Update runtime files

runtime/spell/sr/README_sr.txt

@@ -40,6 +40,20 @@ continues monotonically after the last flag number in cyrillic .aff file.
 A couple of words in cyrillic dict used a latin codepoints for 'a' and 'e',
 that was also corrected.
 
+You should be able to reproduce this with these steps:
+  * Save the existing sr.aff and sr.dic files, if you have them, they will be
+    overwritten.
+  * Create a subfolder "new".
+  * Put 4 files downloaded from LibreOffice dictionaries GitHub repository in
+    it: sr.aff, sr-Latn.aff, sr.dic and sr-Latn.dic
+  * Open Vim and cd into "new"
+  * Execute:  :so ../convert.vim
+  * The resulting sr.aff and sr.dic are created in the parent spell folder
+    (here).
+  * Now one can generate spl file as usual using the merged dic and aff
+    files:
+        env LANG=sr_RS.UTF-8 vim -u NONE -e -c "set enc=utf-8" -c "mkspell! ../sr sr" -c q
+
 
 Ivan Pešić
 28.06.2022.

runtime/spell/sr/convert.vim

@@ -0,0 +1,30 @@
+:e sr.aff
+:normal gg
+:normal wgu$
+:3d
+:4d
+:normal G
+:normal o
+:r sr-Latn.aff
+:%s#^\(SFX\|PFX\).*[а-џa-ž]\zs$# .#g
+:normal G
+?SET
+:.,+5d
+:.,$s#^\(SFX\|PFX\) \zs\(\d\+\)#\= eval(submatch(2) .. ' + 1903')#
+:w ../sr.aff
+:bd!
+:e sr.dic
+:%s#a#а#g
+:%s#e#е#g
+:normal G
+:normal o
+:r sr-Latn.dic
+:normal 201dd
+:.,$s#/\zs\(\d\+\)\(,\(\d\+\)\)\?$#\=(submatch(2) == '') ? eval(submatch(1) + '1903') : eval(submatch(1) + '1903') .. ',' .. eval(submatch(3) + '1903')#
+:normal {
+:normal dd
+:normal gg
+:normal C502898
+:w ../sr.dic
+:bd!
+:q!