Bug#824840: libhunspell api change after upgrade to 1.4

[Rene Engelhard <rene@debian.org>]

forwarded 824840 https://github.com/hunspell/hunspell/issues/387
retitle 824840 get_wordchars_utf16 documentation wrong in hunspell(3) after API change
found 824840 1.4.0~git20160415-1
thanks

Hi,

On Fri, May 20, 2016 at 12:13:41PM +0200, Jeroen Ooms wrote:
>   const std::vector<w_char>& get_wordchars_utf16() const;
> 
> However 'man 3 husnpell' still lists the old API:
> 
>   unsigned short * get_wordchars_utf16(int *" len)

That is a bug, yeah, if it wasn't a accidential change.

AFAICS that was done in

commit 6f9e403ab758c9db9ddcebe86f237e336f4a5624
Author: Caolán McNamara <caolanm@redhat.com>
Date:   Thu Apr 14 17:26:28 2016 +0100

    convert ignorechars_utf16 to vector

which did

@@ -3680,9 +3675,8 @@ const char* AffixMgr::get_wordchars() const {
   return wordchars;
 }
 
-const w_char* AffixMgr::get_wordchars_utf16(int* len) const {
-  *len = wordchars_utf16.size();
-  return &wordchars_utf16[0];
+const std::vector<w_char>& AffixMgr::get_wordchars_utf16() const {
+  return wordchars_utf16;
 }
 
 // is there compounding?

and

-  const w_char* get_wordchars_utf16(int* len) const;
+  const std::vector<w_char>& get_wordchars_utf16() const;

so it probably was intended.

> The API change was not advertised elsewhere either. It is not clear if
> the change was intended, but it breaks anything that uses this
> function.

Well, the SONAME didn't change for no reason.

If the API changed with intention so be it, and as said, the SONAME changed.

> Upstream issue: https://github.com/hunspell/hunspell/issues/387

Marking as forwarded.

Regards,

Rene