- Jul 02, 2004
-
-
rtoy authored
want to do that.
-
- Jun 29, 2004
-
-
rtoy authored
This causes (let ((a t) (b t) (c t) (d t) (e t) (f t)) (setf (values a (values b c) (values d) (values e f)) (values 0 1 2 3 4 5 6)) (list a b c d e f)) to return the (0 1 2 3 4 5) instead of (0 1 nil 2 3 nil).
-
rtoy authored
values. BIGNUM-LOAD-BYTE is significantly faster than LDB when extracting out 32-bit sized pieces. o Initial start at fixing BIGNUM-DEPOSIT-BYTE. Still broken from certain combinations of signed NEW-BYTE and BIGNUM and various border cases for BYTE-SPEC. Many debug prints still in the code.
-
- Jun 21, 2004
-
-
rtoy authored
From Eric Marsden.
-
- Jun 20, 2004
-
-
pmai authored
The old defmodule handler and the module:*-library.* loading have been factored out to their own module-provider functions, which are by default on *module-provider-functions*. Note however that this implies one functional change for defmodule defined modules: The loading of the files constituting such a module are now _NOT_ wrapped with without-package-locks. Wrapping of module:*-library.* files with without-package-locks is only retained temporarily, until proper changes to the various modules have been made. Also adjusted documentation of require to correctly state the default for *require-verbose* which has always been t, not nil as previously claimed. This change is still missing documentation in the user manual.
-
pmai authored
ensure that its effects kick in early enough during startup, thus silencing loading of siteinit, siteinit-loaded files, etc. Reported by JBThiel against the OS X port.
-
- Jun 18, 2004
-
-
rtoy authored
o Update %unary-ftruncate and the corresponding deftransform because it was not properly returning the correct sign for signed zeroes. If the number was in a good range, we used truncate, which loses the sign of zero. (Should we have leave it in and add an explicit check for a zero result and adjust the sign appropriately? That would allow us to use the fast builtin instructions at the expense of a test for zero and a fix.)
-
rtoy authored
cause garbage to be retained with gencgc on x86 (but not sparc), eventually causing an out-of-heap error. From Helmut Eller, via cmucl-imp.
-
- Jun 13, 2004
-
-
emarsden authored
Fixes to REINITIALIZE-GLOBAL-TABLE: ignore invalid entries in the *GLOBAL-TABLE*, and reload files in the same order as they were initially loaded. From Lynn Quamm.
-
emarsden authored
Avoid a hang when calling SOFTWARE-VERSION on Linux kernel version 2.6.x. The hang is due to a bug in certain files in the proc filesystem, where the select() system call does not work correctly.
-
- Jun 10, 2004
-
-
cwang authored
-
- Jun 09, 2004
-
-
rtoy authored
faster bignum reader. Contributed by Mark Wooding on cmucl-imp, May 24, 2004.
-
rtoy authored
they are arithmetically equal. Do this by adding 0.0 to the number, which does nothing to the number, except convert -0.0 to 0.0.
-
rtoy authored
it return 1 (of the right type) in these cases. We still produce a FP trap, if enabled, though. Is that right?
-
rtoy authored
returned appropriately instead of +0.0. This is to conform to IEEE754.
-
- Jun 04, 2004
-
-
rtoy authored
-
- Jun 02, 2004
- Jun 01, 2004
- May 24, 2004
- May 18, 2004
-
-
rtoy authored
but the following gives wrong results: (macroexpand '(multiple-value-bind (&rest x) (foo y) d)) (MULTIPLE-VALUE-CALL #'(LAMBDA (&OPTIONAL &REST X &REST #:G858) (DECLARE (IGNORE #:G858)) D) (FOO Y)) Fix from Juan Jose Garcia Ripoll.
-
rtoy authored
Upcase path components when given a logical pathname. Preserves print/read consistency and fixes (I think) an issue with ASDF and logical pathnames
-
rtoy authored
-
rtoy authored
-
- May 17, 2004
- May 15, 2004
-
-
rtoy authored
instead of a random value, just like on sparc.
-
- May 14, 2004
-
-
rtoy authored
itself. Only for sparc currently. Doesn't lazily compute the symbol hash yet. Simple test shows a 5% increase in compilation speed, despite making make-symbol significantly slower. * src/code/hash-new.lisp (internal-sxhash): Use the symbol-hash slot instead of computing the hash value. * src/compiler/generic/new-genesis.lisp (allocate-symbol): Write out the sxhash value of the symbol into the symbol-hash slot. * src/compiler/globaldb.lisp (info-hash): Update to use the symbol hash instead of computing the sxhash. * src/code/symbol.lisp (make-symbol): Compute the symbol hash when creating the symbol. * src/compiler/sparc/cell.lisp ((symbol-hash)): Add vop to extract out the symbol hash from a symbol. * src/compiler/generic/objdef.lisp: Rename the unused slot to hash, so we can make it the symbol hash.
-
- May 12, 2004
-
-
rtoy authored
comment for full details, but the issue is that Lisp says mixing a real and a complex requires converting the real to complex before doing the operation. But Kahan's algorithms assume that this doesn't happen, like z-1 should not be computed as z-(1+0*i). One place where this was wrong was for acos(2 +/- 0i). Kahan says acos(2+0i) is +0 - i*acosh(2) and acos(2-0i) is +0 + i*acosh(2). We had this backwards for the above reason. I think this was caused by the erroneous deftransforms for real op complex which were removed sometime ago, causing these function to compute the wrong thing.
-
- May 06, 2004
-
-
rtoy authored
source location for defvar and friends.
-
rtoy authored
the following enters an infinite recursion if it gets compiled two times: (def-alien-type nil (struct foo (f (* (function (values) (* (struct foo))))))) During the second compile, the existing type is compared with the to be defined type. This enters a recursion because (struct foo) is recursive and the comparison is done again. The current code has a "depth" counter and simply returns true if it exceeds 10. But it only works for pointers to records and loops forever for function pointers. The patch below should fix this. It keeps a hashtable of the already compared types and returns true if a pair is already in the hashtable.
-
- May 05, 2004
-
-
rtoy authored
function name for the dispatch function, but CMUCL only accepted a function. Allow either a function or symbol for the dispatch function, and change the declared type appropriately in the pprint-dispatch-entry structure slot.
-
- May 04, 2004
-
-
rtoy authored
neatly. (Currently only for the time report, not space report.)
-
- Apr 28, 2004
-
-
emarsden authored
On Linux/AMD64, we need to tell the platform linker to use the 32-bit linking mode instead of the default 64-bit mode. This can be done either via the LDEMULATION environment variable, or via the "-m" command-line option. - in EXT:LOAD-FOREIGN, add LDEMULATION=elf_i386 to the environment inherited by the linker This assumes that the LDEMULATION environment variable will be ignored by the platform linker on Linux/i386 platforms.
-
- Apr 23, 2004
-
-
rtoy authored
sizes up to MAX-STREAM-ELEMENT-SIZE (currently 1024). Ported from SBCL, with minor changes.
-
emarsden authored
- add a non-blocking version of CONNECT-TO-INET-SOCKET, which avoids having CMUCL blocked in in the connect() system call, allowing other threads to run. The new function is called CONNECT-TO-INET-SOCKET/NON-BLOCKING, and CMUCL still uses the old function internally. From Michael Hannemann.
-
rtoy authored
with FLONUM-TO-DIGITS someday?)
-