In this case we have a very small data example to work from. I did put in a print so the OP can see the default sort order and determine whether or not this applies more generally to his/her problem or not. Version numbers often, but not always have leading zero's or a fixed number of digits.

In this case we have a very small data example to work from.

A frustratingly small data sample, and if this were $work, I (and I suspect you, too) would instantly be busy calling on the phone or composing an e-mail to the "client" to answer the question "Just what the heck do you really want here?"

... in the OP's case, the default sort order does work.

The legal and ethical obligations of work don't apply in this place, but I still think there is, so to speak, a "moral" obligation (appropriate to a monastery!) to offer to a supplicant defensive measures suitable to the question. After all, none of the example version numbers have leading zeros, and how did, e.g., 'abc-4.2.11-8.el7_2.x86_64' get where it is without first going through 'abc-4.2.9-8.el7_2.x86_64', and might it not yet go to 'abc-4.2.11-10.el7_2.x86_64'?

("I didn't get where I am today without going through 'abc-4.2.9-8.el7_2.x86_64'!" Old Reginald Perrin reference. Never mind...)

---

Give a man a fish:  <%-{-{-{-<

Ok, fair enough.

If it can be determined exactly what the "version" string is, then I perhaps would use an algorithm like below. One pass through the data, no sorting. Use a hash table with a key of the "generic name" and value of the version. If the version is greater, then replace the hash value with the greater version.

Of course this can become very complex in the real world because version naming conventions vary widely.

However, I think this was a successful thread in that the main issues have been discussed and some techniques demo'd. I did this quickly and perhaps there are some minor point to quibble about. But I think the general idea of this approach is solid.

#!/usr/bin/perl
use warnings;
use strict;

my %versions;

while (my $line = <DATA>)
{
   next if $line =~ /^\s*$/; #skip blank lines
   $line =~ s/\s*$//;
     
   my ($name, $version, $platform) = 
       $line =~ /^([a-zA-Z-]+)-([\d\._-]+)\.(.*)$/;
   
   # set a default value if needed.
   $versions{"$name:$platform"} //= "0.0.0-0";
      
   if (cmp_verA2B ($versions{"$name:$platform"}, $version)<0) 
   {                                                        
      $versions{"$name:$platform"} = $version;
   }
}

foreach my $version (sort keys %versions) #print results
{
   my $max_ver = $versions{$version};
   my ($prefix,$suffix) =split ':', $version;
   print "$prefix\-$max_ver\.$suffix\n";
}

sub cmp_verA2B
{
  my ($verA, $verB) = @_; 
  
  my ($majorA,$minorA,$subVerA) = split /\./,$verA;
  my ($majorB,$minorB,$subVerB) = split /\./,$verB;
  my ($subVerA1, $subVerA2) = split ('-',$subVerA);
  my ($subVerB1, $subVerB2) = split ('-',$subVerB);
  
     $majorA <=> $majorB
              or
     $minorA <=> $minorB
              or
     $subVerA1 <=> $subVerB1
               or
     $subVerA2 <=> $subVerB2
}

=Prints:
abc-4.2.11-8.el7_2.x86_64
samba-common-libs-4.2.12-7.el7_2.x86_64
xyz-libs-4.2.13-7.el7_2.x86_64
=cut


__DATA__
samba-common-libs-4.2.12-7.el7_2.x86_64
samba-common-libs-4.2.10-3.el7_2.x86_64
xyz-libs-4.2.10-7.el7_2.x86_64
samba-common-libs-4.2.10-6.el7_2.x86_64
abc-4.2.11-7.el7_2.x86_64
xyz-libs-4.2.11-7.el7_2.x86_64
abc-4.2.11-8.el7_2.x86_64
samba-common-libs-4.2.10-6.2.el7_2.x86_64
xyz-libs-4.2.12-7.el7_2.x86_64
xyz-libs-4.2.13-7.el7_2.x86_64
samba-common-libs-4.2.10-8.el7_2.x86_64
[download]