Commit Graph

83 Commits

Author SHA1 Message Date
Ian Wienand
8e822768f9 Add flag to disable EPEL
Add DIB_EPEL_DISABLED flag that allows installation of the EPEL repo,
but to have it disabled by default.  This will help when you have
unavoidable EPEL dependencies, but want to make sure you only pull
specific things in with "--enablerepo" calls when installing those
packages.

Change-Id: Iedf6167a7cd69418255ebbee095aea04c50d73fd
2017-04-05 05:59:40 +10:00
Jenkins
6e7d7e77ad Merge "Use stevedore for plugin config of block device" 2017-04-04 08:13:45 +00:00
Hironori Shiina
eaa0569ac5 Fix code-block in README of rhel7 element
A code-block in README of rhel7 element is not rendered as expected.
This patch fixes it to be rendered correctly.

Change-Id: Ie8f4c05edd1dd93314290682e4b2734622894e15
2017-04-03 03:56:22 +00:00
Paul Belanger
87236af75c Have simple-init enable network.service
When a glean is running on centos with multiple NICs, it will try to
systemctl enable network.service multiple times for each interface.
Because of systemd magic, it is possible for the systemctl command to
fail in a race condition.

glean shouldn't be enabling network.service during boot in
pre-networking phases (Ib2b618dd975ca44e9c6b0a2c9027642ffc46b9b0).  I
have proposed I8319f1ed6498a9d447950c2b4b34bca59e7b97e4 to remove this
and document the behaviour.

This also bring across suse's version
(I20bffabd333ea290d8712ec2a467f2b2d5678f3a)

Change-Id: I89d9443cb61e287bd0d9da3f48315272218ee335
Signed-off-by: Paul Belanger <pabelanger@redhat.com>
2017-03-28 19:28:51 +11:00
Andreas Florath
57c9e0bb41 Use stevedore for plugin config of block device
This patch introduces stevedore plugin mechanism for use
with the block device layer.  This makes it possible that
other projects pass in their own block device plugins.

Change-Id: Id3ea56aaf75f5a20a4e1b6ac2a68adb12c56b574
Signed-off-by: Andreas Florath <andreas@florath.net>
2017-03-24 21:42:47 +00:00
Jenkins
f128d08cf0 Merge "Use correct Ubuntu distro url on non-x86 arches" 2017-03-24 04:42:59 +00:00
Jenkins
94e2973785 Merge "functests: skip qcow2 generically but add specific test" 2017-03-24 01:07:10 +00:00
Mikhail S Medvedev
4d4417c9a6 Use correct Ubuntu distro url on non-x86 arches
diskimage-builder usually provides defaults that work out of the box.
One default that does not work outside of x86 land is Ubuntu distro
mirror url. Considering there are only two valid default options, we can
automatically choose a better default.

This patch changes behavior only for architectures known to be using
http://ports.ubuntu.com/ubuntu-ports. All others still would use
http://archive.ubuntu.com/ubuntu as default. It provides some guarantee
that we do not introduce a regression.

Change-Id: If95a64bac0c88f30736da4bae7f1fdce126c0bf6
2017-03-23 09:27:25 -05:00
gecong1973
8ff6f51b81 Typo fix: curent => current
Change-Id: I5ce5a98c355310d6f5f6c09e19993b01ed025f0e
2017-03-23 15:02:20 +08:00
Ian Wienand
76ae374413 functests: skip qcow2 generically but add specific test
We somewhat discussed skipping qcow2 generation previously in
I9372e195913798a851c96e62eee89029e067baa1.  As recent issues with PPC
testing have shown, we are not actually testing the "vm" element and
hence the bootloader path in the functional tests.

I don't think we need to test this on every element; it overlaps
somewhat with the testing done by the nodepool jobs which build full
images and boot them.  I also didn't want to introduce a separate run
for this.  Thus it seems valuable to at least have one element
enhanced to do this installation and conversion in our default tests
for basic sanity.

This disables qcow generation by default, as per the other change, but
allows an element to drop a file that will override the output
formats.  The Xenial element is modified to produce a qcow2 using
this, and also introduces a dependency on the "vm" element so it tries
to install the bootloader.

We now exit if the .qcow2 fails to build as well.

