To whom it may concern;

  Spent some time chasing a red herring around the'get_suffix' function. 
The patch included aims to simplify the logic considerably.

There are a number of interesting points to note:

   Under normal situations (".gz") (could be modified by using -S) the 
predicate:
<code>
         if (z_len < suflen && strequ (z_suffix, *suf + suflen - z_len))
</code>

   Will never be true. Modifications to the suffix string, possibly at 
build time,
or on the command line will trigger the boolean to get set, executing:
<code>
     known_suffixes[suffix_of_builtin
                    ? sizeof known_suffixes / sizeof *known_suffixes - 2
                    : 0] = z_lower;
</code>

   This will drop the duplicated pointer into the next-to-last element of 
the array,
rather than the front. The const qualifier on  known_suffixes on reading 
that it is const but this is not the case.

    strlen is a size_t, there may potentially be sign extension issues. 
It could be possible to craft a pointer as an argument to a mapped 
memory region if it was large enough to trigger a sign extension error, 
however it's extremely unlikely.

Take care,
John SETH Thielemann
JOSTALY Technologies
https://www.jostaly.com
223-231-3511