- Nov 30, 2014
-
-
Raymond Toy authored
The tests still fail (along with others), but the test suite will finish. * code/float-trap.lisp: * Update SET-FLOATING-POINT-MODES: * When we clear out bits in float-invalid-op-1-byte, clear out the float-invalid-op-2-byte. Not necessary for this fix, but those bits signal other invalid operations from sqrt and such. * When clearing the exceptions, clear the sticky exceptions summary bit too. * Fix SIGFPE-HANDLER: * As above, clear out the bit in float-invalid-op-2-byte and the exceptions summary bit. * Only clear out the new exception bits, as done for SSE2. * Mask out the float-invalid-op-2-byte and the summary bit before setting the new modes. * compiler/ppc/parms.lisp * Define float-exceptions-summary-byte. * code/exports.lisp: * Export FLOAT-INVALID-OP-2-BYTE
-
- Nov 29, 2014
-
-
Raymond Toy authored
(floating-point status and control register).
-
- Nov 28, 2014
-
-
Raymond Toy authored
Fixes ticket:95, along with changeset:[2b39eb]
-
Raymond Toy authored
The check for invalid gf lambda lists happens during macroexpansion, so update the tests to catch this during macroexpansion. This fixes three of the issues in ticket:95.
-
- Nov 27, 2014
-
-
Raymond Toy authored
* src/code/irrat.lisp: * In log10, return the correct type for the result. Previously, we would return a double even if a single were desired. * Fix up a few more cases where we failed to handle log of a huge rational correctly for log base 2 and base 10. * tests/irrat.lisp: * Add some tests to check that log returns the correct type of number for the log base 2 and 10 of very large rationals.
-
- Nov 26, 2014
-
-
Raymond Toy authored
* src/code/irrat.lisp: * Add log10 function to handle the computation of the base 10 log of big rationals that might not fit in a double-float. * Fix some issues where CL:LOG wasn't handling logs of big rationals. (A regression). * tests/trac.lisp: * Update trac.8 test to include logs base 10.
-
Raymond Toy authored
* code/exports.lisp: * Export %LOG2. * code/irrat.lisp: * Define %log2 * Use %log2 instead of log2. (This needs work) * compiler/float-tran.lisp: * Use %log2 instead of log2 in the deftransforms.
-
Raymond Toy authored
* log2.c: * New file containing the parts of e_pow.c that implemented a log2 function for use in pow(). * GNUmakefile: * Compile log2.c as part of the build.
-
- Nov 25, 2014
-
-
Raymond Toy authored
* Fix the log10.result-types test because we return correctly rounded results for these few tests.
-
Raymond Toy authored
general case.
-
Raymond Toy authored
-
- Nov 24, 2014
-
-
Raymond Toy authored
Since log2 and log10 use basically the same natural log implementation, factor that out the common part into its own routine.
-
Raymond Toy authored
6.64. Tests were already added to tests/irrat.lisp.
-
Raymond Toy authored
-
Raymond Toy authored
-
Raymond Toy authored
double-double. * src/code/irrat.lisp: * For the following two cases, cmucl generated an error instead of computing the log * Base is a double-double but number is not. * Number is a double-double but base is not. * tests/irrat.lisp: * Add some tests for log2 and log10.
-
- Nov 22, 2014
-
-
Raymond Toy authored
-
- Nov 15, 2014
-
-
Raymond Toy authored
%enumerate-directories. This makes this part of the function the same as the version from 18a. Don't see any real reason why stat was required anyway.
-
- Nov 13, 2014
-
-
Raymond Toy authored
-
- Nov 12, 2014
-
-
Raymond Toy authored
-
Raymond Toy authored
* Add :elf to *features* so extern-alien-name does the right thing for solaris/x86 * Add char-bytes and lowtag-bits to the info env hacks. (Otherwise, cross compile fails because char-bytes is undefined.)
