This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] fix #19444 - build failures with -O1 due to -Wmaybe-uninitialized
- From: Joseph Myers <joseph at codesourcery dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: Martin Sebor <msebor at gmail dot com>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Thu, 14 Jan 2016 18:49:04 +0000
- Subject: Re: [PATCH] fix #19444 - build failures with -O1 due to -Wmaybe-uninitialized
- Authentication-results: sourceware.org; auth=none
- References: <56968319 dot 3070802 at gmail dot com> <alpine dot DEB dot 2 dot 10 dot 1601131740310 dot 17296 at digraph dot polyomino dot org dot uk> <CAMe9rOpBfq9q2X3+7GKe4APKuR6zPnzxDqsv86_aLAw+uN+vhg at mail dot gmail dot com> <alpine dot DEB dot 2 dot 10 dot 1601131756510 dot 17296 at digraph dot polyomino dot org dot uk> <CAMe9rOoHDQCOjs9HT-6fo5ayrLd-4d7VuyZa9xJMJkza9kE5Jg at mail dot gmail dot com> <alpine dot DEB dot 2 dot 10 dot 1601141802590 dot 9114 at digraph dot polyomino dot org dot uk> <CAMe9rOr4k5StujVk2vA8HyErJtrhoFvckrT9TtL=nxro4JdKFQ at mail dot gmail dot com>
On Thu, 14 Jan 2016, H.J. Lu wrote:
> This is the right direction, but may not work with -Os. I am working
> on a patch to add bits/string-1.h to define _STRING_ARCH_unaligned,
> which is included unconditionally.
Suppose you apply all Wilco's patches related to string inlines (including
those that remove many of them from bits/string2.h). Are there any
remaining references to _STRING_ARCH_unaligned in installed headers other
than the definition?
If so, what are they? If not, it should be moved out of installed headers
as I said (and bits/ is a namespace exclusively for installed headers).
Installed headers should not define things only relevant for building
glibc, or contain _LIBC conditionals, except where we don't have a cleaner
alternative for how to avoid internals creeping into installed headers.
The first preference should be that internal things go in internal headers
(which includes the include/ wrappers, though arguably those should be
slimmed down as well with more internal interfaces going in separate
internal headers that are explicitly included, not in a header with the
same name as one that gets installed).
--
Joseph S. Myers
joseph@codesourcery.com