[Slime-devel] M-. and interning

[Gábor Melis]

Helmut Eller <eller.helmut at gmail.com> writes:

> On Tue, Jan 27 2015, Zach Beane wrote:
>
>> If I load a file like this:
>>
>>   (defpackage xach (:use cl))
>>   (in-package xach)
>>   (defun attack () 'attack)
> [...]
>> Why does ATTACK get interned in the current package?
>
> To find the position in the file, M-. skips over the first two forms
> with *read-suppress* = t.  But the last form is READ with
> *read-suppress* = nil and with the current package.  That causes the
> interning.
>
> For a while it was possible to read the last form with *read-suppress*
> as READ returned stuff like (nil (nil nil nil)) which was good enough to
> find the beginning positions of lists, but later the reader in SBCL
> changed and since then it only returns nil and we had to disable
> *read-suppress*.
>
> Instead of returning non-nil values, our source location recording
> wrappers could probably store things like (1 (2 3 4)) in a global
> variable.  Then we could again use *read-suppress*.
>
> However, it looks like the patch to better support reader-macros needs
> to bind *read-suppress* to nil while skipping over forms and will likely
> cause more unwanted interning.
>
> Helmut

Actually, the patch I'm cleaning up changes *PACKAGE* after reading the
IN-PACKAGE form, so it interns stuff in the XACH package where it's a
noop, so it fixes Xach's issue.