Re: [bug-gnu-libiconv] iconv fails converting 0xCC in EBCDIC-AT-DE charset.

[Bruno Haible]

Roland Tapken wrote on 2014-11-04:
> while converting RPG scripts from an IBM's iSeries I found a converting
> error.
> 
> See the following Base64-encoded line from an RPG script with EBCDIC 1141
> charset:
> 
> QEBAQEBAXMSBo6SUQEBAQEBAzEDBpKOWmUBAQEBAQEBAQEBAzE
> DChaKDiJmFiYKklYdAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBA
> QEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQA==
> 
> In ISO-8859-1 this reads:
> 
>       *Datum      ¦ Autor           ¦ Beschreibung
> 
> But iconv (Version "(Ubuntu EGLIBC 2.19-0ubuntu6.3) 2.19") fails to convert
> ¦ from EBCDIC 1141 (0xCC) to ISO-8859-1 (0xA6):
> 
> $ echo QEBAQEBAXMSBo6SUQEBAQEBAzEDBpKOWmUBAQEBAQEBAQEBAzE\
> DChaKDiJmFiYKklYdAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBA\
> QEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQA== | base64 -d |
> iconv -f EBCDIC-AT-DE -t ISO-8859-1
>       *Datum      iconv: illegal input sequence at position 18

According to the iconv version you gave, you are using iconv from GNU libc,
not from GNU libiconv. Therefore your bug report should be directed to
https://sourceware.org/bugzilla/ (pick component 'glibc').

It could be that glibc is wrong in its EBCDIC-AT-DE implementation.
Or that the Wikipedia section 
http://de.wikipedia.org/wiki/Extended_Binary_Coded_Decimals_Interchange_Code#EBCDIC_1141
is wrong.
Or that EBCDIC-1141 and EBCDIC-AT-DE are in fact different encodings.

Bruno