- Oct 18, 2015
-
-
Raymond Toy authored
This allows us to build all of cmucl, but the utilities don't yet build completely.
-
Raymond Toy authored
In particular, use the 64-bit versions of most functions.
-
Raymond Toy authored
WIP; many items have been moved, but not all, and not all things have been checked. This current code doesn't succeed in building itself. The second build crashes with a type error coming from unexpected-eof-error.
-
- Oct 17, 2015
-
-
Raymond Toy authored
Covers 1, 2, and 4-bit arrays. Manually verified that the cmucl 21a fails these tests, as expected, when the index is a multiple of the number of elements per 32-bit word.
-
Raymond Toy authored
For simple-arrays of 1, 2, or 4-bit elements, data-vector-set-c was incorrectly merging the new value into the array when the index is a multiple of the number of elements per (32-bit) word. Thus, for 4-bit elements, the new value was not merged in when the index is a multiple of 8. In these cases, there's no need to shift the array value or the new value to move them into the correct place. When the shift is zero, the code accidentally removed the part that merges in the new value. Fix #10.
-
- Oct 11, 2015
-
-
Raymond Toy authored
when printing random bytes and words.
-
Raymond Toy authored
The break inst is really the int3 instruction, so disassemble it as int3.
-
Raymond Toy authored
Remove old comment, and remove the min OSX version on ppc too.
-
Raymond Toy authored
MIN_VER requires the 10.5 SDK be available. Add -Wl,-no_pie to turn off warning (and add comment).
-
Raymond Toy authored
-
Raymond Toy authored
-
Raymond Toy authored
When making an executable on some random machine, we can't depend on the user having the 10.5 SDK around. So remove the flag, but add -Wl,-no_pie to suppress the warning about PIE disabled due to absolute addressing.
-
Raymond Toy authored
Left this out in the previous commit.
-
Raymond Toy authored
-
- Oct 10, 2015
-
-
Raymond Toy authored
Noticed by Joram Schrijver who gave the correct path.
-
Raymond Toy authored
It's really nice to see the exit code of the process when printing out the process structure. Fix a typo too: "tings" -> "things".
-
Raymond Toy authored
obj_run_linker() returns the return code from the call to system(). Exit lisp with this return code to indicate if running the executable linker script worked or not.
-
- Oct 05, 2015
-
-
Raymond Toy authored
* src/lisp/GNUMakefile * Don't add exec-final.o to lisp.a. (But still remove exec-init.o from it.) * src/tools/linker.sh * When linking the executable, need to link exec-final.o too. * bin/make-main-dist.sh * Need to install exec-init.o and exec-final.o. exec-init.o is needed to link a normal lisp; exec-final.o is used for the executable image. * Fix modes on some installed files. exec-init.o, exec-final.o, and lisp.a don't need to be executable.
-
- Oct 01, 2015
-
-
Raymond Toy authored
-
Raymond Toy authored
-
Raymond Toy authored
Add bootfile for building cmucl with the new version too.
-
- Sep 01, 2015
-
-
Raymond Toy authored
The previous implementation cleared *herald-items* when -quiet was given, but if the init file loaded up items that added to *herald-items*, then the items were still printed. Plus, it seems odd that -quiet actually clears *herald-items*. So, just don't print herald if -quiet is given, but also preserve *herald-items* so that (print-herald) would actually print the actual herald.
-
- Aug 30, 2015
-
-
Raymond Toy authored
-
Raymond Toy authored
-
Raymond Toy authored
-
Raymond Toy authored
Instead of using random integers in the switch statement, use the fdlibm enum values. Also fix a compiler warning that ret might be used uninitialized. Fix this by adding a default case, which should never happen.
-
Raymond Toy authored
Gets rid of compiler warning too.
-
Raymond Toy authored
-
Raymond Toy authored
These are needed for asdf and slime, respectively. These were updated for unix.lisp, but not unix-glibc2.lisp. (Time to merge them into one!!!!)
- Aug 27, 2015
-
-
Raymond Toy authored
Apply patch from Helmut, as is.
-
- Aug 01, 2015
-
-
Raymond Toy authored
o Add declaration for LIST-ELT* that the index is a kerrnel:index. o Clean up ELT to directly call LIST-ELT* instead of having an inlined version. o Fix typo: issue.5 is really issue.4. o Add tests for invalid indices for ELT and (SETF ELT) for both lists and vectors.
-
- Jul 26, 2015
-
-
Raymond Toy authored
Then when the pcl tests are compiled, the fasls are stored somewhere else instead of in the tests/pcl directory. This means the src directory isn't polluted with fasls when we make a source tarball.
-
- Jul 22, 2015
-
-
Raymond Toy authored
-
- Jul 16, 2015
-
-
Raymond Toy authored
From Fred Gilham.
- Jun 24, 2015
-
-
Raymond Toy authored
This requires using boot-2015-06-1 to make the change. Regenerated cmucl.pot too.
-
Raymond Toy authored
Fix #7 The change to use extended names like (flet frob) broke the detection of local names for get-setf-expansion. Fix it. This reverts behavior back to what 18a used to do. A test for this is also added.
-
- Jun 22, 2015
-
-
Raymond Toy authored
I always forget how to run one set of tests. Add some hints.
-
Raymond Toy authored
Fix #3. The lookup for the names was not handling things like (FLET SQUARE TEST/PRESENT) correctly. Use VALID-FUNCTION-NAME to get the function name instead of a plain EQUAL test.
-