-
- Nov 08, 2014
-
-
Raymond Toy authored
* Config.solaris: * Add appropriate value for DEPEND_FLAGS
-
Raymond Toy authored
* GNUmakefile: * Auto-generate dependencies using same approach as for lisp. * Config.Darwin: * Config.NetBSD: * Config.solaris_sunc: * Config.x86: * Add appropriate value for DEPEND_FLAGS
-
Raymond Toy authored
Now that motifd can run as a 64-bit binary again, build the 64-bit binary on linux. Clean up list of LIBS as well. Simple tests show this works on Linux.
-
- Nov 07, 2014
-
-
Raymond Toy authored
codepoint and glyph iterators.
-
- Nov 06, 2014
-
-
Raymond Toy authored
* src/code/string.lisp: * Add WITH-STRING-CODEPOINT-ITERATOR to allow iterating over the codepoints in a string. * Add WITH-STRING-GLYPO-ITERATOR to allow iterating over the glpyhs in a string. * src/code/exports.lisp: * Export WITH-STRING-CODEPOINT-ITERATOR and WITH-STRING-GLYPH-ITERATOR. * src/code/loop.lisp: * Support (loop for cp being the codepoint of string ...) which extracts the consecutive codepoints from the string. Allow codepoints, code-point, and code-points as aliases of codepoint. * Support (loop for g-string being the glyph of string ...) which extracts each glyph (as a string) from the string. Allow glpyhs as an alias. * tests/extended-loop.lisp: * New file of tests for the new loop paths. * src/i18n/locale/cmucl.pot: * Update.
-
Raymond Toy authored
-
Raymond Toy authored
This shows up when creating executables. Executables don't have the magic number in the core file to indicate whether the core suports sse2 or x87, so the fpu_type is never set to the actual type, and fpu_mode is set to the value of fpu_type. But there are asserts in the C code to check that fpu_mode is SSE2 since we have dropped support for x87. One case where this happens is building maxima with a cmucl executable on darwin. An assert triggers preventing maxima from running. This change fixes that.
-
- Nov 03, 2014
-
- Nov 02, 2014
-
-
Raymond Toy authored
-
Raymond Toy authored
It's not used anymore; ansi-loop.lisp is CMUCL's loop implementation.
-
Raymond Toy authored
Apparently this is from old-loop.lisp, which isn't used anymore.
-
Raymond Toy authored
-
- Nov 01, 2014
-
-
Raymond Toy authored
* src/pcl/simple-streams/external-formats/aliases: * Remove the :utf alias. * src/general-info/release-21a.txt: * Update.
-
- Oct 30, 2014
-
-
Raymond Toy authored
it's not true.
-
- Oct 29, 2014
-
-
Raymond Toy authored
Fixes from Robert Swindells.
-
- Oct 27, 2014
-
-
Raymond Toy authored
-
Raymond Toy authored
RXmStringEtLtoR. Fix from Richard Kreuter. This makes his clm-young::chooseone demo work reliably. Richard writes: The detail that this patch affects is non-deterministic: RXmStringGetLtoR in xmstring.c calls message_write_boolean on a pointer, message_write_boolean in datatrans.c calls combine_type_and_data with the pointer and the boolean type tag, and combine_type_and_data in datatrans.h IORs the boolean type tag with the bits 25 to 32 from the start of the pointer. This gave a decoding error on the Lisp side in TOOLKIT-READ-VALUE, because the high 8 bits of the 32 bits being decoded indexed past the end of *TYPE-TABLE*. (But other incorrect outcomes are possible, e.g., the result of combine_type_and_data could index a valid, but wrong, type code, etc.)
-
- Oct 25, 2014
-
-
Raymond Toy authored
Fixes from Robert Swindells.
-
Raymond Toy authored
Merge fix from Richard Kreuter's clm-text-callback branch. This fixes an issue where text callbacks get nil. The test case is clm-young::numeric from Richard Kreuter. Without this patch, running (clm-young::numeric) and entering digits and characters causes errors[1]. With this patch, digits can be entered, and non-digits are correctly rejected. [1] On linux. For whatever reason, this works on darwin/x86 even without this patch.
-