]> code.ossystems Code Review - openembedded-core.git/commit
libtool: avoid running automake/autoconf --version
authorRobert Yang <liezhi.yang@windriver.com>
Thu, 5 Feb 2015 09:31:28 +0000 (01:31 -0800)
committerRichard Purdie <richard.purdie@linuxfoundation.org>
Thu, 5 Feb 2015 09:45:49 +0000 (09:45 +0000)
commiteb9d896db2fc67bac8efd258744d06fbbee87f06
treea4c489322dd6ff000b303d98f317a76df5f85b12
parent2104111edc85d057eb4fadecd007f5c592803da6
libtool: avoid running automake/autoconf --version

The libtool would be very slow if run "automake/autoconf --version", for
example, when compile xz-native (make -j1, only compile, no confiure or
install):

before patched: 19s
after patched:  11s

Use plain text to instead of running them.

NOTE: it is still a little slower than libtool 2.4.2 when compile
xz-native because of other parts:
make -j1: about 2s slower
make -j32: about 0.4s slower

If we run to do_install:
(PARALLEL_MAKE = "-j32")
libtool 2.4.2:
$ bitbake xz-native -cinstall && bitbake xz-native -ccleansstate && time bitbake xz-native -cinstall
real    0m21.092s
user    0m28.292s
sys     0m3.932s

libtool 2.4.5:
$ bitbake xz-native -cinstall && bitbake xz-native -ccleansstate && time bitbake xz-native -cinstall
real    0m21.380s
user    0m31.140s
sys     0m4.068s

About 0.3s, slightly different.

But when using /bin/bash as CONFIG_SHELL, the new libtool would be much slower:
real    0m23.106s
user    0m44.044s
sys     0m4.280s

About 2s slower, for the big package like cairo, it is more slower (about 6s),
unfortunately, /bin/bash is most default CONFIG_SHELL for the recipes since
configure checks first check bash.

Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
meta/recipes-devtools/libtool/libtool-2.4.5.inc
meta/recipes-devtools/libtool/libtool/avoid-running-automake-autoconf-version.patch [new file with mode: 0644]