summaryrefslogtreecommitdiff
path: root/.travis.yml
Commit message (Collapse)AuthorAgeFilesLines
* For now, we can't use -Werror.Guy Harris2020-05-311-1/+1
| | | | | | | | | | There are a ton of "hey, that snprintf might truncate output!" warnings (yes, we know, the API has a bunch of hard length limits, so we'll have to have a different API, with mallocated message buffers, to eliminate them, and suppressing the warnings runs the risk of turning into a game of Whack-A-Mole), as well as some breakloop warnings on Linux that need an API change to add a version of pcap_breakloop() that can return an error indication.
* Build with -Werror on Travis.Guy Harris2020-05-311-1/+1
| | | | | See if there are any remaining warnings. If there are, we'll try to squash them; if not, we'll hope this keeps more from coming in.
* Try adding output to the Coverity build to see if that fixes issues.Guy Harris2020-05-121-1/+1
| | | | | | | For some reason, on the 64-bit PowerPC/Power API, z/Architecture, and 64-bit ARM buildbots, the Coverity bilds time out after the "if this is a Coverity branch, quit before doing anything" step. Print a line, to see whether that satisfies Travis.
* Pick up some changes made to the tcpdump .travis.yml.Guy Harris2020-04-181-1/+27
| | | | | | | | | | | | | | | | | | | | | | | | 1) Don't bother with "GCC" on macOS. The "gcc" on the Travis buildbots is just Apple's "gcc", which is an alternate front end to Clang, rather than being GCC-for-macOS. 2) Only do 64-bit x86 Coverity build. If we're doing a Coverity build, only do it once. Now that we're building for multiple instruction sets, we need to check the instruction set to ensure that. 3) Cut down the build matrix for macOS. macOS currently runs only on 64-bit x86, so eliminate all the other architecture builds for it. (It may run on 64-bit ARM in the future, but is unlikely ever to run on z/Architecture, and probably won't be re-introduced to the world of the Power ISA, much less little-endian Power ISA.)
* Travis CI: Remove the useless package libusb-1.0-0-devFrancois-Xavier Le Bail2020-04-161-1/+0
| | | | It is not needed for usbmon USB capture support on Linux.
* Travis CI: Replace libdbus-glib-1-dev by libdbus-1-devFrancois-Xavier Le Bail2020-04-131-1/+1
| | | | | | The workaround for 'Disallowing packages: libdbus-1-dev' is now useless. This reverts commit b8056c38e7512a6282f2e8806cead97042b2e8da.
* Travis CI: Add more CPU architecturesFrancois-Xavier Le Bail2020-04-071-0/+6
| | | | | | Add ppc64le, s390x and arm64 architectures. https://docs.travis-ci.com/user/multi-cpu-architectures/
* Travis CI: Update printing the list of devel libs in Linux buildsFrancois-Xavier Le Bail2020-02-251-1/+1
| | | | This change improves readability.
* Travis CI: Revert the workaround for Coverity certificate problemFrancois-Xavier Le Bail2020-02-111-2/+0
| | | | It seems no longer needed.
* Xcode 11 currently blows up trying to build for IA-32.Guy Harris2019-09-081-1/+0
| | | | | I need to further tweak the configure script to detect the failure on Mojave/Xcode 11.
* While we're updating build environments, let's go with Xcode 11.Guy Harris2019-09-081-0/+1
|
* Travis CI: Use Linux Bionic (Ubuntu 18.04)Francois-Xavier Le Bail2019-09-071-1/+1
| | | | gcc 7.4.0 (on Xenial was 5.4.0)
* Travis CI: Workaround for the Coverity certificate problemFrancois-Xavier Le Bail2019-04-261-0/+2
| | | | | The error message was: "curl: (60) server certificate verification failed."
* Travis CI: Remove "sudo: false"Francois-Xavier Le Bail2019-04-241-1/+0
| | | | | | | From Travis CI documentation: Container-based infrastructure has been fully deprecated. Please remove any "sudo: false" keys in your .travis.yml file to use the fully-virtualized Linux infrastructure.
* Travis CI: Use Linux XenialFrancois-Xavier Le Bail2019-03-291-0/+2
|
* Travis CI: Use ccache and parallel make.Denis Ovsienko2019-03-221-0/+3
|
* Travis CI: Add debug infosFrancois-Xavier Le Bail2019-03-211-0/+11
| | | | | | Add a folded 'cat Makefile' (without the mkdep part for autotools). Add a folded 'cat config.h'. Add a folded 'cat config.log' (for autotools).
