[Mesa-dev] [PATCH] Revert "configure.ac: Disable compiler optimizations when --enable-debug is set"

Emil Velikov emil.l.velikov at gmail.com
Tue Jan 8 11:17:55 PST 2013

On 08/01/13 17:27, Marek Olšák wrote:
> On Tue, Jan 8, 2013 at 6:11 PM, Matt Turner <mattst88 at gmail.com> wrote:
>> On Tue, Jan 8, 2013 at 8:38 AM, Marek Olšák <maraeo at gmail.com> wrote:
>>> This reverts commit 122345876479cf5cf553e38162ab105658614ab7.
>>> Some distributions use --enable-debug for testing packages and the commit
>>> results in terrible CPU performance. It can be embarrassing for us.
>>> Those who do not want optimizations should set the CFLAGS and CXXFLAGS
>>> environment variables. In my opinion, --enable-debug should only set -DDEBUG.
>>> ---
>> Dan's two patches from yesterday handle this while still setting -O0
>> if CFLAGS don't contain -O*. Do those patches work for your purposes?
>> The problem here is that we're just going back and forth on this
>> behavior and both behaviors are reasonable for different things.
> The problem is that we try to set -O0 for --enable-debug and I don't
> like that and I'm not the only one. If you want -O0, you should add
> another flag e.g. --disable-optimizations. The default behavior should
> always be -O2, because I always want optimizations, but I also want to
> sometimes turn on assertions and other run-time checks depending on
I never thought that such a small change would cause such a lengthy
discussion. Possibly I should have expected it considering that it
causes quite noticeable change.

But to be honest, with all sanity I was not expecting distros to build
packages with --enable-debug :P

Considering all the arguments, as well as the annoying regression this
has caused on Open64 builds please revert the commit.

Dan's patches are quite nice, although the default optimization level
(for gcc at least) is already -O0

> Also, people expect a certain behavior from --enable-debug and we
> shouldn't change that.
> Adding --disable-optimizations is the way to go for those who need -O0.
This option sounds like a reasonable approach, that I will use if
available (patch will is coming shortly)

> Marek
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev

More information about the mesa-dev mailing list