ID:               50189
 User updated by:  yoarvi at gmail dot com
 Reported By:      yoarvi at gmail dot com
 Status:           Open
 Bug Type:         *General Issues
 Operating System: Solaris 10 (SPARC)
 PHP Version:      6SVN-2009-11-16 (SVN)
 New Comment:

Updated patch using WORDS_BIGENDIAN (suggested by christopher dot jones
at oracle dot com)

Index: ext/standard/php_smart_str.h
===================================================================
--- ext/standard/php_smart_str.h        (revision 290471)
+++ ext/standard/php_smart_str.h        (working copy)
@@ -86,10 +86,17 @@
        smart_str_appendc_ex((dest), (c), 0)

 /* appending of a single UTF-16 code unit (2 byte)*/
+#ifndef WORDS_BIGENDIAN
 #define smart_str_append2c(dest, c) do {       \
        smart_str_appendc_ex((dest), (c&0xFF), 0);      \
        smart_str_appendc_ex((dest), (c&0xFF00 ? c>>8 : '\0'), 0);     
\
 } while (0)
+#else
+#define smart_str_append2c(dest, c) do {       \
+       smart_str_appendc_ex((dest), (c&0xFF00 ? c>>8 : '\0'), 0);     
\
+       smart_str_appendc_ex((dest), (c&0xFF), 0);      \
+} while (0)
+#endif

 #define smart_str_free(s) \
        smart_str_free_ex((s), 0)


Previous Comments:
------------------------------------------------------------------------

[2009-11-16 16:35:08] yoarvi at gmail dot com

ext/sqlite3/libsqlite/sqlite3.c uses
#if defined(i386) || defined(__i386__) || defined(_M_IX86)\
                             || defined(__x86_64) ||
defined(__x86_64__)


Is that better?

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

[2009-11-16 13:07:50] tokul at users dot sourceforge dot net

If is not "#if (defined(i386) || defined(__i386__) || defined(_X86_))
" vs others.

It is little endian vs big endian. I suspect that code should not
assume that all other archs are big endian.

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

[2009-11-16 12:20:44] yoarvi at gmail dot com

Description:
------------
zspprintf() incorrectly represents strings/chars as unicode characters
on Solaris (SPARC).

There are byte ordering differences for unicode representations between
x86 and SPARC:

For example, the unicode representation (i've grouped them in sets of
2chars) of '/tmp' on x86 is
'/''\0' 't''\0' 'm''\0' 'p''\0'

and on SPARC it is
'\0''/' '\0''t' '\0''m' '\0''p'

http://marc.info/?l=php-internals&m=125811990106419&w=2 has some more
details.

the problem seems to be in the smart_str_append2c macro that
zspprintf()/xbuf_format_converter end up using.

The following patch fixes the problem:
Index: ext/standard/php_smart_str.h
===================================================================
--- ext/standard/php_smart_str.h        (revision 290471)
+++ ext/standard/php_smart_str.h        (working copy)
@@ -86,10 +86,17 @@
        smart_str_appendc_ex((dest), (c), 0)

 /* appending of a single UTF-16 code unit (2 byte)*/
+#if (defined(i386) || defined(__i386__) || defined(_X86_))
 #define smart_str_append2c(dest, c) do {       \
        smart_str_appendc_ex((dest), (c&0xFF), 0);      \
        smart_str_appendc_ex((dest), (c&0xFF00 ? c>>8 : '\0'), 0);     
\
 } while (0)
+#else
+#define smart_str_append2c(dest, c) do {       \
+       smart_str_appendc_ex((dest), (c&0xFF00 ? c>>8 : '\0'), 0);     
\
+       smart_str_appendc_ex((dest), (c&0xFF), 0);      \
+} while (0)
+#endif

 #define smart_str_free(s) \
        smart_str_free_ex((s), 0)



Reproduce code:
---------------
% sapi/cli/php
ext/spl/tests/DirectoryIterator_getBasename_basic_test.php

Expected result:
----------------
getBasename_test

Actual result:
--------------
php goes into an infinite loop


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


-- 
Edit this bug report at http://bugs.php.net/?id=50189&edit=1

Reply via email to