This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Fix another -Os strcoll build issue
- From: Adhemerval Zanella <adhemerval dot zanella at linaro dot org>
- To: libc-alpha at sourceware dot org
- Date: Mon, 26 Feb 2018 15:25:33 -0300
- Subject: Re: Fix another -Os strcoll build issue
- Authentication-results: sourceware.org; auth=none
- References: <alpine.DEB.2.20.1802231740550.24555@digraph.polyomino.org.uk>
On 23/02/2018 14:41, Joseph Myers wrote:
> While there are now clean -Os build and test results on x86_64 (given
> my patch <https://sourceware.org/ml/libc-alpha/2018-02/msg00602.html>,
> pending review), testing with -Os with build-many-glibcs.py shows the
> build is still failing with -Os everywhere except for x86_64, x86 and
> s390x.
>
> There are a variety of different build failures, but the most common
> seem to be in strcoll / wcscoll, similar to existing such cases where
> DIAG_* are used to disable -Wmaybe-uninitialized. There are various
> different failures even within those functions. This patch fixes one
> particular case that seems quite common, where the warning appears at
> the declarations of seq1 and seq2.
>
> Tested with build-many-glibcs.py that this fixes the -Os build for
> aarch64-linux-gnu with GCC 7.
>
> 2018-02-23 Joseph Myers <joseph@codesourcery.com>
>
> * string/strcoll_l.c: Include <libc-diag.h>.
> (STRCOLL): Ignore -Wmaybe-uninitialized for -Os around
> declarations of seq1 and seq2.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
>
> diff --git a/string/strcoll_l.c b/string/strcoll_l.c
> index 4a63c56..c001ff4 100644
> --- a/string/strcoll_l.c
> +++ b/string/strcoll_l.c
> @@ -24,6 +24,7 @@
> #include <stdint.h>
> #include <string.h>
> #include <sys/param.h>
> +#include <libc-diag.h>
>
> #ifndef STRING_TYPE
> # define STRING_TYPE char
> @@ -291,7 +292,17 @@ STRCOLL (const STRING_TYPE *s1, const STRING_TYPE *s2, locale_t l)
>
> int result = 0, rule = 0;
>
> + /* With GCC 7 when compiling with -Os the compiler warns that
> + seq1.back_us and seq2.back_us might be used uninitialized.
> + Sometimes this warning appears at locations in locale/weightwc.h
> + where the actual use is, but on architectures other than x86_64,
> + x86 and s390x, a warning appears at the definitions of seq1 and
> + seq2. This uninitialized use is impossible for the same reason
> + as described in comments in locale/weightwc.h. */
> + DIAG_PUSH_NEEDS_COMMENT;
> + DIAG_IGNORE_Os_NEEDS_COMMENT (7, "-Wmaybe-uninitialized");
> coll_seq seq1, seq2;
> + DIAG_POP_NEEDS_COMMENT;
> seq1.len = 0;
> seq1.idxmax = 0;
> seq1.rule = 0;
>