jasder
/
antlr
1
0
Fork 1
Code Issues Pull Requests Releases 1 Wiki Activity
antlr/.travis.yml

206 lines
4.2 KiB
YAML
Raw Normal View History

and again
2015-09-18 23:00:39 +08:00
sudo: true
Explicitly lay out the matrix for Travis. This means that we only have one run on macOS, using the default JDK. Travis does not support JDK-switching on macOS.
2016-11-07 08:42:30 +08:00
divide and conquer
2016-11-28 00:56:14 +08:00
language: java
divide and conquer
2016-11-28 01:13:36 +08:00
Remove ~/.m2/repository/org/antlr from the Travis cache. Caching ~/.m2 is useful; caching org/antlr inside there is not. It causes unnecessary re-submission of the cache every build.
2017-10-29 04:13:43 +08:00
before_cache:
- rm -rf $HOME/.m2/repository/org/antlr
Add cache entries to .travis.yml. This is recommended by the Travis docs to speed up the parts of the tests where unpack / install has a significant cost, as opposed to the download. This changeset adds entries for ~/.m2 (for Maven) and ~/Library/Caches/Homebrew for (brew on macOS).
2017-10-13 00:46:38 +08:00
cache:
Add Travis-side caching of the dotnet package that we download from Microsoft. This avoids hitting the Microsoft server every time. The download isn't super-reliable, so this should help with build reliability. This extends the cache timeout, because the package is large and the upload was timing out.
2017-10-27 03:12:21 +08:00
timeout: 600
Add cache entries to .travis.yml. This is recommended by the Travis docs to speed up the parts of the tests where unpack / install has a significant cost, as opposed to the download. This changeset adds entries for ~/.m2 (for Maven) and ~/Library/Caches/Homebrew for (brew on macOS).
2017-10-13 00:46:38 +08:00
directories:
- $HOME/.m2
Add Travis-side caching of the dotnet package that we download from Microsoft. This avoids hitting the Microsoft server every time. The download isn't super-reliable, so this should help with build reliability. This extends the cache timeout, because the package is large and the upload was timing out.
2017-10-27 03:12:21 +08:00
- $HOME/Library/Caches/Antlr4
Add cache entries to .travis.yml. This is recommended by the Travis docs to speed up the parts of the tests where unpack / install has a significant cost, as opposed to the download. This changeset adds entries for ~/.m2 (for Maven) and ~/Library/Caches/Homebrew for (brew on macOS).
2017-10-13 00:46:38 +08:00
- $HOME/Library/Caches/Homebrew
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stages:
- smoke-test
- main-test
- extended-test
Explicitly lay out the matrix for Travis. This means that we only have one run on macOS, using the default JDK. Travis does not support JDK-switching on macOS.
2016-11-07 08:42:30 +08:00
matrix:
divide and conquer
2016-11-28 02:22:18 +08:00
include:
divide and conquer
2016-11-28 00:56:14 +08:00
- os: linux
divide and conquer
2016-11-28 02:47:48 +08:00
compiler: clang
Replace oraclejdk7 with openjdk7 in .travis.yml. Oracle JDK 7 support has been withdrawn. https://github.com/travis-ci/travis-ci/issues/7964 http://www.webupd8.org/2017/06/why-oracle-java-7-and-6-installers-no.html
2017-09-29 06:48:25 +08:00
jdk: openjdk7
Revert "clean up per travis weblint" This reverts commit 49de9e8f9509c8a804610876c2522d757b821798.
2017-02-21 03:59:22 +08:00
env:
- TARGET=cpp
- CXX=g++-5
Split the C++ tests into three groups. Running these as a single test still intermittently times out after 40 minutes, even with the other recent csets.
