[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Radical idea, test backwards compatibility instead of recompile

On Thu, Oct 05, 2000 at 12:17:23AM -0400, Greg Stark wrote:
> Ben Collins <bcollins@debian.org> writes:
> > You people are missing one important bit of information. libc.so.6 is
> > perfectly backward compatible. 
> Perfectly!?  Wow, I think TeX was the last piece of software to make that
> claim and it was only after years of testing. You're ready to claim that after
> 24 hours of testing. Amazing.

Theoretically perfect. IOW, let's not assume anything is wrong with it.
Test the upgrades, compile against it, and report bugs. As far as I'm
concerned, it's starting with a clean slate, and only through testing is
it going to be know for sure....for instance, being in unstable and people
using it.

Ulrich's own words (paraphrased) "this release has less known bugs than
2.1.3", which says a lot coming from the person that did most of the

> What I'm suggesting is that we learn a different lesson from the libdb
> breakage. The general reaction has been to run around recompiling packages to
> work around a broken bit of backwards compatibility. Now that we've determined
> that it was the backwards compatibility that failed I'm suggesting we revert
> those mistaken unnecessary recompiles and actually test that glibc is
> backwards compatible. Instead of going on faith that it's "perfect".

If you recompile them, they will be the same as they are now. The libdb in
2.1.94-3 is *runtime* only, meaning you cannot compile against it.
Downgrade to the potato version and see if that still works (2.1.94-4 will
have the conflicts removed, just for that testing).


/  Ben Collins  --  ...on that fantastic voyage...  --  Debian GNU/Linux   \
`  bcollins@debian.org  --  bcollins@openldap.org  --  bcollins@linux.com  '

Reply to: