice_*()
331 _g_thread_mem_private_get():
332 _g_thread_mem_private_set(): added accessors for private memory,
333 needed because the ordinary GPrivate implementation relies on GArray
334 and GSList and therefore indirectly on working g_slice_*() allocations.
337 g_thread_foreach(): new public API function to loop over all existing threads.
351 removed GAllocator and free list usages and accompanying locks.
352 use g_slice_*() API to allocate and cache small bits of memory.
354 * glib/ghook.h: removed GMemChunk field from public API.
357 * glib/glist.h: deprecate allocator API, provide _free1() for consistency.
359 * glib/gnode.h: deprecate allocator API.
361 * glib/gmain.c: reordered GPollRec fields so g_slice_free_chain() can
362 be used for poll rec lists.
364 * glib/grel.c: removed mem chunk usage, and allocated tuples via g_slice_*().
365 g_relation_destroy(): free all tuples from the all_tuples hash table,
366 this effectively maintains the life time track keeping of tuples.
367 g_relation_delete_tuple(): free tuples which are removed from the
368 all_tuples hash table. this fixes a temporary leak that was present
369 in the memchunk code until the destruction of the relation.
371 2005-10-29 Matthias Clasen <mclasen@redhat.com>
373 * tests/convert-test.c: Add some tests for conversions between
374 UTF-8, UCS-4 and UTF-16.
376 * glib/gutf8.c (g_utf8_to_ucs4, g_utf8_to_utf16): Fix handling
377 of len == -1, noticed by Morten Welinder.
379 2005-10-27 Erdal Ronahi <erdal.ronahi@gmail.com>
381 * configure.in: Added ku (Kurdish) to ALL_LINGUAS
383 2005-10-26 Matthias Clasen <mclasen@redhat.com>
385 * glib/gutf8.c (g_ucs4_to_utf8): Don't set items_read twice
386 in the error case, and add some documentation. (#319806, Morten
389 2005-10-19 Manish Singh <yosh@gimp.org>
391 * configure.in: Use AC_CHECK_FUNCS for _NSGetEnviron, to get the
392 config.h symbol automatically. Fixes bug #313731.
394 2005-10-19 Tor Lillqvist <tml@novell.com>
396 * glib/gwin32.c (g_win32_get_package_installation_directory):
397 Return a g_strdup()ed copy of the value stored in the hash table,
398 so that it can be g_free()d without leaving a dangling pointer in
399 the hash table. (#319232)
401 2005-10-06 Matthias Clasen <mclasen@redhat.com>
403 * glib/gunicollate.c (g_utf8_collate_key_for_filename): Handle
404 all-zero sequences correctly. (#317930, Sebastien Bacher)
406 2005-10-05 Matthias Clasen <mclasen@redhat.com>
408 * glib/gmarkup.c (g_markup_escape_text): Doc addition
410 * Makefile.am: Add Collation tests.
412 * tests/collate/*: Inputs and expected outputs for collation tests.
414 * tests/run-collate-tests.sh: Script to run collation tests.
416 * tests/unicode-collate.c (main): Rework slightly to make
417 it usable in unit tests. Also test g_utf8_collate_key_for_filename().
419 2005-10-01 Behdad Esfahbod <behdad@gnome.org>
421 * docs/reference/glib/tmpl/unicode.sgml:
422 * glib/gen-unicode-tables.pl:
424 * glib/gunichartables.h:
426 * tests/casefold.txt:
427 * tests/casemap.txt: Updated to Unicode 4.1. There are five new
428 GUnicodeBreakType types. That may break some applications, like
431 2005-09-26 Matthias Clasen <mclasen@redhat.com>
433 * glib/gstrfuncs.c (g_ascii_strtoull): Add details to the
434 docs. (#314393, Matthew F. Barnes)
437 * glib/gprintf.h: Remove g_snprintf() and g_vsnprintf(), since
438 they are already declared in glib.h. This doesn't break documented
439 use of gprintf.h, but should probably be pointed out in the
440 release notes for 2.10. (#314232, Behdad Esfahbod)
442 Tue Sep 20 13:16:04 2005 Tim Janik <timj@imendio.com>
444 * glib/gpattern.c (g_pattern_ph_match): applied significant recursion
445 complexity optimization, based on a patch from Matthias Clasen.
