Re: Configuring gdb_wchar.h

[Eli Zaretskii <eliz@gnu.org>]

> Cc: gdb-patches@sources.redhat.com
> From: Tom Tromey <tromey@redhat.com>
> Date: Tue, 14 Apr 2009 10:31:13 -0600
> 
> >>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:
> 
> Tom> Please try the appended patch.
> 
> Eli> It compiles and links okay, thanks.  Do you want me to test some
> Eli> feature(s) to make sure the resulting binary works as expected?
> 
> Sure.  If you have dejagnu, try charset.exp.

I ran some of the tests in charset.exp by hand, and I see some
potential issues.

For example, when I set the host charset to UTF-8 and the target
charset to EBCDIC-US, I get an error message:

    Cannot convert between character sets `EBCDIC-US' and `UTF-8'

Is this expected?  Is this again a consequence of the list of charsets
not coming from libiconv, but from the default?

Also, this result seems to be different from what charsets.exp
expects:

    Breakpoint 1, main () at charset.c:173
    173        }
    (gdb) set host-charset ASCII
    (gdb) set target-charset ISO-8859-1
    (gdb) print iso_8859_1_string[69]
    $2 = -94 '\242'
    (gdb) print iso_8859_1_string+70
    $3 = 0xbdb6 "\21"
    (gdb) print iso_8859_1_string[70]
    $4 = 17 '\21'

Perhaps I misunderstood, but this part of the test seems to expect
something different, at least for the second and third print commands:

    # Test handling of characters in the target charset which
    # can't be translated into the host charset.
    if {! [string compare $target_charset iso-8859-1]} {
        gdb_test "print iso_8859_1_string\[69\]" \
                 " = \[0-9-\]+ '\\\\242'" \
                 "print character with no equivalent in host character set"
        gdb_test "print iso_8859_1_string + 70" \
                 " = ${hex} \"\\\\242.*\"" \
                 "print string with no equivalent in host character set"
    }

I also noticed that on a GNU/Linux system, "set target-charset TAB"
shows the list of possible charsets with a ^J character at the end of
each line.