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: Don't include function names in test data in generated libm-test.c


On 05/22/2013 05:31 PM, Joseph S. Myers wrote:
This patch makes the strings in the test structures in the generated
libm-test.c that are used to generate test names contain only the
function arguments, and not the function name, with the full test name
being computed at runtime (the same way as is already done for "Real
part of" and "Imaginary part of" test names).  Where a test covers
multiple outputs of a function, only one string is now needed in this
structure, rather than a separate name for each test.

This will allow test arrays to be shared between multiple functions
with identical semantics for their test data (remainder, drem; lgamma,
lgamma_r, gamma; isfinite, finite; exp10, pow10).  It will also allow
the START_DATA and END_DATA calls to be removed, and START and END to
be changed into ordinary macros instead of something specially
processed by gen-libm-test.pl (in the case of END, separate macros for
real and complex functions).  When conversion to MPFR-based generation
of expected results for each format allows accurate expected results
for each rounding mode, it will also allow for test structures to be
shared between rounding modes (containing one string, one test input,
four expected results).

Tested x86_64 and x86.

ok, thanks,

Andreas

--
 Andreas Jaeger aj@{suse.com,opensuse.org} Twitter/Identica: jaegerandi
  SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
   GF: Jeff Hawn,Jennifer Guild,Felix Imendörffer,HRB16746 (AG Nürnberg)
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126


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