Change-Id: I1a6acefe52f8c696c39b2d592fdc7ae32a87e6fe
2017-03-23 13:49:24 +11:00
Jenkins
2ea3a434f7 Merge "Add default PPC block-device layout" 2017-03-23 01:03:13 +00:00
Jenkins
f7999f86e4 Merge "Capture output in _exec_sudo" 2017-03-23 01:01:13 +00:00
Ian Wienand
19fcd263d0 Add default PPC block-device layout
Add a default PPC block-device layout.  I've extracted this into
separate yaml files for ease of editing and to facilitate things like
longer comments.

This is not sufficient to get PPC images working, but it is required.

Change-Id: I09e5d1ed92260bdb632333f5203dd7e70d512dc8
2017-03-23 09:44:01 +11:00
Ian Wienand
a8f2eaded8 Capture output in _exec_sudo
The stdout of the script is captured, so anything coming out from
these commands needs to be captured.  Move to check_process and show
the output as part of an error log in failure case.

Change-Id: I1150375cdc479d4f19b8ddeb49a824ab16fdf831
2017-03-23 09:31:37 +11:00
Yolanda Robla
8a3a3ffe6e Fix typo in CMDLINE env var for bootloader
The correct spelling of the env var needs to be
DIB_BOOTLOADER_DEFAULT_CMDLINE

Change-Id: I4ae2aed928723588e736b63283a54eca5f4b4c3b
2017-03-22 12:02:37 +01:00
Jenkins
f1d53f2e31 Merge "Use sphinx warning-is-error" 2017-03-15 23:56:59 +00:00
Ian Wienand
fea6ab1624 Use sphinx warning-is-error
Sphix 1.5 (I9e7261c4124b71eeb6bddd9e21747b61bbdc16fa) includes
"warning-is-error" which supersedes pbr's warnerrors.  Enable this and
fix up the resulting failures

 - trailing lines for lists in element_deps directive
 - missing README's that are linked
 - syntax error and highlighting in building instructions

Change-Id: I6549551b4a9bf47076c9811a7a38a666cbea2a50
2017-03-14 14:49:49 +11:00
Michael Johnson
fae8004e4d squash-package-install to use the correct python
99-squash-package-install in the package-installs element does not
know which python environments the requirements were installed into.
This can cause it to select the wrong python to run the
package-installs-squash script.

Co-Authored-By: Adam Harwell <flux.adam@gmail.com>
Change-Id: I5fab0e192c3a2dad8f60e821c184479e24e33bcd
2017-03-13 20:24:39 +11:00
Ian Wienand
85985cdadc Make our virtualenv source py3 safe
execfile() has gone in python3, and if you google various
stackoverflow results, python-dev mailing list threads and other
projects it seems runpy is considered one of the better solutions.
This should be py2.7 safe too.

Change-Id: I18077ba9d603752492cc81f260e12710981f4dff
2017-03-13 20:22:27 +11:00
Ian Wienand
b04c58eca4 blockdevice.py: python3 fixes
keys() is an iterator in python3, so we need to make it list before
finding the first element

Change-Id: Ic158c7b2901c1a34ff417d3432fcefed4760ce24
2017-03-13 19:33:07 +11:00
Jenkins
8b19f7f99d Merge "Fix dib-init-system for Debian Jessie and Debian Stretch" 2017-03-13 03:29:41 +00:00
Andreas Florath
20389d755f Fix dib-init-system for Debian Jessie and Debian Stretch
On Debian Jessie and Debian Stretch systemctl is in /bin.
If the package systemd-sysv is not installed the script
dib-init-system did not find the init system.
This patch fixes the problem: it also looks in /bin
for systemctl and if found decides for systemd.

Change-Id: I5a18052a070bad5e16b14672237a1e2b38513949
Signed-off-by: Andreas Florath <andreas@florath.net>
2017-03-11 14:04:13 +00:00
Yolanda Robla
11275c212b Send custom parameters in bootloader to GRUB_CMDLINE
Depending on the types of deployment (security, nfv...) some extra
kernel flags are needed on the images. This change exposes the
DIB_BOOTLOADER_DEFAULT_CMDLINE parameter, defaulting to the
existing 'nofb nomodeset vga=normal', that will allow to modify
these boot params.

