This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH v3] Add tests for the long double version of ecvt and fcvt
- From: Florian Weimer <fweimer at redhat dot com>
- To: "Gabriel F. T. Gomes" <gabriel at inconstante dot eti dot br>
- Cc: <libc-alpha at sourceware dot org>
- Date: Wed, 12 Dec 2018 14:23:49 +0100
- Subject: Re: [PATCH v3] Add tests for the long double version of ecvt and fcvt
- References: <20181108222916.27414-1-gabriel@inconstante.eti.br> <87bm5sulmt.fsf@oldenburg2.str.redhat.com> <20181212110741.40805913@tereshkova>
* Gabriel F. T. Gomes:
> On Tue, 11 Dec 2018, Florian Weimer wrote:
>
>>* Gabriel F. T. Gomes:
>>
>>> * misc/tst-efgcvt-template.c: Renamed from misc/tst-efgcvt.c.
>>
>>I think you should add a file comment with a brief list of all the
>>customization macros.
>
> Is the following text all right?
>
> /* This template provides testing for the *cvt family of functions,
> which deal with double or long double types. In order to use the
> template, the following macros must be defined before inclusion of
> this template:
>
> FLOAT: The floating-point type, i.e. double or long double.
>
> ECVT: Appropriate *ecvt function for FLOAT, i.e. ecvt or qecvt.
> FCVT: Likewise for *fcvt, i.e. fcvt or qfcvt.
> ECVT_R: Likewise for *ecvt_r, i.e. ecvt_r or qecvt_r.
> FCVT_R: Likewise for *fcvt_r, i.e. fcvt_r or qfcvt_r.
>
> NAME: Macro to stringify the name of the function, e.g. ecvt,
> possibly adding prefixes or suffixes to it, e.g. qecvt.
>
> PRINTF_CONVERSION: The appropriate printf conversion specifier with
> length modifier for FLOAT, i.e. "%f" or "%Lf".
>
> EXTRA_ECVT_TESTS: Additional tests for the ecvt or qecvt function
> that are only relevant to a particular floating-point type and
> cannot be represented generically. */
Looks fine to me.
>>Do you expect NAME to be customizable in the future?
>
> No more than what is already implemented, i.e. when FLOAT is double,
> errors print the function name as ecvt or fcvt, but when FLOAT is long
> double, you get qecvt, qfcvt, etc. For the long double transition on
> powerpc64le, I do *not* plan to make this test print the internal name of
> the functions, e.g. __qecvtieee128.
Can you move NAME into the template file, then? Does it really differ
across test variants?
Thanks,
Florian