diff options
author | Karl Williamson <khw@cpan.org> | 2021-06-30 13:06:50 -0600 |
---|---|---|
committer | Karl Williamson <khw@cpan.org> | 2021-07-25 10:15:29 -0600 |
commit | fc4ef3c83cb2080bed88ceb3d8dc349ac321f012 (patch) | |
tree | 887ff078617549eedab14558fa4b4b0d4eaf972b /perl.h | |
parent | 4399381d2fd192c7677b390040305777edee6254 (diff) | |
download | perl-fc4ef3c83cb2080bed88ceb3d8dc349ac321f012.tar.gz |
perl.h: Add some comments
Diffstat (limited to 'perl.h')
-rw-r--r-- | perl.h | 10 |
1 files changed, 6 insertions, 4 deletions
@@ -5929,10 +5929,12 @@ static U8 utf8d_C9[] = { * evident in the table. The class numbers for start bytes are constrained so * that they can be used as a shift count for masking off the leading one bits. * It would make the code simpler if start byte FF could also be handled, but - * doing so would mean adding nodes for each of continuation bytes 6-12 - * remaining, and two more nodes for overlong detection (a total of 9), and - * there is room only for 4 more nodes unless we make the array U16 instead of - * U8. + * doing so would mean adding two more classes (one from splitting 80 from 81, + * and one for FF), and nodes for each of 6 new continuation bytes. The + * current table has 436 entries; the new one would require 140 more = 576 (2 + * additional classes for each of the 10 existing nodes, and 20 for each of 6 + * new nodes. The array would have to be made U16 instead of U8, not worth it + * for this rarely encountered case * * The classes are * 00-7F 0 |