447 * tests/patterntest.c: more tests, mostly from matthias.
449 2005-09-20 Matthias Clasen <mclasen@redhat.com>
451 * glib/gqueue.c (g_queue_insert_sorted): Correct the docs.
452 (#316703, Mark Drago)
454 Mon Sep 19 17:23:23 2005 Tim Janik <timj@imendio.com>
456 * glib/gpattern.c: applied a patch from matthias which checks on the
457 upper bound of GPatternSpec length to optimize matches.
460 * tests/patterntest.c: added more match cases.
462 2005-09-16 Tor Lillqvist <tml@novell.com>
464 * glib/gstrfuncs.c (g_ascii_strcasecmp, g_ascii_strncasecmp): Add
465 warning to doc comment that these functions should not be used on
466 encodings like CP932.
468 2005-09-14 Matthias Clasen <mclasen@redhat.com>
470 * tests/keyfile-test.c: Add a test for grup names of length 1.
472 * glib/gkeyfile.c (g_key_file_line_is_group): Accept group names
473 of length 1. (#316309)
475 2005-09-12 Matthias Clasen <mclasen@redhat.com>
477 * glib/gmarkup.c (g_markup_escape_text): Clarify docs.
479 2005-09-11 Sebastian Wilhelmi <seppi@seppi.de>
481 * tests/refcount/Makefile.am (INCLUDES): Link the the refcount
482 tests to the system thread library $(G_THREAD_LIBS). Fixes #313744
485 2005-09-11 Kjartan Maraas <kmaraas@gnome.org>
487 * glib/gmain.c: (g_child_watch_prepare), (g_child_watch_check),
488 (child_watch_helper_thread): Remove some dead code. Closes
491 2005-09-07 Tor Lillqvist <tml@novell.com>
493 * glib/Makefile.am: Create also a console version of the
494 gspawn-win32-helper program, gspawn-win32-helper-console.exe.
495 It's otherwise identical to gspawn-win32-helper.exe, except marked
496 as a console application (linked without the -mwindows option).
498 * glib/gspawn-win32.c (do_spawn_directly, do_spawn_with_pipes):
499 Drop the dont_wait parameter. Its truth value correlated 100% with
500 the NULLness of the exit_status parameter anyway, so it's enough
501 to check whether exit_status is NULL. Invert the sense of the
502 dont_return_handle parameter and rename it to do_return_handle, to
503 make the code easier to read by avoiding double negations.
505 (g_spawn_sync_utf8, g_spawn_async_with_pipes_utf8): Modify calls
506 to do_spawn_with_pipes() accordingly.
508 (do_spawn_with_pipes): If we have a console, use the console
509 version of the helper program, otherwise use the GUI one. This
510 avoids extra console windows opening up in some situations. (In
511 case a console application uses the GUI gspawn-win32-helper.exe to
512 spawn another console application we would get a separate console
513 for the spawned console application).
515 * glib-zip.in: Distribute also gspawn-win32-helper-console.exe.
517 2005-09-05 Matthias Clasen <mclasen@redhat.com>
519 * glib/gmappedfile.c (g_mapped_file_new): Report an error
520 if the file is too large. (#315275, Kjartan Maraas)
522 * glib/gkeyfile.c (g_key_file_load_from_fd): The return value
523 of read() is signed. (#315273, Kjartan Maraas)
525 2005-08-31 Tor Lillqvist <tml@novell.com>
527 * glib/gutils.h: Wrapping atexit() is a bad idea on Windows, where
528 the EXE and each DLL have their own atexit function chains.
530 #define g_atexit as atexit instead. This means it has a
531 better chance of doing what the caller wants. For instance,
532 gtkhtml calls g_atexit() registering a function in gtkhtml
533 itself. This caused a crash when g_atexit() was implemented as a
534 function in the GLib DLL. The gtkhtml DLL was already unloaded by
535 the time the GLib DLL got unloaded.
537 * glib/gutils.c: #undef the #define mentioned above, to also get a
538 real g_atexit() into the DLL for backward compatibility. Document
539 the Windows behaviour of g_atexit(), and document the varying ways
540 atexit() can behave in the context of dynamically loaded modules
543 2005-08-31 Matthias Clasen <mclasen@redhat.com>
547 * glib/gdataset.c: Add string interning functions.
