Re: Slink to potato upgrade
At 09:22 +0100 1999-03-22, Richard Braakman wrote:
ijr@po.cwru.edu wrote:
No program *needs* to be compiled to work with the new libraries. Programs
compiled against glibc2.0 can work with libraries compiled against glibc2.1,
this is not an issue.
This is demonstrably not true. Go run slink's elm on a potato system if
you don't believe me.
We need to investigate *why* elm breaks.
The vast, vast majority of programs that break are those that are poorly
written and make use of internal libc symbols.
That's no excuse. Binary compatibility is binary compatibility. If we're
going to stop supporting poorly written code, we might as well give up now.
It happens that some programs accidentally reference internal
symbols, but this is no reason to bump the soname IMO.
Furthermore, glibc's idea of what is "internal" is not exactly standard.
I still remember the problems with fsck and llseek.
The llseek thing was a *mistake*, there is exactly one reason libc
has it at all, libdb needs it. The standards are quite clear that
symbols beginning with _ or __ are reserved namespace. There are some
well known exceptions.
--
Joel Klecker (aka Espy) <URL:http://web.espy.org/>
<URL:mailto:jk@espy.org> <URL:mailto:espy@debian.org>
Reply to: