Improved check-localedef script

[Zack Weinberg]

Here is an improved version of the check-localedef script I posted the
other week.  It now takes only about 1.5 seconds to process all the
files in localedata/locales/ (instead of seven seconds with the old
parser), which is fast enough that I think it would be reasonable to
run it during 'make check'.  Also, many bugs have been fixed.
Especially, the "can we encode this string in the charset that the
file is annotated with" test now actually _runs_...

... and finds dozens and dozens of errors. The full list is attached,
but here's a small sample:

localedata/locales/ur_PK... (charset: cp1256)
  localedata/locales/ur_PK:114: string not representable in cp1256:
      062C 0646 0648 0631 06CC
  localedata/locales/ur_PK:115: string not representable in cp1256:
      0641 0631 0648 0631 06CC
  localedata/locales/ur_PK:117: string not representable in cp1256:
      0627 067E 0631 06CC 0644

These are the abmon strings, so I think it really would be a problem...

zw
-------------- next part --------------
A non-text attachment was scrubbed...
Name: check-localedef.py
Type: text/x-python
Size: 16166 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20170803/84654941/attachment.py>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: check-localedef.errs
Type: application/octet-stream
Size: 52061 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20170803/84654941/attachment.obj>