build: Initial bazel build file changes for Windows by sesmith177 · Pull Request #3884 · envoyproxy/envoy

sesmith177 · 2018-07-18T14:52:51Z

Description:
This is the first step in breaking up #3786 into smaller chunks. This contains the Bazel config / compiler options to allow Envoy to build on Windows. Future PRs will address the external deps build scripts and PGV.

Risk Level:
Low
Testing:
Ran bazel build //source/exe:envoy-static and bazel test //test/... on Linux
Docs Changes:
None
Release Notes:
None

This just contains the "non-controversial changes" to get Envoy building on Windows. Future PRs will address: 1. getting external deps to build on Windows 2. getting PGV working on Windows Signed-off-by: Sam Smith <sesmith177@gmail.com> Signed-off-by: Amin Jamali <ajamali@pivotal.io> Signed-off-by: Matthew Horan <mhoran@pivotal.io> Signed-off-by: Akshat Gokhale <agokhale@pivotal.io>

junr03 · 2018-07-18T15:26:11Z

cc @laszlocsomor

jmillikin-stripe · 2018-07-18T16:29:10Z

LGTM

laszlocsomor · 2018-07-19T07:26:32Z

bazel/BUILD

+    values = {"cpu": "x64_windows"},
+)
+
+config_setting(


Why do you define 3 different config_setting rules for Windows for the compilation_mode options? They are all used the same way, you could substitute them with //bazel:windows_x86_64.

We need the 3 different config_setting rules for this select here: https://github.com/greenhouse-org/envoy/blob/windows-build-pr/bazel/envoy_build_system.bzl#L36-L41

Specifically, we want to never append -ggdb3 to the copts on Windows, but we want to keep the selection mechanism on Linux. If we just add //bazel:windows_x86_64 to that select, it will fail since multiple options will be true (e.g. //bazel:windows_x86_64 and //bazel:dbg_build).

As far as we know, this is the best way to select on multiple conditions -- did we miss some other way to do this?

Ah, I see. No, I think you're right. Thanks for the explanation!

laszlocsomor · 2018-07-19T07:35:32Z

ci/build_setup.ps1

+echo "ENVOY_BAZEL_ROOT: $env:ENVOY_BAZEL_ROOT"
+echo "ENVOY_SRCDIR: $env:ENVOY_SRCDIR"
+
+$env:BAZEL_BASE_OPTIONS="--nomaster_bazelrc --output_base=$env:ENVOY_BAZEL_ROOT --bazelrc=$env:ENVOY_SRCDIR\windows\tools\bazel.rc --batch"


--batch is deprecated. You'll have to manually shut down the Bazel server afterwards.

laszlocsomor · 2018-07-19T07:35:56Z

ci/build_setup.ps1

+echo "ENVOY_SRCDIR: $env:ENVOY_SRCDIR"
+
+$env:BAZEL_BASE_OPTIONS="--nomaster_bazelrc --output_base=$env:ENVOY_BAZEL_ROOT --bazelrc=$env:ENVOY_SRCDIR\windows\tools\bazel.rc --batch"
+$env:BAZEL_BUILD_OPTIONS="--strategy=Genrule=standalone --spawn_strategy=standalone --verbose_failures --action_env=HOME --action_env=PYTHONUSERBASE --jobs=$env:NUM_CPUS --show_task_finish $env:BAZEL_BUILD_EXTRA_OPTIONS"


Does $HOME have any value on Windows? (Also in BAZEL_TEST_OPTIONS below.)

It does not. It doesn't look like we need any of the other action_env or test_env environment variables on Windows, so we'll remove them as well

laszlocsomor · 2018-07-19T07:40:34Z

windows/setup/workstation_setup.ps1

+
+trap { $host.SetShouldExit(1) }
+
+Invoke-WebRequest -UseBasicParsing "https://aka.ms/vs/15/release/vs_buildtools.exe" -OutFile "$env:TEMP\vs_buildtools.exe"


FYI: For file downloads, Start-BitsTransfer is faster than Invoke-WebRequest. You may be able to use that.

laszlocsomor · 2018-07-19T07:46:34Z

windows/tools/bazel.rc

@@ -0,0 +1,6 @@
+# Windows/Envoy specific Bazel build/test options.
+
+build --experimental_shortened_obj_file_path