549 2005-08-28 Matthias Clasen <mclasen@redhat.com>
551 * glib/giochannel.c: Unify some near-duplicate strings. (#314654,
554 2005-08-26 Matthias Clasen <mclasen@redhat.com>
556 * configure.in: Bump version to 2.9.0
558 2005-08-25 Tor Lillqvist <tml@novell.com>
560 Make also the g_spawn*() functions take parameters in the GLib
561 file name encoding, i.e. UTF-8, on Windows. Has no impact on Unix
562 API or ABI. Like the other GLib API that was earlier changed to
563 use UTF-8 on Windows, the names of the functions that take UTF-8
564 have _utf8 suffixes added by using preprocessor macros in the
565 header file. The old names are kept for functions with the old
566 behaviour, taking parameters in the system codepage, for DLL ABI
569 * glib/gspawn.h: On Win32 add the suffix _utf8 to the names of the
570 g_spawn*() functions.
572 * glib/gspawn-win32.c: Use wide-char API on NT-based
573 Windows. Convert parameters from UTF-8 to wide chars (NT) or
574 system codepage (Win9x) and call the C library _wspawn*() or
575 spawn*() functions respectvely. Add DLL ABI stability versions
576 that take parameters in the system codepage.
578 * glib/gspawn-win32-helper.c: On NT-based Windows use the
579 wide-char versions of argv and envp, and use wide-char API to
580 change directory and spawn the program to run. Remove the verbose
581 debugging output, it was too complex to modify for the wide-char
582 features. (Just add temporary debugging printouts if needed, no
583 need to have them permanently in the source.)
585 * glib/gspawn.c: Corresponding documentation updates.
587 * glib/glib.symbols: Corresponding changes: Mark the ABI stability
588 symbols as PRIVATE, add the new _utf8-suffixed ones.
590 2005-08-24 Stepan Kasal <kasal@ucw.cz>
592 * glib/gtypes.h (G_MININT64): Cast the constant to gint64; it is
593 guint64 otherwise and that can produce warnings about comparison
594 between signed and unsigned.
596 2005-08-23 Matthias Clasen <mclasen@redhat.com>
598 * glib/gutils.c: Fix the crt_externs.h include.
600 2005-08-23 Stepan Kasal <kasal@ucw.cz>
602 * NEWS: Fix spelling of my first name.
604 2005-08-23 Matthias Clasen <mclasen@redhat.com>
608 * === Released 2.8.1 ===
612 2005-08-20 Hans Breuer <hans@breuer.org>
614 * glib/makefile.msc.in : link with ws2_32.lib
616 2005-08-18 Tor Lillqvist <tml@novell.com>
618 * configure.in: Check for <sys/wait.h>
620 * glib/gbacktrace.c: Include <sys/wait.h> on if HAVE_SYS_WAIT_H.
622 2005-08-18 Ross Burton <ross@burtonini.com>
625 Optimise single-character insertions.
630 * tests/string-test.c:
631 Add tests for new optimisation, and fix a leak.
633 2005-08-17 Matthias Clasen <mclasen@redhat.com>
635 * configure.in: Check for crt_externs.h and _NSGetEnviron.
637 * glib/gutils.c: On Darwin, include crt-externs.h and
638 define environ using _NSGetEnviron(). (#313731)
640 2005-08-16 Stepan Kasal <kasal@ucw.cz>
642 * glib/gutils.c (g_get_any_init): Move the body of the big if...
643 (g_get_any_init_do): ... to this new function.
644 (g_get_any_init): Declare as inline.
645 (g_get_any_init_locked): New inline function, does the locking.
646 Make use of these two throughout the code.
648 2005-08-15 Matthias Clasen <mclasen@redhat.com>
650 * glib/gbacktrace.c (g_on_error_stack_trace): Wait for
651 the child process and then simply return. This makes
652 The "S" option work as documented in g_on_error_query().
653 (#313125, Matthew F. Barnes)
655 * glib/gunicode.h: Update the link to Unicode category
656 values. (#313369, Behnam Esfahbod)
658 * glib/gqueue.c (g_queue_find_custom): Clarify docs
659 a little. (#311727, Tristan van Berkom)
661 * glib/abicheck.sh, gobject/abicheck.sh: Make the
662 check work on ia64 too, where some symbols ended up
663 in yet another section.