c1make-global-variable -> make-global-var (and move to cmpvar) -- this function
does not create c1form so this name was wrong
cmp-env-declare-special -> declare-special (and move to cmpenv-var) -- this
function does not only declare special but it also creates an instance of a
variable - move that to a new file cmpenv-var.lsp
Use proper package accessors instead. This was mostly already done. Removal of
package imports make it easier to tell when symbols do not belong to cmp.
It is worth noting that *active-protection* and *pending-actions* are generally
not used too (because *pending-actions* are never modified), but if we want to
add some useful semantics to with-compilation-unit one day then we'll need both.
For empty structs, the size was off by one. While this did not affect
structs that were created directly with the structure constructor, for
structs that were externalized with MAKE-LOAD-FORM-SAVING-SLOTS
instances which were one element too large were created.
The pattern in the long form of the define-method-combination may contain * as
a list element meaning "any" qualifier. For example:
(define-method-combination foo ()
((bar (:xxx * :yyy)))
...)
In this case qualifiers `:xxx 3 :yyy' and `:xxx :zzz :yyy' will match.
The comment mentioned that aux variables (the sixth value) are returned the
same way as requireds, optionals and keywords however factually that was not
the case - the number of variables was not the first element of the list. This
commit updates the function and all its callers.
Previously for example (prin1 1.0e-8) would result in 1.e-8. However,
the ANSI standard (see CLHS section 22.1.3.1.3) specifies that there
has to be at least one digit after the decimal point therefore we now
print 1.0e-8.
Since our ancient LGPL licensed copy of libgmp only knows about
PowerPC apple devices, build it using the fallback plain C
implementation as on x86.
Closes#689.
Autoconf detects that the system function is present but compilation
fails with
error: 'system' is unavailable: not available on iOS
Therefore we manually disable this function for iOS.
Closes#691.
The problem encountered in 3f64e2e88b
affects not only the case of a logarithm where one argument is a
rational and the other a long or double float, but also cases where
both arguments are floating point numbers of different lengths.
For (log x y) where one of the two arguments is a double or long float
and the other a rational number, defining (log x y) as (/ (log y) (log
x)) is imprecise since intermediate single float results will be used
for the rational argument. Prevent this by computing the logarithm of
the rational to the same precision as that of the other argument.
Fixes#688.