FYI, this flag:

is false by default in Bazel 0.15.2

will be true by default in Bazel 0.16.0

will be unsupported by Bazel 0.17.0

- --batch is deprecated - the --action_env and --test_env settings aren't necesssary on Windows Signed-off-by: Sam Smith <sesmith177@gmail.com> Signed-off-by: Akshat Gokhale <agokhale@pivotal.io>

Signed-off-by: Akshat Gokhale <agokhale@pivotal.io> Signed-off-by: Sam Smith <sesmith177@gmail.com>

will be removed in future versions of Bazel Signed-off-by: Sam Smith <sesmith177@gmail.com> Signed-off-by: Akshat Gokhale <agokhale@pivotal.io>

laszlocsomor · 2018-07-19T15:36:32Z

LGTM

htuch

Awesome, thanks.

rgs1 · 2018-07-24T21:23:35Z

@sesmith177 @htuch this broke the example filter:

~/src/envoy-filter-example (master) > bazel build //:envoy
Starting local Bazel server and connecting to it...
...................
DEBUG: /home/rgs/.cache/bazel/_bazel_rgs/978e967c6368412c05ae0af37ca6ae3c/external/io_bazel_rules_go/proto/def.bzl:138:3: You no longer need to call proto_register_toolchains(), it does nothing
ERROR: /home/rgs/src/envoy-filter-example/BUILD:10:1: no such package 'bazel': BUILD file not found on package path and referenced by '//:envoy'
ERROR: /home/rgs/src/envoy-filter-example/BUILD:10:1: no such package 'bazel': BUILD file not found on package path and referenced by '//:envoy'
ERROR: Analysis of target '//:envoy' failed; build aborted: no such package 'bazel': BUILD file not found on package path
INFO: Elapsed time: 15.514s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (5 packages loaded)

htuch · 2018-07-24T21:24:45Z

@sesmith177 could you investigate? Thanks.

rgs1 · 2018-07-24T21:36:50Z

@sesmith177 to repro (on Linux):

git clone https://github.com/envoyproxy/envoy-filter-example.git 
cd envoy-filter-example
git submodule update --init
bazel build //:envoy

sesmith177 · 2018-07-24T21:37:48Z

OK, I can repro

sesmith177 · 2018-07-24T21:50:02Z

It looks like some of the references to "//bazel:windows_x86_64" in https://github.com/envoyproxy/envoy/blob/master/bazel/envoy_build_system.bzl weren't properly prefixed with the repository name.

I noticed that some select statements in that file hardcode "@envoy", e.g.

envoy/bazel/envoy_build_system.bzl

Line 87 in e89c9d6

"@envoy//bazel:coverage_build": [],

and some use a repository variable, e.g.:

envoy/bazel/envoy_build_system.bzl

Line 35 in e89c9d6

repository + "//bazel:opt_build": [] if test else ["-ggdb3"],

. Is there a reason that some are hardcoded and some are not?

Anyways, fixing these prefixes up allows the build to succeed. How should we proceed from here?

@htuch @rgs1

htuch · 2018-07-24T22:04:02Z

I would use @envoy everywhere. If you can get the small fix PR up I can approve and we can merge quickly, no need to roll back.

sesmith177 · 2018-07-24T22:29:18Z

Done, submitted: #3947

rgs1 · 2018-07-25T00:30:00Z

@sesmith177 hmm still seeing this with our internal build (which operates like the filter example):

