Max Nikulin <manikulin@gmail.com> writes:

>> -		 (org-export-file-uri raw-path))
>> +                 (pcase (org-export-file-uri raw-path)
>> +                   (`(and (pred #'file-name-absolute-p) ,path)
>> +                    path)
>
> I am in doubts if this branch may be executed at all. For absolute file 
> names `org-export-file-uri' adds "file:" prefix so I expect that 
> `file-name-absolute-p' is always false.

You are indeed right.
See the attached new version of the patch.

> This time I am not lucky enough to accidentally face a case with invalid 
> link URI. However I do not think that for the link
>
>       [[file:file-b.org]]
>
> "../file-b.org" is appropriate default description. I do not expect 
> "../" here.

I also noticed this, but I feel that it is reasonable. In odt,
"file-b.org" implies a resource inside the odt archive. "../file-b.org"
will highlight that it is an external file.

I'd prefer to leave the "../file" description for now and only do
something when/if we hear from the ox-odt users.