Change-Id: I67d191fa5ca44a57f776cb9739a02dd71212969c
Closes-Bug: #1668890
2017-03-08 15:43:46 +01:00
Yolanda Robla
96504a4de0 Use OrderedDict for partitions instead of simple dictionary
The order of the partitions is important, it needs to be preserved.
If using a simple dict, this is not happening. As a consequence,
checks like 'primary partition being first' are failing because the
dictionary sorts the partitions randomly.
Switched to OrderedDict solved the problem, as it preserves the
ordering it gets from the yaml blob.

Change-Id: Icfa9bd95ffd0203d7c3f6af95de3a6f848c2a954
2017-03-07 15:55:20 +01:00
Andreas Florath
866a06f92d Refactor: block-device partitioning cleanup
Now that the main partitioning refactor patch is merged, there is
a small relict of handling partitions still in the disk-image-create
main.

This patch moves the functionality from disk-image-create to the
block-device/partitioning module: it is mostly a rewrite of the
original bash code in python.

Change-Id: Ia73baeca74180a7bc9ea487da03ff56d6a3070ce
Signed-off-by: Andreas Florath <andreas@florath.net>
2017-03-07 18:43:09 +11:00
Ian Wienand
f1776d72aa Merge branch 'master' into merge-branch
Change-Id: Ib3c4b0030c94459a8c2eca836a16e42edf1accd4
2017-03-02 13:49:29 +11:00
Andreas Florath
003fc17208 Fix #1627402: fix grub2 element for ubuntu xenial
Looks that the special handling for Ubuntu is not needed any longer
(its a pity that there are no detailed comments...).

The grub2 element is a second implementation of the bootstrap element
- but because there are some features that come only here, e.g. efi
boot, it should be working as long as this is not implemented in the
bootloader element.

Change-Id: I74269116ea30b84f3259805720d5cd1616f960c5
Signed-off-by: Andreas Florath <andreas@florath.net>
Closes-Bug: #1627402
2017-02-11 18:44:18 +00:00
Ian Wienand
7a155e08bf Merge branch 'master' into merge-branch
Change-Id: I28e4c7837d84e8b66eff3d182666c5a87a9e3c9b
2017-02-09 13:35:53 +11:00
Andreas Florath
fdd2c4b236 Semi-automatic doc generation of element dependency
Currently there is no description of dependencies in the generated
documentation of the elements: therefore a user of an element does not
know which other elements are automatically included and e.g. which
configuration options are available. In addition there are some
copy&pastes of parts of the README.rst scattered thought different
Ubuntu and Debian specific elements.

This patch adds a semi-automatic generation of dependency information
of all elements.  Nevertheless these are not automatically included.
The author of the element's README.rst can decide if and where the
dependency information should appear and can use the descriptor

.. element_deps::

for this.

This patch adds the dependency information for some Debian and
Ubuntu patches - and creates the base for later removing the
duplicated parts.

A call is added to element_dependencies._find_all_elements() to
populate reverse dependencies for Element objects.

(This is a reworking of I31d2b6050b6c46fefe37378698e9a330025db430 for
the feature/v2 branch)

Change-Id: Iebb83916fed71565071246baa550849eef40560b
2017-02-09 09:50:30 +11:00
Jenkins
0f224d3ae6 Merge "Check return of _load_state" into feature/v2 2017-02-06 23:41:50 +00:00
Andreas Florath
df1f583903 Check return of _load_state
In error scenarios there might be no loadable state available.
This patch adds some robustness handling this scenarios
by checking the result and possible bailing out.

Change-Id: I02e2731b14bec22c22db08d60d8d15db1911a84e
Signed-off-by: Andreas Florath <andreas@florath.net>
2017-02-04 19:21:48 +00:00
Andreas Florath
a8953dd277 block-device: change top level config from dict to list
With the old configuration structure it was only possible
to use one image and one partition layout.  The new
block-device configuration uses a list at top level;
therefore it is possible to use multiple instances
of each element type.

