GitLab

  Fix inconsistency in handling illegal byte sequences.
  Convert surrogate codepoints and U+FFFE, U+FFFF to U+FFFD.
  Also, a few extra mappings.
Detail:
  enc_utf8.c: 0x80 is a continuation byte. Map stray ones to U+FFFD.
              Reset the count of expected continuation bytes to 0 when
              encountering illegal byte sequences. Previously, if the character
              callback returned non-zero, this count would not be reset, thus
              leaving the codec in an inconsistent state. Additionally, we no
              longer consume the illegal continuation byte: instead, we process
              it as a start byte next time round.
  encoding.c: Do not load extension tables for ISO-8859-{1,2,9,10,15,16}
              If these are needed, it's probably best that different charset
              names are used rather than overloading 8859-n.
  iso2022.c:  Permit SS2/3 escape sequences for EUC encode/decode.
              Disable C1 chara...

> Summary:
>   Merged changes from branch tree
>   Reversed previous change
> Detail:
>
> * Merged a few changes/fixes from the Unicode library in
>  branch's tree.
>
> * Reversed Steve's change from version 0.50. The change wasn't
>  necessary, and with the changed definition of NOT_USED in this
>  version, it compiles fine with cc 5.45.
>
> * Small comment change in unix.c. It now states that the file
>  isn't equivalent to any in the branch tree.
>
> Admin:
>   Built and briefly tested using TextConv utility on Risc PC.

Version 0.52. Tagged as 'Unicode-0_52'

Detail:
  Copyright messages changed from E-14 to Pace throughout, filename
   placed at top of file throughout, instead of in just some files.

  Merged branch's fixes into our code base, plus made it possible to
   get nice debug output in branched tree, and vfprintf() to stderr in
   RISC OS tree.  Exactly same source used in branched tree now (apart
   from OS specific files riscos.c and unix.c moving into layers
   directory structure).

Admin:
  Built for branched, both Unix and RISC OS.
  Built in RISC OS tree, and compiled into TextConv.


Version 0.45. Tagged as 'Unicode-0_45'

Changed encoding_table_remove_unused() so that it takes a parameter giving the depth from which to start purging.

Fixed ISO2022 write code to free search tables.
Added unix.c for unix-targeted builds.
Updated cross-compile build.
Added unix-targeted build of library and textconv tool
in ccsolaris directory.

Version 0.36. Tagged as 'Unicode-0_36'

Added encoding_set_flags().
Proper handling of byte order marks in UTF-16 and UCS-4.
Fixed UTF-16 surrogate writing.
Adjusted various MIME charset identifiers.
Incorporated latest Unicode Character Database (2.1.8).
Added "current system alphabet" encoding.
Created "TextConv" command line character set conversion utility.

Version 0.34. Tagged as 'Unicode-0_34'

Version 0.33. Not tagged

Reinstated use of data->data relocations.

Version 0.32. Not tagged

Version 0.01. Not tagged