This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: glibc 2.28 - Ready to cut branch on Wednesday?
- From: DJ Delorie <dj at redhat dot com>
- To: "Carlos O'Donell" <carlos at redhat dot com>
- Cc: libc-alpha at sourceware dot org, palmer at sifive dot com
- Date: Mon, 30 Jul 2018 22:38:48 -0400
- Subject: Re: glibc 2.28 - Ready to cut branch on Wednesday?
"Carlos O'Donell" <carlos@redhat.com> writes:
> DJ, Palmer,
> What's the status for risc-v? How is the testing going?
Jim Wilson and I have tested on hardware, with vastly different results.
Darius Rad tested on QEMU, Jim's have the least failures. We're trying
to narrow down the reasons for the differences, but a lot of the
failures are related to the new POSIX floating point rules, where RISC-V
and glibc differ in what's expected (by the testsuite) vs required (by
POSIX).
This is nothing new to us and only really affect boundary conditions,
where NaNs, Inf, and exceptions live.
We haven't yet found any failures attributable to glibc itself, though,
but that doesn't preclude them. Regenerating ULPs made a minor
difference, but I hesitate to update the ULP tables as there are a lot
of off-by-ULP errors that are *not* fixed by updating the tables.