2017-10-20 15:26:35 +08:00
- GROUP=LEXER
stage: main-test
addons:
apt:
sources:
- ubuntu-toolchain-r-test
- llvm-toolchain-precise-3.7
packages:
- g++-5
- uuid-dev
- clang-3.7
- os: linux
compiler: clang
jdk: openjdk7
env:
- TARGET=cpp
- CXX=g++-5
- GROUP=PARSER
stage: main-test
addons:
apt:
sources:
- ubuntu-toolchain-r-test
- llvm-toolchain-precise-3.7
packages:
- g++-5
- uuid-dev
- clang-3.7
- os: linux
compiler: clang
jdk: openjdk7
env:
- TARGET=cpp
- CXX=g++-5
- GROUP=RECURSION
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: main-test
divide and conquer
2016-11-28 02:50:30 +08:00
addons:
apt:
sources:
- ubuntu-toolchain-r-test
- llvm-toolchain-precise-3.7
packages:
- g++-5
- uuid-dev
- clang-3.7
divide and conquer
2016-11-28 02:44:07 +08:00
- os: osx
compiler: clang
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
env:
- TARGET=cpp
- GROUP=LEXER
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: extended-test
start with the slow builds
2016-11-28 03:04:30 +08:00
- os: osx
compiler: clang
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
env:
- TARGET=cpp
- GROUP=PARSER
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: extended-test
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
- os: osx
compiler: clang
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
env:
- TARGET=cpp
- GROUP=RECURSION
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: extended-test
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
- os: osx
compiler: clang
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
env:
- TARGET=swift
- GROUP=LEXER
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: main-test
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
- os: osx
compiler: clang
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
env:
- TARGET=swift
- GROUP=PARSER
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: main-test
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
- os: osx
compiler: clang
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
categorise tests to fix Travis timeout
2017-06-27 00:50:49 +08:00
env:
- TARGET=swift
- GROUP=RECURSION
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: main-test
adding travis configuration for swift under linux
2017-05-13 12:43:52 +08:00
- os: linux
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
dist: xenial
adding travis configuration for swift under linux
2017-05-13 12:43:52 +08:00
compiler: clang
fix failing build
2017-06-27 23:09:36 +08:00
env:
- TARGET=swift
- GROUP=ALL
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: extended-test
[dotnet] update to .netcore v1.0.1; add osx build
2017-04-04 22:26:40 +08:00
- os: osx
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
dotent also times out
2017-06-28 22:18:19 +08:00
env:
- TARGET=dotnet
- GROUP=LEXER
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: extended-test
dotent also times out
2017-06-28 22:18:19 +08:00
- os: osx
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
dotent also times out
2017-06-28 22:18:19 +08:00
env:
- TARGET=dotnet
- GROUP=PARSER
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: extended-test
dotent also times out
2017-06-28 22:18:19 +08:00
- os: osx
Move Travis tests to Swift 4.2 and Xcode 10.1. Move Travis Swift, .NET, and C++ macOS tests from Xcode 9 to Xcode 10.1. Move Travis Linux Swift tests from Trusty to Xenial and to Swift 4.2.1. Remove the dirname call when setting JAVA_HOME, it seems to be incorrect. Print the JAVA_HOME that Travis ends up with. Moving to Xcode 10.1 / Swift 4.2 exposes the issue discussed in #2372 so this cset will fail automated tests without the relevant fixes.
2018-11-08 05:11:06 +08:00
osx_image: xcode10.1
dotent also times out
2017-06-28 22:18:19 +08:00
env:
- TARGET=dotnet
- GROUP=RECURSION
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: extended-test
divide and conquer
2016-11-28 01:36:46 +08:00
- os: linux
Replace oraclejdk7 with openjdk7 in .travis.yml. Oracle JDK 7 support has been withdrawn. https://github.com/travis-ci/travis-ci/issues/7964 http://www.webupd8.org/2017/06/why-oracle-java-7-and-6-installers-no.html
2017-09-29 06:48:25 +08:00
jdk: openjdk7
divide and conquer
2016-11-28 02:22:18 +08:00
env: TARGET=java
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: extended-test
- os: linux
jdk: openjdk8
env: TARGET=java
stage: extended-test
divide and conquer
2016-11-28 02:22:18 +08:00
- os: linux
jdk: oraclejdk8
env: TARGET=java
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: smoke-test
divide and conquer
2016-11-28 01:36:46 +08:00
- os: linux
Replace oraclejdk7 with openjdk7 in .travis.yml. Oracle JDK 7 support has been withdrawn. https://github.com/travis-ci/travis-ci/issues/7964 http://www.webupd8.org/2017/06/why-oracle-java-7-and-6-installers-no.html
2017-09-29 06:48:25 +08:00
jdk: openjdk7
divide and conquer
2016-11-28 02:22:18 +08:00
env: TARGET=csharp
csharp target builds faster so use it for main test
2018-01-01 14:43:29 +08:00
stage: main-test
[csharp] update travis for dotnet
2017-01-25 08:24:36 +08:00
- os: linux
[dotnet] for linux target, bring up to jdk8 since jdk7 no longer available for Trusty; for linux and osx targets, bring up dotnet from 1.0.3 to 1.0.4; Trusty image now has mvn 3.3.9 builtin, removed install code
2017-06-26 20:18:48 +08:00
jdk: oraclejdk8
Revert "clean up per travis weblint" This reverts commit 49de9e8f9509c8a804610876c2522d757b821798.
