在gentoo环境下安装tensorflow

谷歌的tensorflow包在gentoo环境下安装挺非周折，作者折腾了一天，总算搞定了。

我的gentoo是2013年安装的，计划安装tensorflow 0.5.0，上网了解到tensorflow最好的使用环境是python，虽然我不了解python，但有其它语言做基础，觉得安装没什么难度。于是不加思索直接下了个tensorflow包，解压后就想试用，在python环境里：

可想而知，碰了一鼻子灰。

查谷歌文档：http://wiki.jikexueyuan.com/project/tensorflow-zh/get_started/os_setup.html，觉得我这个环境（主要是有两个版本的python，2.7.5和3.2），最好方法用virtualenv安装，不容迟疑，那就赶快装个virtual 吧！

$ emerge virtualenv

由于存在两个版本的python，好像virtualenv要我决定使用哪个版本python才emerge得到，按照网络上所说打入：

$PYTHON_TARGETS="python2_7" emerge virtualenv

emerge说我有很多包block virtualenv，特别是python block的最狠。在网上到处找，很是烦人，终于在gentoo的官网上找到一丝踪迹，https://www.gentoo.org.cn/mirrorlist/，原来是我的gentoo太老了，长时间不更新，包被阻塞。于是按照gentoo的官网修改/etc/portage/make.conf

# These settings were set by the catalyst build script that automatically# built this stage.# Please consult /usr/share/portage/config/make.conf.example for a more# detailed example.CFLAGS="-O2 -pipe"CXXFLAGS="${CFLAGS}"# WARNING: Changing your CHOST is not something that should be done lightly.# Please consult http://www.gentoo.org/doc/en/change-chost.xml before changing.CHOST="x86_64-pc-linux-gnu"# These are the USE flags that were used in addition to what is provided by the# profile used for building.USE="bindist mmx sse sse2 gtk2 gtk3 -gnome qt4 dvd alsa cdr -minimal branding d$SYNC="rsync://mirrors.ustc.edu.cn/gentoo-portage/"GENTOO_MIRRORS="http://mirrors.aliyun.com/gentoo/"PORTDIR="/usr/portage"DISTDIR="${PORTDIR}/distfiles"PKGDIR="${PORTDIR}/packages"INPUT_DEVICES="evdev synaptics"VIDEO_CARDS="intel"

添加了红色这两行。然后运行命令：emerge-webrsync。

 wenxu ~ # emerge-webrsyncFetching most recent snapshot ...Trying to retrieve 20160425 snapshot from http://mirrors.aliyun.com/gentoo ...Fetching file portage-20160425.tar.xz.md5sum ...Fetching file portage-20160425.tar.bz2.md5sum ...Fetching file portage-20160425.tar.gz.md5sum ...20160425 snapshot was not foundTrying to retrieve 20160424 snapshot from http://mirrors.aliyun.com/gentoo ...Fetching file portage-20160424.tar.xz.md5sum ...Fetching file portage-20160424.tar.xz.gpgsig ...Fetching file portage-20160424.tar.xz ...Checking digest ...Getting snapshot timestamp ...Syncing local tree ...Number of files: 208119Number of files transferred: 766Total file size: 410.83M bytesTotal transferred file size: 2.97M bytesLiteral data: 2.97M bytesMatched data: 0 bytesFile list size: 5.18MFile list generation time: 0.001 secondsFile list transfer time: 0.000 secondsTotal bytes sent: 6.41MTotal bytes received: 42.08Ksent 6.41M bytes  received 42.08K bytes  49.84K bytes/sectotal size is 410.83M  speedup is 63.65Cleaning up ... * IMPORTANT: 25 news items need reading for repository 'gentoo'. * Use eselect news to read news items.wenxu ~ # 

这是最新的portage树和distfiles列表了。关键是再次emerge virtualenv，它居然干开了

心里挺乐哈，按照官网说明，就差几个命令的事！

$ virtualenv --system-site-packages ~/tensorflow$ cd ~/tensorflow

$ source ./bin/activate

这些当然没问题，包括这个也没太大问题：

pip install https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.5.0-cp27-none-linux_x86_64.whl

只是墙的谷歌有时抽风，试了很多次才安装完成。但问题又来了，pip是pip了，可见不着tensorflow的影子，

 $ pip install https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.5.0-cp27-none-linux_x86_64.whlDownloading/unpacking https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.5.0-cp27-none-linux_x86_64.whl  Downloading tensorflow-0.5.0-cp27-none-linux_x86_64.whl (10.9MB): 10.9MB downloadedRequirement already satisfied (use --upgrade to upgrade): numpy>=1.9.2 in ./lib/python2.7/site-packages (from tensorflow==0.5.0)Requirement already satisfied (use --upgrade to upgrade): six>=1.10.0 in ./lib/python2.7/site-packages (from tensorflow==0.5.0)Installing collected packages: tensorflowSuccessfully installed tensorflowCleaning up...

