RE: lsbsi call summary 28-Jan-03

> Pending completion of (2) and (3), rebuilds on
> all three architectures, and testing to make sure
> that nothing has regressed, the 1.3 lsbsi looks
> ready to release - this puts it approximately
> one week away.

In the midst of optimism about a release looking
close, I forgot an issue that surfaced again
last week.

The glibc as built in the lsbsi (on ia32 only)
is missing four symbols that are required by
the current LSB spec.  It would be nicer if
these could have been kicked out of the spec,
where they perhaps don't belong, because then
they wouldn't be in the LSB stub libc, and thus
would be satisfied statically out of libgcc,
which *is* included on the link line by lsbcc.

However, that didn't happen.

These are symbols which were included in glibc
by necessity, but as the libgcc compiler-support
library comes into play, are no longer needed
in glibc.  glibc and gcc handshake on whether
to include them in glibc or not at build time;
if glibc 2.2.5 is built by gcc 3.2, they won't
be included.  I don't know how to put them back,
and thus I don't know yet how to fix a critical
failure when running certain conforming apps
inside the ia32 lsbsi.

Any suggestions are welcome - we'll need to fix
this before releasing, or we don't meet the
quality goal of running all the appbat apps.

-- mats