hi!
for reference see this UCW common pitfall note:
Consider this code in a render method:
(iter (for newval in '("a" "b" "c"))
(let ((newval-1 newval))
(<ucw:a :action (setf (value test) newval-1)
(<:as-html newval))))
Seems like it should work fine, the variable is even rebound, but in fact it
won't work due to something that can be considered a bug, but a non-trivial to
fix one.
segv it's iterates 'fault'. iterate walks and macroexpands
its body but
doesn't modify the lexical environment while it does
that, so it causes with-call/cc
to get expanded with the iter's env, and not the
iter's env + the let binding.
attila_lendvai ehh
segv (fwiw arnesi's own code walker doesn't do that either)
segv this explains why it works fine with loop and dolist
segv and dolist*
attila_lendvai is it easy to fix? i can push to iterate, but have no
experience with envs
segv attila_lendvai: i requires looking through the various
lisps to figure out how to
do that, and fixing iterate's code-walker to do the
proper environment augmentation
for let, let*, flet, labels, macrolet and symbol-macrolet forms.
segv doable, but not trivial.
segv at that point it'd probbaly be easier to rework
iterate to use arnesi's code walker
(which would be easier to fix since some of the non
portable environment stuff is already there)
attila_lendvai hm, doesn't sound like something i would chop my axe in... :)
attila_lendvai iterate is non-trivial in itself
segv yup.
Workarounds:
- use loop or dolist until iterate is fixed.
- in the ajax branch you can disable with-call/cc wrapping (which is
not needed here):
(iter (for newval in '("a" "b" "c"))
(let ((newval newval))
(<ucw:a :action (register-action () ;; :with-call/cc is #f
by default
(setf (value self) newval))
(<:ah newval))))
--
- attila
"- The truth is that I've been too considerate, and so became
unintentionally cruel...
- I understand.
- No, you don't understand! We don't speak the same language!"
Ingmar Bergman - Smultronstället (Wild Strawberries)