还好，find一下就有了分晓。

 $ find . -name tensorflow./lib/python2.7/site-packages/tensorflow

网上还说tensorflow依赖这几个包：Numpy、SciPy、scikit-learn和Matplotlib。来，pip它们。

该死的墙，我X墙妈八次！没梯子，这时才想到梯子太有用了！用关键词“镜像”搜索，有了，http://www.xuebuyuan.com/2101031.html，国内有，还有pip命令的用法，这网页的兄弟，太谢谢你了！

于是我打入：pip install -i http://pypi.tuna.tsinghua.edu.cn/simple numpy

爽死了，感受到车从泥巴路上了高速！

再pip scipy、scikit-learn、matplotlib，都没问题，我笑的合不拢嘴！可是，进入python，打入import tensorflow as tf命令，问题大了，说我的glibc版本2.15，太低，需要2.17以上，glibc可不好惹，那是命门，搞不好泰山崩于前！要很小心求证，操作，我跌入冰窟窿。。。饭都吃不香，得用emerge，一提emerge我心里总犯毛，太容易block了，而且我不会解决。这前面就像毛驴的前头的萝卜，很难够得着。网上有人说gentoo的安装机制不会对glibc升级构成威胁，于是我放心去emerge了。emerge...emerge，我等阿等，终于等来了successful。

兴奋！赶快去python里打入:import tensorflow as tf试一试，又来错误了，说libcxx版本低，继续emerge sys-libs/libcxx，惨了，无比惨，居然说gcc版本低，编译了一半撂摊子了，不带这样玩的吧？难不成我要升级gcc？

升就升呗，没见世面阿？

打入：emerge gcc

哗啦啦的，屏幕滚动，好不欢快哦！放心吃饭去，晚饭还没吃呢！等回来，悲剧了，是emakeBUILD错误，无法进行下去。这也太不靠谱了，用错误信息搜索，有说警告级别太高，应该降低，有说sanbox版本低导致，都试了，错误依然不改！

不死心，来官网https://wiki.gentoo.org/wiki/Upgrading_GCC/zh-cn上，居然被我撞上了，原来升级gcc先要升libtool。

有了救命稻草，当然不犹豫:

$ emerge --oneshot libtool 

$ emerge gcc

1461627207:  *** emerge  gcc1461627236:  >>> emerge (1 of 1) sys-devel/gcc-4.9.3 to /1461627236:  === (1 of 1) Cleaning (sys-devel/gcc-4.9.3::/usr/portage/sys-devel/gcc/gcc-4.9.3.ebuild)1461627326:  === (1 of 1) Compiling/Merging (sys-devel/gcc-4.9.3::/usr/portage/sys-devel/gcc/gcc-4.9.3.ebuild)1461632519:  === (1 of 1) Merging (sys-devel/gcc-4.9.3::/usr/portage/sys-devel/gcc/gcc-4.9.3.ebuild)1461632536:  >>> AUTOCLEAN: sys-devel/gcc:4.9.31461632582:  === (1 of 1) Updating world file (sys-devel/gcc-4.9.3)1461632582:  === (1 of 1) Post-Build Cleaning (sys-devel/gcc-4.9.3::/usr/portage/sys-devel/gcc/gcc-4.9.3.ebuild)1461632582:  ::: completed emerge (1 of 1) sys-devel/gcc-4.9.3 to /1461632582: === Unmerging... (sys-devel/gcc-4.7.3-r1)1461632591:  >>> unmerge success: sys-devel/gcc-4.7.3-r11461632591:  *** Finished. Cleaning up...1461632592:  *** exiting successfully.1461632593:  *** terminating.

(wang)wenxu@wenxu ~/wang/lib/python2.7/site-packages/tensorflow $ pythonPython 2.7.5 (default, Oct 10 2013, 05:44:34) [GCC 4.7.3] on linux2Type "help", "copyright", "credits" or "license" for more information.>>> import tensorflow>>> hello=tensorflow.constant("Hello,tensorflow!")>>> se=tensorflow.Session()I tensorflow/core/common_runtime/local_device.cc:25] Local device intra op parallelism threads: 2I tensorflow/core/common_runtime/local_session.cc:45] Local session inter op parallelism threads: 2>>> print se.run(hello)Hello,tensorflow!>>> 

大功告成。