This sounds like a good suggestion to me.
Ingo Strüwing wrote:
> Hi Chuck, Jørgen,
>
> Chuck Bell, 28.04.2009 19:07:
> ...
>> COMMENTARY
>> ----------
>> Very well done. :)
>
> Jørgen Løland, 29.04.2009 11:13:
> ...
>> COMMENT:
>> --------
>> Very nicely written patch. Well done!
>
> Thank you for the worm words. :-)
>
> Some time during my vacation, after I pushed the patch, I got aware that
> my solution might not be optimal:
>
> The current patch does not clean up ../ from paths if the preceding path
> element is a symbolic link. This is important and correct, but perhaps
> not sufficient.
>
> The clean up of ../ is also done if the previous path element references
> a nonexistent file system object. This is acceptable only as long as the
> object remains non-existent. However, if it becomes a plain file, or
> some other non-directory object, it might have been replaced by later
> path elements and the resulting path would become invalid. For example,
> assume this path:
>
> dir/non-existent-file.txt/../existent-file.txt
>
> The resulting path would be dir/existent-file.txt, which does not match
> what standard file operations would do. ../ is a valid path element only
> if the preceding element is an existing directory.
>
> So I ask you, if you agree that I change safe_cleanup_cat_path() so that
> it does not clean up ../ if the preceding path element is not a real
> (non-symlink) directory?
>
> Regards
> Ingo
--
Jørgen Løland