2017-02-21 03:59:22 +08:00
dist: trusty
dotent also times out
2017-06-28 22:18:19 +08:00
env:
- TARGET=dotnet
- GROUP=LEXER
csharp target builds faster so use it for main test
2018-01-01 14:43:29 +08:00
stage: extended-test
dotent also times out
2017-06-28 22:18:19 +08:00
- os: linux
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
jdk: openjdk8
dotent also times out
2017-06-28 22:18:19 +08:00
dist: trusty
env:
- TARGET=dotnet
- GROUP=PARSER
csharp target builds faster so use it for main test
2018-01-01 14:43:29 +08:00
stage: extended-test
dotent also times out
2017-06-28 22:18:19 +08:00
- os: linux
jdk: oraclejdk8
dist: trusty
env:
- TARGET=dotnet
- GROUP=RECURSION
csharp target builds faster so use it for main test
2018-01-01 14:43:29 +08:00
stage: extended-test
divide and conquer
2016-11-28 01:36:46 +08:00
- os: linux
Replace oraclejdk7 with openjdk7 in .travis.yml. Oracle JDK 7 support has been withdrawn. https://github.com/travis-ci/travis-ci/issues/7964 http://www.webupd8.org/2017/06/why-oracle-java-7-and-6-installers-no.html
2017-09-29 06:48:25 +08:00
jdk: openjdk7
divide and conquer
2016-11-28 02:22:18 +08:00
env: TARGET=python2
more python 2 to main test stage since it’s a standalone target
2018-01-01 14:33:38 +08:00
stage: main-test
divide and conquer
2016-11-28 01:36:46 +08:00
- os: linux
Replace oraclejdk7 with openjdk7 in .travis.yml. Oracle JDK 7 support has been withdrawn. https://github.com/travis-ci/travis-ci/issues/7964 http://www.webupd8.org/2017/06/why-oracle-java-7-and-6-installers-no.html
2017-09-29 06:48:25 +08:00
jdk: openjdk7
divide and conquer
2016-11-28 02:22:18 +08:00
env: TARGET=python3
try generic python 3.5
2017-02-21 02:36:36 +08:00
addons:
fix formatting
2017-02-21 02:42:02 +08:00
apt:
sources:
- deadsnakes # source required so it finds the package definition below
packages:
update python from 3.5 to 3.6
2018-11-08 04:39:30 +08:00
- python3.6
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: main-test
divide and conquer
2016-11-28 01:36:46 +08:00
- os: linux
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
dist: trusty
jdk: openjdk8
divide and conquer
2016-11-28 02:22:18 +08:00
env: TARGET=javascript
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: main-test
divide and conquer
2016-11-28 01:36:46 +08:00
- os: linux
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
dist: trusty
jdk: openjdk8
divide and conquer
2016-11-28 02:22:18 +08:00
env: TARGET=go
Split Travis tests into three phases. This gives a "smoke-test" phase which just runs the Java tests, to get a quick check that compilation is working. It is followed by the "main-test" phase that tests each target on their preferred platform (e.g. Swift on macOS, Python on Linux) followed by "test-extended" which runs all the remaining tests (all the other Java variants, and any other platforms supported by each target). This means that those tests in the later phases won't run unless the earlier phases have passed. This should vastly improve our Travis turnaround. This changeset also changes the matrix so that we get some coverage across all of openjdk7,8 and oraclejdk7,8 and moves the Go and Javascript tests to Trusty just so that they are more up-to-date. This adds one additional Java test run, so now we compare oraclejdk8 and openjdk8 in case of problems in that regard. Note that the test matrix has not been extended to cover oraclejdk9. This currently fails for us with a build issue, so it is not included in this changeset. This changeset also removes the meaningless custom apt configuration in the section for macOS .NET tests.
2017-10-13 12:46:20 +08:00
stage: main-test
divide and conquer
2016-11-28 02:22:18 +08:00
divide and conquer
2016-11-28 00:56:14 +08:00
before_install:
Switch all the macOS Travis jobs to use the Xcode 9 image. We require Xcode 9 for the Swift target, and the C++ and .NET tests shouldn't care either way. We no longer need to install cmake: this is already in the Travis Xcode 9.0 image (and it barfs if we try to install it again). We no longer need the rvm bug workarounds; a newer version of rvm is in the Xcode 9.0 image so this is no longer a problem. The openssl package is already installed and up-to-date now, so there is no need to install that. These changes mean that we don't need before-install-osx-cpp.sh or before-install-os-swift.sh at all, and before-install-osx-dotnet.sh gets a bit shorter. .travis.yml gets a small change to handle the fact that some platform / target combos don't need before-install any more.
2017-09-29 06:40:35 +08:00
- f="./.travis/before-install-$TRAVIS_OS_NAME-$TARGET.sh"; ! [ -x "$f" ] || "$f"
Explicitly lay out the matrix for Travis. This means that we only have one run on macOS, using the default JDK. Travis does not support JDK-switching on macOS.
2016-11-07 08:42:30 +08:00
add travis ci file
2015-06-23 08:13:20 +08:00
script:
Wrap the run-tests-$TARGET.sh with a travis_wait call. This means that it will write a line to the log each minute while the tests are running, up to the specified timeout. This works around the fact that one of Maven / Surefire / JUnit are buffering our test output until a whole test class has run. This is a problem in particular with the TestLeftRecursion tests because they are taking over 9 minutes on their own. After 10 minutes of silence Travis assumes that a job is broken and kills it. We've been hitting that timeout regularly.
2017-10-13 12:41:56 +08:00
- cd runtime-testsuite; travis_wait 40 ../.travis/run-tests-$TARGET.sh