Change-Id: I9db4327486b676887d6ce09609994116dbebfc89
Signed-off-by: Andreas Florath <andreas@florath.net>
2017-02-03 20:15:39 +00:00
Ian Wienand
3f8800832a Release 1.27.0
-----BEGIN PGP SIGNATURE-----
 
 iQEcBAABAgAGBQJYk8+9AAoJEBty/58O8cX8LdIH+wU/VrEVs0XYohiL6DUgabzs
 112U3UUihH5xMc/ca9Tarx+XwEvfMZkwYN2Qr0JoRJjmSt2AL6AezUhGSV+98vaY
 iQEccaFDFYlyDHm4V2r7N1xwS0B3mx87FPqVQQSUKlc3JsQxCy4o9RtD9aM8Gvqy
 +gAxMxL3p3O131K0Rvb0U5lC1FLgft9SuljCV8i5nU4/HdoryD6hedz2/ss8a9KG
 KKEdBKvPBKn73+nb8peQD/VXpej9C31r87q5VEjUsZkJ7gduY/qYLlGGgoBQqAXN
 WQ/ef1RkQKW5ba2jsjnk7fdOrA0+wYENxorR2WecuZbe2ieXw6fP3lYiD6VeWsM=
 =IUuh
 -----END PGP SIGNATURE-----

Merge tag '1.27.0' into merge-branch

Release 1.27.0

Change-Id: I9f6948636cae6d375d1d8315976504021f5a3bbb
2017-02-03 11:49:45 +11:00
Ian Wienand
adf39c52cf Release 1.21.1
-----BEGIN PGP SIGNATURE-----
 
 iQEcBAABAgAGBQJYW2GoAAoJEBty/58O8cX8uSAH/15dJsglP6Zie7jSSJcR6k+e
 PJembHn9qrqrCjmJ5EwakojySaaLhwEJKvlP54OU9v7pmUXL9gJtK2OzW54LQ41g
 xBHIu0Pg4z7juyHm9+1P2Sr7Mzs1pVSEbsIYpDYUU19eghI1EAeIj3I1woKgajN7
 JlI61j3r67G6EAVtPOnmD1jvXS8CrtjiJ9wtWTH20pWfmksovg/GuXUCZrLkAAhO
 NcK35CdMii1Hkr7XOH424La/Ar+3qfUX18ZkbJY6yHzkq/ityTzzKOFjAaDl2Jg9
 WNc+SLCVYpPhPwgt7miTywamUNj3ZviA5/Hd8fuLXmtHSLQ23WOtBiaQMLtwXHs=
 =8dIl
 -----END PGP SIGNATURE-----

Merge tag '1.26.1' into merge-branch

Release 1.21.1

Change-Id: Ib9eb3dd1d384fc5b6a9846608216e056c57a173a
2017-02-02 20:36:23 +11:00
Ian Wienand
19f13958f7 Handle blank lines in element-deps
When merging feature/v2 we have re-enabled the centos-minimal test
which has a blank line in the element-deps.  The modified get_elements
functions added in If97bcd21e45de1b5ed91023fdc441a4617051a6b were not
handling this.

Fixing this, we can extract this into some common code for
element-deps and element-provides.

Also, I noticed while debugging that the element list is an iterator
from reverse(), so printing it doesn't help.  Fix up the debug output.

Change-Id: I8e9b079f586e8a36dba3797a0fc03ca1b2041a04
2017-02-02 18:18:46 +11:00
Ian Wienand
bfca36c772 Release 1.25.2
-----BEGIN PGP SIGNATURE-----
 
 iQEcBAABAgAGBQJYV1yqAAoJEBty/58O8cX8hLwIAKP66w6MdPN8PDgUOteui/Sx
 N0UFKJ9yR4GQOAP0NffPLjch5/g0iJLs3eFKOhtGC1LjbDjpVgjX8vW18ib8wBZK
 GemOZPF3uxg8FROrZF1vpoDy/cHgL1YV10hCnwdjN/r9rb8zOuSabqjW+Dennj2n
 fZ0SJfa8Owfudn3YxGuOymVb/wMtEloDmVGBEI1Y+h7osELCCDi3OXmwsA8qMsdl
 cTwbeugBs4PlOVbZUK/JKGuwIHKgPnDYzYu5KpXw77/MdjGT0fo5Tlq5AOBDI2sC
 9JOFEBDli4Ro05VwvI58ADMpvvOax+9EvOhLbB1dRPdZl21Iyb6gOdy2PUbFO0c=
 =aKxq
 -----END PGP SIGNATURE-----

