Re: [PATCH] Ghost specifiers
Some time ago, William M. Perry wrote...
|+
| "Kirill M. Katsnelson" <kkm(a)kis.ru> writes:
|
| > So, IMO the correct behavior in your sample would be that aset signals;
| > although if you write (setq nbar (copy-sequence (symbol-name 'bar))) it
| > is ok, as it mangles with a copy of the string.
|
| This is totally bogus behaviour. We should be returning a duplicate copy
| of the string to begin with. Unless we make this do the 'right thing' and
| reintern 'bar' as 'xar', which would be an even bigger crock.
|-
This is clearer, but might be too expensive. Actually, this is what is done in
other places (so we deal with a bug here) like Fspecifier-fallback, and this
is the very thing Ben is going to change.
A related topic: should not copy-tree copy strings and bit vectors as well?
Kirill