[Buildroot] disable-optimization in libgcrypt?

[Danomi Manchego]

On Fri, May 29, 2015 at 4:07 PM, Bernhard Reutner-Fischer
<rep.dot.nop at gmail.com> wrote:
> On May 29, 2015 9:34:26 PM GMT+02:00, Danomi Manchego <danomimanchego123 at gmail.com> wrote:
>>All,
>>
>>We noticed that libgcrypt.mk explicitly disables optimizations:
>>
>>   LIBGCRYPT_CONF_OPTS = \
>>    --disable-optimization \
>>    --with-gpg-error-prefix=$(STAGING_DIR)/usr
>>
>>Looking through the history, it has been so since 2007, commit
>>595e070364fa578dcc44bea7256ef635368bad9c by Bernhard Reutner-Fischer.
>>The commit log does not explain why - "remove duplicate includedir".
>>
>>I don't see optimization being disabled in OE, Yocto, PTXdist.
>>
>>Does anyone know why optimization is disabled in libgcrypt?
>
> I do not remember if i disabled it on purpose, maybe an ICE?
> Or maybe I wanted the buildroot flags to take effect rather than the -O3 (?) the package wanted to, back then. The latter sounds more likely, now that I think of it.

The buildroot optimization setting is not making it down to the
compile lines.  We got a tremendous run time improvement when we
removed the --disable-optimization.

It looks like the --disable-optimization actively kills any -O passed
in on CFLAGS in libgcrypt's configure script:

# This is handy for debugging so the compiler doesn't rearrange
# things and eliminate variables.
# Check whether --enable-optimization was given.
if test "${enable_optimization+set}" = set; then :
  enableval=$enable_optimization; if test $enableval = no ; then
                         CFLAGS=`echo $CFLAGS | sed 's/-O[0-9]//'`
                       fi
fi


The optimization level set in the menuconfig appears to survive when I
remove this option from the .mk file.

Seems like that's the way to go.

Danomi -


> Please drop it if it fits the current buildroot style.
>
> Cheers,
>
>