[10d2c0]: contrib / sb-bsd-sockets / NEWS  Maximize  Restore  History

Download this file

136 lines (81 with data), 4.2 kB

Changes in 0.58 - Sun Jan 12 00:53:53 GMT 2003

Fix db-sockets.asd so that it doesn't recompile alien.so every single

Announce anon-cvs repo for people to get in-between versions

MSG_NOSIGNAL is a linuxism, I'm told.

Changes in 0.57 - Wed Sep 11 12:27:32 2002

Fix for compilation bug reported by Andreas Fuchs.  Don't use 0.56, it
was a mistakenly uploaded file

Changes in 0.55 - Tue Sep 10 23:42:27 2002

Fix for a unix-domain sockets problem, courtesy of David Lichteblau

Changes in 0.54 - Wed Mar 6 2002

New version mostly due to new packaging format: this is now a
vendor-neutral cclan (vn-cclan) package.  See INSTALL file

Fixed bug in af_file support.

Changes in 0.53 - Thu Jan 31 2002

By popular request (two people, at last count) this works in CMUCL again.
Also, some documentation updates, a really silly bug in make-instance 
fixed, and support for the TCP_NODELAY socket option

Changes in 0.52 - Tue Jan 8 2002

Very few.  This release was put out a few days after 0.5.1 because
0.5.1 is less than 0.42, and various packaging tools tend to get
confused to see version numbers go backwards.

Changes in 0.5.1 - Mon Jan 7 2002

Support for AF_FILE (formerly known as Unix-domain) sockets; both
stream and datagram.  

MAKE-INET-SOCKET has been deprecated (but is still there).  New code is
encouraged to write (make-instance 'inet-socket ...) instead

Fairly pervasive low-level changes to avoid leaking quite as much
memory.  May also have fixed a file descriptor leak in the process.

Changes in 0.42

Repackaged to be a debian-like package, and use
common-lisp-controller, which required a reasonably large amount of
thrashing around renaming files and so on.

New function GET-HOST-BY-ADDRESS returns a HOST-ENT just like

Tested on SBCL (x86),  SBCL (Alpha),
CMUCL 18c+ 2.5.2 (x86)

Changes in 0.41 - Sun Jan 7 2001

Cleanups in the tests for more intelligible failure messages

SOCKET-ERROR conditions now inherit from ERROR not CONDITION - as
otherwise IGNORE-ERROR doesn't ignore them, which is unexpected

Tested on debian cmucl 2.4.19 , sbcl pre-0.6.9 snapshot of Nov 30 2000.

The latter doesn't build without manual intervention:

 error in function SB-C::%DEFCONSTANT:
    The constant INET-ADDRESS-ANY is being redefined.

(just continue)

Changes in 0.4 - Mon Jul 3 2000

Now works (passes tests) in

- Solaris 2.6 SPARC (CMUCL 18b)
- Debian x86 GNU/Linux (Debian CMUCL 2.4.19)
- Debian x86 GNU/Linux (SBCL 0.6.5)

Some CMUCL-on-FreeBSD changes (mostly involve commenting stuff out).  Doesn't 
work, though (but might in SBCL/FreeBSD)

The Solaris changes comprised disabling bits and fixing an 
endianness problem. 

Changes in 0.37 -  Sat May 20 2000

Changes from Martin Atzmueller to make it compile more cleanly in SBCL

Changes in 0.36 -  Thu May 11 2000

Some documentation cleanups


EINTR now generates a INTERRUPTED-ERROR condition

Changes in 0.35  - Mon May 1 2000

MAKE-INET-SOCKET now can take a keyword for PROTOCOL: it lowercases
the symbol's name, then looks it up using GET-PROTOCOL-BY-NAME

A bad bug in the CMUCL code (which caused the EXTENSIONS package to
disappear - oops...) was found and fixed

Changes in 0.3  -  Apr 17 2000 

Now works with SBCL (0.6.1, 0.6.2) in addition to CMUCL.

Fixed to actually work with a READ-SEQUENCE implementation that does
the right thing instead of the (suspected buggy) implementation in
CMUCL. At least, the Hyperspec doesn't give me any particular cause
for belief that READ-SEQUENCE can return before reading as much as the
user asks it to, which is what we were using it for hitherto.

The Makefile got a lot bigger. defs-to-lisp.lisp got a lot smaller.

Standard make target creates "sockets-system.x86f" which contains all
the code in a single file

If you want to build it on SBCL you'll need a working defsystem for
said platform first. This involves some fiddling around: first you
need to get it from CLOCC on Sourceforge then you need to patch it
with this diff. Unless you're looking at a version newer than 1.12, in
which case they might have patched it already before you