Merge tag '1.25.2' into merge-branch

Release 1.25.2

Change-Id: I698bcf2e82117bd81649cd065a7af5cac85990c7
2017-02-02 11:20:00 +11:00
Ian Wienand
b6e631360f Release 1.24.0
-----BEGIN PGP SIGNATURE-----
 
 iQEcBAABAgAGBQJYP+r7AAoJEBty/58O8cX8ZcMIAI7Yy2BfGqY/jfyLYNryATSu
 g78McMxAeOiPYII+KKyVkeWsChKJUafLZr/ppO95pm7ZLXeHQeC4MHDI68RmghIX
 HVXn4hPYzwlZ+Yr4hqaEf11H+vw0TDxo2a5bxlQZw+FCx9+3lkR60QxWf8a6QCxv
 nLqKIhtpwGi9mIyvAya+MaOmHMHqbNYgIBHKP0yXhTdxoHZoJWZeXN1rkelwAF9N
 TS1+KwIk+uemn1ZlqzYDWj8euyFyAv8L0WlF0ztCKe7TAFvNeMo26fLykeKHVDEo
 GC5R8MshiOtihPC972rNIponIzd62AetTlDxfK+gcUZO21pNTrlgdLX5Jn04fo0=
 =BpQc
 -----END PGP SIGNATURE-----

Merge tag '1.24.0' into merge-branch

Release 1.24.0

Change-Id: I23cfa6f8dc66bb6a1d957a49a5b468cfd47c005b
2017-01-31 14:14:19 +11:00
Andreas Florath
ec7f56c1b2 Refactor: block-device handling (partitioning)
During the creation of a disk image (e.g. for a VM), there is the need
to create, setup, configure and afterwards detach some kind of storage
where the newly installed OS can be copied to or directly installed
in.

This patch implements partitioning handling.

Change-Id: I0ca6a4ae3a2684d473b44e5f332ee4225ee30f8c
Signed-off-by: Andreas Florath <andreas@florath.net>
2017-01-24 19:59:10 +00:00
Jenkins
6c0eb3d905 Merge "Add get_elements; preserve backwards compat for expand_dependencies" into feature/v2 2017-01-20 10:25:54 +00:00
Jenkins
1b7c415b4a Merge "Activate virtualenv in disk-image-create" into feature/v2 2016-12-20 03:41:19 +00:00
Gregory Haynes
5da7574aee Add get_elements; preserve backwards compat for expand_dependencies
expand_dependencies() was a public interface so we should try and
preserve backwards compat. However, since the interface is really
broken, add a new exported function "get_elements" that instack can
switch to.  This returns the canonical list of elements without
duplicates, and gives the path to each element too.

This highlighted that the unit tests were really a bit wrong.  They're
testing inner functions when we have an "API" in the get_elements()
function.  Convert all unit-tests to use this function instead.  Since
this is a library call, convert the sys.exit() calls to raised
exceptions.

Refactor the variable output into a separate function so we can do a
sanity check on it.

The added flake8 ignores are for the "over-indented for ... indent"
which happens a lot with these new longer lines.  Most other projects
ignore them.

This is an alternative proposal to
I15609389c18adf3017220fc94552514d195b323a

Change-Id: If97bcd21e45de1b5ed91023fdc441a4617051a6b
2016-12-20 08:18:59 +11:00
Ian Wienand
5adfc9365a Fix --version display
We do not have the concept of "not installed" in v2, so remove the
obsolete code looking for a now non-existent variable.

Also, log the version at startup.  This can help when
debugging from logs

Change-Id: I964c4cf207c10666afc5bc7ab9f2bfb9b1897c1e
2016-12-01 04:37:04 +00:00
Ian Wienand
04208e7c79 Activate virtualenv in disk-image-create
Because we're still fundamentally a python program calling a
shell-script, there's some oddities like not having the virtualenv
bin/ in the $PATH if we call disk-image-create directly.

We can detect this, however, and activate the virtualenv before we
fork the disk-image-create shell script so everything "just works".

See also nodepool change I0537cbf167bb18edf26f84ac269cbd9c8a1ea6a2

Change-Id: Ibfea6cf6a6fd0c7f1e468d501c61ae0b58992042
2016-11-30 06:11:09 +11:00
Ian Wienand
448a2602fe Merge remote-tracking branch 'origin/master' into merge-branch
Change-Id: I05cc51c699008018d557ed3874d71af26fd60240
2016-11-29 07:43:46 +11:00
Ian Wienand
7d5afecfd9 Merge remote-tracking branch 'origin/master' into merge-branch
Change-Id: Ibab1bb95521292ae818bd91f7073c3749a2cc0cb
2016-11-18 13:53:56 +11:00
Jenkins
cc3a28755d Merge "Move dib-run-parts into diskimage-builder" into feature/v2 2016-11-15 00:43:49 +00:00
Ian Wienand
3c6e589226 Remove execute perm from disk-image-create
Remove the x bit from lib/disk-image-create; because it's called
directly by the entry-point, it doesn't need to be exectuable.
This should also be clearer that you're not supposed to run it
by hand.

