This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: bz1311954 - multilib variations in LC_COLLATE files, with fixes
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: DJ Delorie <dj at redhat dot com>
- Cc: "Carlos O'Donell" <codonell at redhat dot com>, Rafal Luzynski <digitalfreak at lingonborough dot com>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Wed, 10 Apr 2019 10:52:09 -0700
- Subject: Re: bz1311954 - multilib variations in LC_COLLATE files, with fixes
- References: <ca9e8bf9-f36f-4669-3f41-4e0cf53b2c67@redhat.com> <xnpnqkuauy.fsf@greed.delorie.com>
On Wed, Mar 20, 2019 at 9:09 PM DJ Delorie <dj@redhat.com> wrote:
>
> "Carlos O'Donell" <codonell@redhat.com> writes:
> > Very minor. Thanks, I think this means that using 50% is the simplest
> > and easiest option then. Care to post patches that use 50% with >> ?
>
> iconv, localedef: avoid floating point rounding differences
>
> Two cases of "int * 1.4" may result in imprecise results, which
> in at least one case resulted in i686 and x86-64 producing
> different locale files. This replaced that floating point multiply
> with integer operations. While the hash table margin is increased
> from 40% to 50%, testing shows only 2% increase in overall size
> of the locale archive.
>
> https://bugzilla.redhat.com/show_bug.cgi?id=1311954
>
> diff --git a/iconv/iconvconfig.c b/iconv/iconvconfig.c
> index 0201450742..1e6066cdf0 100644
> --- a/iconv/iconvconfig.c
> +++ b/iconv/iconvconfig.c
> @@ -1079,9 +1079,9 @@ write_output (void)
>
> /* Create the hashing table. We know how many strings we have.
> Creating a perfect hash table is not reasonable here. Therefore
> - we use open hashing and a table size which is the next prime 40%
> + we use open hashing and a table size which is the next prime 50%
> larger than the number of strings. */
> - hash_size = next_prime (nnames * 1.4);
> + hash_size = next_prime (nnames + nnames >> 1);
> hash_table = (struct hash_entry *) xcalloc (hash_size,
> sizeof (struct hash_entry));
> /* Fill the hash table. */
> diff --git a/locale/programs/ld-collate.c b/locale/programs/ld-collate.c
> index bb4e2c539d..19b23c2453 100644
> --- a/locale/programs/ld-collate.c
> +++ b/locale/programs/ld-collate.c
> @@ -2401,8 +2401,8 @@ collate_output (struct localedef_t *locale, const struct charmap_t *charmap,
>
> runp = runp->next;
> }
> - /* Add 40% and find the next prime number. */
> - elem_size = next_prime (elem_size * 1.4);
> + /* Add 50% and find the next prime number. */
> + elem_size = next_prime (elem_size + elem_size >> 1);
>
> /* Allocate the table. Each entry consists of two words: the hash
> value and an index in a secondary table which provides the index
This caused:
https://sourceware.org/bugzilla/show_bug.cgi?id=24442
--
H.J.