Re: Trying to fix slang
On Wed, Jan 30, 2002 at 04:16:20AM +0900, Junichi Uekawa wrote:
> The binary incompatibility between utf8 and nonutf8
> versions has not been resolved.
What reasons are there to use the non-UTF8 version? Being able to drop in
different versions is neat, but if there's no reason to use the non-UTF8
version, why lose sleep over it? Make upstream-compatibility packages,
slang1-compat*, keep them binary compatible with upstream, and put the
UTF8 code in libslang1(-dev) as the default. (With a different soname,
so the upstream-compatible package works with binaries compiled
Another idea (adding to the above): change slang1 to be binary-compatible
with slang1-utf8, since the reverse doesn't seem possible. This wouly only
be useful if the non-utf8 package was actually needed (for reasons other
than binary compatibility.)
Of course, the #define UTF8 problem needs to be fixed; the non-UTF8
prototypes should never be used when linking against the UTF8 library.
Any of this would break programs that depend on slang's prototypes, of
course; any programs that actually need this could just use the compat
package directly (if they didn't want to bother adding stubs for the
calls.) This would show up at compile-time, however.
(I don't know about the issues related to the boot floppies--havn't been
following the problem long enough.)