Remove some boilerplate from old file

Change-Id: Ibb6cdae613e6c9cf21dd6aecc8e1f739bc3a2643
2016-11-04 17:11:27 +11:00
Ian Wienand
d65678678e Move dib-run-parts into diskimage-builder
Move dib-run-parts from dib-utils into diskimage-builder directly.

For calling outside the chroot, we provide a standard entry-point
script.  However, as noted in the warning comment, the underlying
script is still copied directly into the chroot by the dib-run-parts
element.  I believe this to be the KISS approach.

This removes the dependency on dib-utils.  We have discussed this
previously and nobody seemed to think retiring dib-utils was going to
be an issue.

This also updates the documentation to not mention dib-utils, or using
disk-image-create via $PATH setup, but rather gives instructions on
installing from pip with a virtualenv.

Change-Id: Ic1e22ba498d2c368da7d72e2e2b70ff34324feb8
2016-11-04 17:07:37 +11:00
Ian Wienand
7446c32197 Move diskimage-image-create to an entry point
It has always been a weird thing that dib is a python package, but
is totally driven by the disk-image-create script.  It creates this
strange division that is hard to explain.

This moves disk-image-create to a regular python entry-point

Currently, this simply exec()s the original disk-image-create script.

However, we now have a (private) interface between disk-image-create
written in python and the driver shell script.  Here's some things we
could do, for example:

* Argument parsing is generally nicer in Python, and then end result
  is mostly just setting environment variables to flag different things
  in the shell script.  I could see us moving the argument-parsing into
  diskimage_builder.disk_image_create:main() and just setting things in
  os.environ before the exec()).

* I7092e1845942f249175933d67ab121188f3511fd sets IMAGE_ELEMENT_YAML in
  disk-image-create by calling-back to element-info.  We can just call
  element_dependencies.find_all_elements() in here an export is to
  os.environ before disk-image-create starts.

* remove need for ramdisk-image-create symlink by just exporting
  IS_RAMDISK based on sys.argv[1] value

* you could even unit test some of this :)

Change-Id: I69ca3d26fede0506a6353c077c69f735c8d84d28
2016-11-02 05:12:11 +00:00
Ian Wienand
97c01e48ed Move elements & lib relative to diskimage_builder package
Currently we have all our elements and library files in a top-level
directory and install them into
<root>/share/diskimage-builder/[elements|lib] (where root is either /
or the root of a virtualenv).

The problem with this is that editable/development installs (pip -e)
do *not* install data_files.  Thus we have no canonical location to
look for elements -- leading to the various odd things we do such as a
whole bunch of guessing at the top of disk-image-create and having a
special test-loader in tests/test_elements.py so we can run python
unit tests on those elements that have it.

data_files is really the wrong thing to use for what are essentially
assets of the program.  data_files install works well for things like
config-files, init.d files or dropping documentation files.

By moving the elements under the diskimage_builder package, we always
know where they are relative to where we import from.  In fact,
pkg_resources has an api for this which we wrap in the new
diskimage_builder/paths.py helper [1].

We use this helper to find the correct path in the couple of places we
need to find the base-elements dir, and for the paths to import the
library shell functions.

Elements such as svc-map and pkg-map include python unit-tests, which
we do not need tests/test_elements.py to special-case load any more.
They just get found automatically by the normal subunit loader.

I have a follow-on change (I69ca3d26fede0506a6353c077c69f735c8d84d28)
to move disk-image-create to a regular python entry-point.

Unfortunately, this has to move to work with setuptools.  You'd think
a symlink under diskimage_builder/[elements|lib] would work, but it
doesn't.

[1] this API handles stuff like getting files out of .zip archive
modules, which we don't do.  Essentially for us it's returning
__file__.

Change-Id: I5e3e3c97f385b1a4ff2031a161a55b231895df5b
2016-11-01 17:27:41 -07:00