17:12:50 Makefile:64: recipe for target '/build/tmp/_bazel_bazel/b570b5ccd0454dc9af9f65ab1833764d/external/envoy_deps_cache_2c744dffd279d7e9e0910ce594eb4f4f/benchmark.dep' failed
17:12:50 make: *** [/build/tmp/_bazel_bazel/b570b5ccd0454dc9af9f65ab1833764d/external/envoy_deps_cache_2c744dffd279d7e9e0910ce594eb4f4f/benchmark.dep] Error 1
17:12:50 + 1532477553.908601192 . /source/envoy/ci/build_container/build_recipes/libevent.sh
17:12:50 ++ 1532477553.909974402 set -e
17:12:50 ++ 1532477553.911713828 VERSION=2.1.8-stable
17:12:50 ++ 1532477553.913660857 curl https://github.com/libevent/libevent/archive/release-2.1.8-stable.tar.gz -sLo libevent-release-2.1.8-stable.tar.gz
17:12:50 ++ 1532477554.316359265 tar xf libevent-release-2.1.8-stable.tar.gz
17:12:50 ++ 1532477554.342869444 cd libevent-release-2.1.8-stable
17:12:50 ++ 1532477554.344297614 mkdir build
17:12:50 ++ 1532477554.346709184 cd build
17:12:50 ++ 1532477554.348043344 build_type=Release
17:12:50 ++ 1532477554.349984389 [[ '' == \W\i\n\d\o\w\s\_\N\T ]]
17:12:50 ++ 1532477554.351489148 cmake -G Ninja -DCMAKE_INSTALL_PREFIX=/build/tmp/_bazel_bazel/b570b5ccd0454dc9af9f65ab1833764d/external/envoy_deps_cache_2c744dffd279d7e9e0910ce594eb4f4f/thirdparty_build/ -DEVENT__DISABLE_OPENSSL:BOOL=on -DEVENT__DISABLE_REGRESS:BOOL=on -DCMAKE_BUILD_TYPE=Release ..
17:12:50 CMake Error: CMake was unable to find a build program corresponding to "Ninja".  CMAKE_MAKE_PROGRAM is not set.  You probably need to select a different build tool.
17:12:50 CMake Error: CMAKE_C_COMPILER not set, after EnableLanguage
17:12:50 -- Configuring incomplete, errors occurred!

on my side... investigating...

rgs1 · 2018-07-25T00:30:51Z

@sesmith177 ah, looks like we need to update our build container to include ninja. sorry for the false alarm.

sesmith177 changed the title ~~Initial bazel build file changes for Windows~~ build: Initial bazel build file changes for Windows Jul 18, 2018

junr03 requested a review from jmillikin-stripe July 18, 2018 15:25

junr03 assigned htuch Jul 18, 2018

laszlocsomor reviewed Jul 19, 2018

View reviewed changes

ajgokhale and others added 3 commits July 19, 2018 09:37

Cleanup bazel build options

1e5ccfc

- --batch is deprecated - the --action_env and --test_env settings aren't necesssary on Windows Signed-off-by: Sam Smith <sesmith177@gmail.com> Signed-off-by: Akshat Gokhale <agokhale@pivotal.io>

Use Start-BitsTransfer instead of Invoke-WebRequest

93dd729

Signed-off-by: Akshat Gokhale <agokhale@pivotal.io> Signed-off-by: Sam Smith <sesmith177@gmail.com>

Add note that experimental_shortened_obj_file_path

8d04a12

will be removed in future versions of Bazel Signed-off-by: Sam Smith <sesmith177@gmail.com> Signed-off-by: Akshat Gokhale <agokhale@pivotal.io>

htuch approved these changes Jul 24, 2018

View reviewed changes

htuch merged commit e89c9d6 into envoyproxy:master Jul 24, 2018

dio mentioned this pull request Aug 1, 2018

problem about Compiling envoy by bazel #4011

Closed

sesmith177 deleted the windows-build-1 branch October 5, 2018 16:14


		trap { $host.SetShouldExit(1) }

		Invoke-WebRequest -UseBasicParsing "https://aka.ms/vs/15/release/vs_buildtools.exe" -OutFile "$env:TEMP\vs_buildtools.exe"

		@@ -0,0 +1,6 @@
		# Windows/Envoy specific Bazel build/test options.

		build --experimental_shortened_obj_file_path

Conversation

sesmith177 commented Jul 18, 2018

Uh oh!

junr03 commented Jul 18, 2018

Uh oh!

jmillikin-stripe commented Jul 18, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

laszlocsomor commented Jul 19, 2018

Uh oh!

htuch left a comment

Choose a reason for hiding this comment

Uh oh!

rgs1 commented Jul 24, 2018

Uh oh!

htuch commented Jul 24, 2018

Uh oh!

rgs1 commented Jul 24, 2018

Uh oh!

sesmith177 commented Jul 24, 2018

Uh oh!

sesmith177 commented Jul 24, 2018

Uh oh!

htuch commented Jul 24, 2018

Uh oh!

sesmith177 commented Jul 24, 2018

Uh oh!

rgs1 commented Jul 25, 2018

Uh oh!

rgs1 commented Jul 25, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

rgs1 commented Jul 25, 2018 •

edited

Loading