This is the mail archive of the
newlib@sourceware.org
mailing list for the newlib project.
Re: Long double complex methods
- From: Joel Sherrill <joel dot sherrill at oarcorp dot com>
- To: Dionna Amalie Glaze <dionnaglaze at google dot com>, Aditya Upadhyay <aadit0402 at gmail dot com>
- Cc: "newlib at sourceware dot org" <newlib at sourceware dot org>
- Date: Wed, 28 Jun 2017 16:57:25 -0500
- Subject: Re: Long double complex methods
- Authentication-results: sourceware.org; auth=none
- References: <CAMZxAxdMKkLBbx9uognWNRxPwQwc89U89F8fsT1ksFq+opi-cQ@mail.gmail.com> <CAAH4kHbk0ACL7uGYAy4PaXD=_rPDviO08y=Zcn9J3vLO2oweKA@mail.gmail.com>
On 6/28/2017 4:36 PM, Dionna Amalie Glaze via newlib wrote:
Gave it a quick look-over (I'm not a core dev, but figured I'd look
since you seem eager). It looks like everything is indeed from the
current NetBSD codebase.
cacoshl.c has #if 0'd code that probably shouldn't be checked in? The
#if 0 is directly from NetBSD, but still #if 0s aren't nice.
clogl.c why not use cargl(z) for rr? The compiler would inline it, and
arg(z) is closer to the math.
cprojl.c More of a note to others: I was confused why you would use
HUGE_VAL instead of __INFINITY, but apparently that was a conscious
change here http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libm/complex/cprojl.c?rev=1.6&content-type=text/x-cvsweb-markup&only_with_tag=MAIN
from http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libm/complex/cprojl.c?rev=1.5&content-type=text/x-cvsweb-markup&only_with_tag=MAIN
I can't argue the rationale for any of the above except to
say that the goal is to import unmodified source from
third party sources. Newlib generally trusts that the
upstream sources are "reputable".
Would you please include additions to libm/test/ for your included
functionality? That probably should be upstreamed to NetBSD too, since
I see they have no tests *gasp*
http://www.mirrorservice.org/pub/NetBSD/NetBSD-current/src/tests/lib/libm/
Agreed they need tests but that would be a NetBSD issue.
Also, although I don't know how to run them, doesn't someone
run glibc tests on newlib? They likely have tests for this for
newlib's purposes.
<hint>I would like to know how to run then</hint> :)
Aditya.. file a NetBSD ticket with the list of methods they
don't have any tests for. I may have to get a NetBSD account
so I can track the ticket.
FWIW Aditya is a Google Summer of Code student for RTEMS working
to improve POSIX compliance for newlib and by implication
RTEMS and Cygwin. Issues with upstream sources with the code
should be filed upstream.
On Wed, Jun 28, 2017 at 1:34 PM, Aditya Upadhyay <aadit0402@gmail.com> wrote:
Hello Developers,
I have ported all long double complex methods with suffix "l" from
NetBSD. I am attaching the patches for all the methods. I am
requesting you to please review these patches and point me any
modification needed.
Thanks & Regards,