bug#27267: 25.1; ffap-file-at-point erroneously returns "" in some cases

bug#27267: 25.1; ffap-file-at-point erroneously returns "" in some cases

[Alex]

[-- Attachment #1: Type: text/plain, Size: 617 bytes --]

Execute ffap-file-at-point with the following buffer contents (| for
cursor position):

7:|

The result is "", which leads to ffap's default input being ~ (not ~/ as
expected).

Similarly, try ffap-file-at-point on:

:6|

This is due to ffap-file-at-point stripping off the colon and numbers,
and sending "" to ffap-file-exists-string. (file-exists-p "") returns t,
so ffap-file-at-point treats "" as an existent file.

The following diff fixes this, but I'm not sure if it's the best
approach. Perhaps ffap-file-at-point just checks that the argument to
ffap-file-exists-string is non-empty each time it's called.



[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: ffap --]
[-- Type: text/x-diff, Size: 445 bytes --]

diff --git a/lisp/ffap.el b/lisp/ffap.el
index 87531110b8..21312b34c3 100644
--- a/lisp/ffap.el
+++ b/lisp/ffap.el
@@ -514,6 +514,7 @@ ffap-file-exists-string
 The optional NOMODIFY argument suppresses the extra search."
   (cond
    ((not file) nil)			; quietly reject nil
+   ((string-empty-p file) nil)		; reject empty string
    ((file-exists-p file) file)		; try unmodified first
    ;; three reasons to suppress search:
    (nomodify nil)

bug#27267: 25.1; ffap-file-at-point erroneously returns "" in some cases

[Lars Ingebrigtsen]

Alex <agrambot@gmail.com> writes:

> Execute ffap-file-at-point with the following buffer contents (| for
> cursor position):
>
> 7:|
>
> The result is "", which leads to ffap's default input being ~ (not ~/ as
> expected).

I get

(ffap-file-at-point)
=> "/home/larsi/"

so this seems like it's fixed.

>
> Similarly, try ffap-file-at-point on:
>
> :6|
>
> This is due to ffap-file-at-point stripping off the colon and numbers,
> and sending "" to ffap-file-exists-string. (file-exists-p "") returns t,
> so ffap-file-at-point treats "" as an existent file.

I get nil from that in the Emacs trunk, which seems correct, so I guess
this has been fixed to, and I'm closing this bug report.  Please reopen
if this is still present on the trunk.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no