This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH] Fixes for a small number of compiler warnings
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: Hans-Peter Nilsson <hp at bitrange dot com>, Ed Schouten <ed at nuxi dot nl>, Binutils <binutils at sourceware dot org>
- Date: Tue, 31 Mar 2015 20:57:19 -0700
- Subject: Re: [PATCH] Fixes for a small number of compiler warnings
- Authentication-results: sourceware.org; auth=none
- References: <CABh_MKmxfVGU9b=vD9yHdUWXtkN7==Ptmu4h3izQPGCzsa=b=g at mail dot gmail dot com> <CAMe9rOrys_QG+0opphc9AXn5=Z-Df4Nk0WUFTJEy-dy_ux8Y=g at mail dot gmail dot com> <CABh_MKmKPHAfVo5Q2BkBnQuGFk-x+5a3j46KtXgF9_af2UHrCg at mail dot gmail dot com> <CABh_MKnE8EBOCuTCofXi2CsWsOVq5xtC+cPqZB-NgotzKFFGLg at mail dot gmail dot com> <20150401003744 dot GN26234 at bubble dot grove dot modra dot org> <alpine dot BSF dot 2 dot 02 dot 1503312120010 dot 87955 at arjuna dot pair dot com> <20150401024707 dot GP26234 at bubble dot grove dot modra dot org> <CAMe9rOrpdXYXG-Lsq3UVCnLxwTk-qFAKZE4Ms5UALquTa9cSNA at mail dot gmail dot com> <20150401032705 dot GQ26234 at bubble dot grove dot modra dot org> <CAMe9rOpi+SxEp8BTacMjuaQ_EcYusd9ks_dbHgNTfUwg0bWCDw at mail dot gmail dot com> <20150401034857 dot GR26234 at bubble dot grove dot modra dot org>
On Tue, Mar 31, 2015 at 8:48 PM, Alan Modra <amodra@gmail.com> wrote:
> On Tue, Mar 31, 2015 at 08:32:35PM -0700, H.J. Lu wrote:
>> On Tue, Mar 31, 2015 at 8:27 PM, Alan Modra <amodra@gmail.com> wrote:
>> > On Tue, Mar 31, 2015 at 07:53:36PM -0700, H.J. Lu wrote:
>> >> This becomes a regression with the older GCC.
>> >
>> > Yes. I'd rather cause that than have warnings with more recent gcc
>> > versions.. I also agree that we shouldn't worry too much about
>> > compilers other than gcc, and I know that self-assignment is something
>> > that C++ programmers need to worry about which is no doubt why the
>> > warning is on for Clang.
>> >
>> > I did consider just reverting the previous patch, but fixing a warning
>> > with a self-assignment just looks horrible to me, particularly when
>> > the warning is about an uninitialized variable. I mean, it looks to
>> > me that a self-assignment actually creates an uninitialized variable
>> > access at the abstract level, when there wasn't one there before!
>>
>> For this particular case, self-assignment causes no harm since
>> the variable will be initialized.
>
> I didn't say there was any harm, just that I dislike the style and why
> I dislike it.
>
>>
>> Will
>>
>> struct bfd_link_hash_entry ehdr_start_save = { 0 };
>>
>> for all compilers?
>
How about something like
struct bfd_link_hash_entry
SILENCE_UNINIT_WARNING_WITH_SELF_ASSIGN (ehdr_start_save);
We define it to self-assign only for those affect GCCs.
--
H.J.