- 14 Aug, 2011 5 commits
-
-
-
-
-
-
in favor of using dispatch tables just like other parts of our compilation process (precompile, pass1, pass2).
-
- 13 Aug, 2011 7 commits
-
-
-
-
in closures. Code elimination! Yay!
-
-
also make sure all compiland children have known class names before processing the body of the compiland.
-
passing in the right values to the reading function.
-
reach the function from the compiland, but the other way around works.
-
- 12 Aug, 2011 9 commits
-
-
at the same time speeding up compilation.
-
-
-
instead of counting them, even though we only needed a HAS-CHILDREN boolean.
-
-
-
-
This commit groups all nested function objects resulting from a COMPILE call into one class loader (instead of a class loader each). Additionally, nested function objects aren't instantiated using reflection anymore, instead, the 'new' instruction is used, winning a factor 100 per local function.
-
-
- 11 Aug, 2011 12 commits
-
-
but about binding establishment (ie assigning a new value to the closure array's slots).
-
read from: don't copy it - ever.
-
(Unbreaks last commit.)
-
WrongNumberOfArguments program errors.
-
A bad regression presumably introduced r13533. Found by @acelent on #abcl (and possibly Theam Yong Chew but I didn't have time to properly respond to his last email).
-
-
-
Remove the wrong-headed use of TRUENAME at compile time which would error unless the test directory had previously been created. Remove the use of the readtime #. macro to create the success conditions for tests which were failing based on the presence of the temporary jar directory. These bugs seem show the need for a more explicit test build-up/tear-down sequence as the complexity in manage Lisp macros here leads to a lot of unncessary bugs.
-
-
-
-
well; this improves printing of "Wrong number of args" errors.
-
- 09 Aug, 2011 1 commit
-
-
Found by: Blake McBride Fixed by: me :-)
-
- 07 Aug, 2011 5 commits
-
-
for structures. In order to make sure they are pre-compiled in our build too, compile defstruct.lisp earlier in the compilation phase. (Saves roughly 20s on my compilation runs.)
-
"adding" the feature.
-
-
-
-
- 06 Aug, 2011 1 commit
-
-
Note: This commit also fixes some failures in the random testing ANSI tests, notably PRINT.RANDOM.SYMBOL.*.
-