This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] ia64: refresh ulps


On Mon, 18 Sep 2017 12:16:24 +0000
Joseph Myers <joseph@codesourcery.com> wrote:

> On Sat, 16 Sep 2017, Sergei Trofimovich wrote:
> 
> > After regen 'make check results':  
> 
> To confirm: you truncated the libm-test-ulps file to regenerate from 
> scratch, so that ulps decreases were possible as well as increases?  
> (Though since ia64 uses mainly ia64-specific function implementations, 
> there are less likely to be decreases than for other architectures 
> regenerating after a long time.)

Yes, it was ran against truncated 'sysdeps/ia64/fpu/libm-test-ulps'.

Or to be precise it was the following command sequence:
    [ ran in glibc-build/; ../glibc/ is a source tree ]
    [ $ ../glibc/configure --prefix=/usr && make && make check ]
    $ echo -n > ../glibc/sysdeps/ia64/fpu/libm-test-ulps
    $ make regen-ulps
    $ cp math/NewUlps ../glibc/sysdeps/ia64/fpu/libm-test-ulps
    [ make check ]

It's best I could reconstruct from https://sourceware.org/glibc/wiki/Regeneration

> > Some of math are still failing and need separate fixing
> > but those are NaN-related. Will be dealt with later.  
> 
> We currently have bugs 10163, 11585, 16401 open for ia64 libm issues 
> (10163 has a patch attached, either never sent to libc-alpha or never 
> reviewed there).  If the testsuite is showing other ia64 libm bugs 
> (generally, any bug that was user-visible in a release and isn't already 
> filed in Bugzilla), could you please file them in Bugzilla?  Then, once a 
> fix for a bug is checked in, it should be resolved as FIXED with the 
> target milestone set to the first mainline release with the fix (so 2.27 
> for any fixes checked in within the next few months), and that information 
> is used to generate the list of fixed bugs for the NEWS file at release 
> time.

Will do. I was not sure to file bugs before regenerating ulps
because many test failures are result of missing data (and not
a behaviour difference).

Thanks!

-- 

  Sergei

Attachment: pgpkt1sOcpoP7.pgp
Description: Цифровая подпись OpenPGP


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]