;Reading at top level in Lisp Listener 2. ;Reading in base 8 in package USER with standard Zetalisp readtable. (load-patches :noselective) Checking whether site configuration has changed... Loading site files on account of SYS: SITE; SITE QFASL > Loading site files on account of SYS: SITE; HSTTBL QFASL > Patches for System (Current version is 125.19): 125.8 pld: (= 4\5 0.8s0) returned NIL because (floatp 0.8s0) returns T, causing 4\5 to be coerced to a float for comparison, not a short-float. Unfortunately, (= 0.8 0.8s0) returns NIL, which can't be easily fixed. We CAN (correctly) coerce 4\5 to a short-float, however, and solve the initial problem... 125.9 pld: Problems with describing an integer type. If you do: (setq a 4) (check-type a (integer 5)) you get: error: a is not an integer greater than 5 Fixed to be error: a is not an integer greater than or equal to 5 If you then do: (check-type a (integer (4))) you get: error: a is not an integer greater than 5 Fixed to be error: a is not an integer greater than 4 Similar fixes for when the upper bound is specified inclusively or exclusively. 125.10 pld: lam:dtp-rplacd-forward really needs to have a value. 125.11 pld: QFILE bug: (fs:host-chaos-interrupt-function) failed to match asynchronous marks with their file handles. One noticable example occured when the server got a Disk Full error and the user was simply left hanging. Now you are given the debugger with useful proceed options including: running Dired, calling (zwei:clean-directory), expunging the directory. 125.12 pld: Fixes to Chaos routing: - packets for this host always go through Loopback interface - packets for other processors on same bus go through Share interface, even if both this processor and the other one have Ethernet interfaces. - packets for other processors on same bus go directly to that processor, not through the Ethernet owner! (matters only to 2X2+ or 3X3) 125.13 pld: more intelligent calculation of line length for finger-arg-prompt 125.14 pld: The 3com microcode driver had a bug: the packet length returned by the 3com board had 20 bytes of overhead added in that was not being subtracted out before setting the int-pkt fill-pointer. This fails now that set-fill-pointer does bounds checking. 125.15 keith: LOAD-PATCH-FILE used to very strictly enforce the restriction that "The use of LOAD in a patch file is not supported." This was done by binding the function LOAD to an error routine. Unfortunately, this had a horrible side-effect: if you warm-booted in the middle of this error, the new "temporary" function value of LOAD became permanent. Now, LOAD itself checks the value of FS:THIS-IS-A-PATCH-FILE and issues a continuable, not fatal, error. You can proceed if you dare. I believe this is correct; sometimes a patch may need to LOAD something like a font file. 125.16 keith: When you get a background error with windows locked, the who-line message now tells the truth about which keys to press to resolve the problem. The key for clearing window-system locks is Terminal Control-Clear-Input, Not M-C-Clear-Input. (I always wondered why that didn't work.) 125.17 keith: Rationalize treatment of various READer end-of-file conditions. 1. Packages not followed by symbols -- reading things like "foo:" used to blow up with a datatype comparison error from (=) . [I type things like that a lot when I stay up too late.] Now, signals a CERROR, so in a listener it just prompts "EOF after package prefix". 2. Used to be, all READ-END-OF-FILE errors had the proceed option that said, "Close off all unfinished lists." What a crock - that hardly applies to strings and (now) unterminated symbols. I defined flavors and separate :NO-ACTION proceed methods for these other read-end-of-file types. Now: a) with an unterminated string, readers offer to return the null string. That's what you got before. b) with an unterminated symbol, readers offer to return NIL. That's new. c) with an unterminated list, you get the same behavior as before. 125.18 saz: Cosmetic change to error strings as processed by QFILE-PROCESS-ERROR-NEW: we now strip off any "."'s which appear at the end of qfile error strings, as some of this function's callers expect to be able to append text to the end of this returned string in constructing longer error messages. 125.19 pld: The NETWORK package doesn't really want to import/export the implementation specific variable ZUNDERFLOW -- better it should use the macro WITHOUT-FLOATING-UNDERFLOW-TRAPS, which is implemented in terms of ZUNDERFLOW on the Lambda, but need not be on other machines... Patches for Tape (Current version is 24.1): 24.1 keith: Improvements to interface of PARTITION-SEARCHER. 1. When it drags someone into the command loop against their will (maybe they mis-typed an argument), give them a message that orients them. 2. Print help note on first entering command loop. Same after every 8 bogus keystrokes (they're obviously confused). 3. Print more complete info on current selections. 4. If they have sufficiently specified a partition to select, give them a message telling them how to exit. T (dribble)