pike.git
/
CHANGES
version
»
Context lines:
10
20
40
80
file
none
3
pike.git/CHANGES:1:
Changes since Pike 7.6.112 ----------------------------------------------------------------------
-
Changes
in Pike:
+
General
changes
in Pike:
-
o Fixed
infinite
loop
that
could
occur
when
feeding
Image
.
XBM
with
-
malformed
data
.
-
o Added
option
"file"
to
Image
.
PS.decode
().
-
o
Image
.
PS.decode
()
now
reads
nonblocking from
the
pipes
.
-
o Added
use
of
custom
backend
in
Image
.
PS
.
+
o Fixed
a
race
condition
in
Stdio.mkdirhier()
+
o
Fixed some memory problems in the backend pointed out by valgrind
+
and others.
+
o Fixed bug ingethrvtime
when
cpu_time_t
isn't 64 bit long
.
+
o
Fixed
bug
in finding out svalue type for trampolines
.
+
o Added
handling
of
ISO-6937:2001
in charset decoder
.
+
o Many dup2
()
fixes in Stdio
.
File.
+
o
Fixed Crypto
.
set_{en,de}crypt_key
()
to
accept
the
force
flag
.
+
o Added
alternate
backend
support
for
Stdio.sendfile().
+
o
More error handling added to Monger
.
+
o More error handling added to cgrep
.
o Added "Regexp.PCRE.Widestring" to the feature list. o Added ADT.Trie data type. o Fixed bug in columns sorting in ADT.Table.
-
o Fixed bug in a special case of GIF color decoding (bug 3536).
+
o Encoder, Decoder and Codec are now available in the Pike module. While the programs were always retrievable from the master object, now they can be inherited in codecs that wish to extend the master codecs. o The XMLRPC module can now serialize Calender objects. Only 200 responses will be attempted to be decoded. Fixed a bug in handling of string values.
-
+
+
Image related changes in Pike:
+
+
o Fixed infinite loop that could occur when feeding Image.XBM with
+
malformed data.
+
o Added option "file" to Image.PS.decode().
+
o Image.PS.decode() now reads nonblocking from the pipes.
+
o Added use of custom backend in Image.PS.
+
o Fixed bug in a special case of GIF color decoding (bug 3536).
+
+
Database related changes in Pike:
+
o In the Sql module, do not show the total row count after a big_query, if unavilable.
-
o Fixed a race condition in Stdio.mkdirhier()
-
o Fixed some memory problems in the backend pointed out by valgrind
-
and others.
-
o Fixed bug ingethrvtime when cpu_time_t isn't 64 bit long.
-
o Fixed bug in finding out svalue type for trampolines.
+
o Don't crash when not sending ssl options to Mysql. o Fix for case where ODBC tried to write error messages with wide strings in them. o Bugfixes for ODBC long fields. o Added unlocking of the interpreter lock for almost all ODBC operations. This is needed for simplistic odbc drivers that take their internal lock for all calls.
-
o Added handling of ISO-6937:2001 in charset decoder.
-
o Many dup2() fixes in Stdio.File.
-
o Fixed Crypto.set_{en,de}crypt_key() to accept the force flag.
-
o Added alternate backend support for Stdio.sendfile().
-
o win32: Fixed bug in RegGetValue{,s}() for fields of type *_SZ of length 0.
-
o Close the kqueue fd when a backend goes away. Should solve all
-
kinds of "no more file descriptor" and "couldn't create backend wakeup
-
pipe" problems on Darwin.
-
o More error handling added to Monger.
-
o More error handling added to cgrep.
+
o Added more methods to enable the buggy unicode decode support in Mysql.mysql: * "broken-unicode" as charset to create() or set_charset(). * set_unicode_decode_mode (-1). Note that all this only applies when compiling with a mysql client lib older than 4.1.0. Otherwise these methods are aliases to enabling the normal unicode decode support. This means that no code using broken-unicode should rely on the BINARY flag on text fields to avoid utf-8 decoding.
-
+
Mostly platform specific changes in Pike:
+
+
o win32: Fixed bug in RegGetValue{,s}() for fields of type *_SZ of length 0.
+
o Darwin: Close the kqueue fd when a backend goes away. Should solve all
+
kinds of "no more file descriptor" and "couldn't create backend wakeup
+
pipe" problems on Darwin.
+
Build fixes: o Mird: Renamed FDATASYNC() to MIRD_FDATASYNC() to avoid clash with NetBSD header-files. o The precompiler doesn't generate broken string width check code. o Fixed some problems with configure tests using rusage and threads. o Added /opt/csw/lib to list of library directories to check. o If compiler is GCC, also add all library directories listed by gcc -print-search-dirs. o Normalize library directory pathnames before adding them.