* Travis CI: Print the devel libs in Linux buildsFrancois-Xavier Le Bail2019-03-151-0/+1
|
* Travis CI: Fix the problem with folded outputsFrancois-Xavier Le Bail2019-03-151-3/+3
| | | | | | Trailing newlines are now needed in travis_fold:start:* strings, because of a Travis change. Based on Travis support proposal.
* Travis CI: Restore pre-DPDK tests setupFrancois-Xavier Le Bail2019-03-151-15/+0
| | | | | | | | | With this change the apt-get command installs again theses libraries: libusb-1.0-0-dev libdbus-glib-1-dev libbluetooth-dev libnl-genl-3-dev libibverbs-dev
* Try adding [trusted=yes] to the sourceline: entry for jessie-backports.Guy Harris2019-02-101-2/+12
| | | | | Thanks and a tip of the Hatlo hat to Anders Broman for the suggestion to use a Debian repository and for this suggestion.
* Forget jessie-backports.Guy Harris2019-02-091-12/+2
| | | | | | | | | Either it or Travis or both appear to require a GPG key to verify the packages, and Debian appears not to make it easy to determine what the URL is to get it (or don't make it available via HTTP). Thanks, fellas. I really appreciate how easy you made it to test building libpcap with a sufficently-modern DPDK on Travis.
* Try pulling libdpdk-dev from the Devian jessie-backports APT repository.Guy Harris2019-02-091-2/+12
|
* Don't bother with Xenial.Guy Harris2019-02-081-262/+26
| | | | | | | | The only reason for trying it was to try building DPDK, and the version of DPDK that comes with it is too old to work with libpcap, so just go back to the old matrix for now. (We may want to revive it to test with newer compilers on Linux and to test with real GCC, not "clang is claiming to be GCC", on macOS.)
* OK, go back to the explicit matrix, and fix package installation.Guy Harris2019-02-081-29/+261
| | | | | | | | | | | It appears that having multiple distributions in dist: causes the build to hang forever in the queueing stage. Go back to the explicit matrix as a way to get both Trusty and Xenial builds. If you have an item in some matrix entries and the same item outside the matrix, the one in the matrix entry overrides the item outside the matrix, so we have to have the APT add-on packages in each matrix entry. Do so.
* OK, without dist: it finished quickly; try putting it back again.Guy Harris2019-02-081-3/+3
|
* Doing the dist: stuff hung forever queueing. Try removing that.Guy Harris2019-02-081-3/+3
|
* Try going back to the old matrix style, but with multiple Ubuntu versions.Guy Harris2019-02-081-106/+17
| | | | | | | That's a bit simpler. Hopefully, that won't cause Travis to get confused because we're building on macOS as well as Ubuntu.
* Will this install dpdk-dev on Xenial but not Trusty?Guy Harris2019-02-081-0/+8
|
* macOS is called "osx", not "macos".Guy Harris2019-02-081-6/+17
| | | | | Add some comments, and get rid of some no-longer-needed stuff, while we're at it.
* Maybe the problem was that we had two "matrix:" lines.Guy Harris2019-02-081-5/+73
|
* OK, shrink it to a tiny matrix and see if that runs *two* jobs, not *one*.Guy Harris2019-02-081-12/+17
|
* OK, revert for now, but dump out the environment before install.Guy Harris2019-02-081-87/+12
| | | | | | | I *suspect* that the custom matrix isn't setting custom environment variables, so CMAKE is neither "yes" nor "no", and we are neither running the configure script nor CMake. See what happens *without* the custom matrix, so we can compare them.
* OK, try *this*.Guy Harris2019-02-081-96/+76
| | | | | | | | | | | The syntax, including the indentation, is taken from a .travis.yml file for a GitHub repository for a project being maintained by "Twisted Matrix Labs", who are apparently responsible for some obscure product called "Twisted". :-) Presumably *that* works. The only things they're *not* doing are 1) tweaking the distribution and compiler and 2) setting multiple environment variables. If Travis can't parse *that*, what's the issue?
* OK, put the &&s back.Guy Harris2019-02-081-20/+20
|
* OK, try undoing the indentation change.Guy Harris2019-02-081-96/+96
|
* Travis says it can't parse that; try removing the &&s, for starters.Guy Harris2019-02-081-20/+20
|
* Fix indentation, put && between environment setting in MATRIX_EVAL.Guy Harris2019-02-081-96/+96
| | | | That's what the examples in the Travis documentation do.
* OK, dump out the environment after the eval to see whether it worked.Guy Harris2019-02-081-0/+1
|
* Maybe *this* is needed for custom matrices.Guy Harris2019-02-081-0/+1
| | | | That's something in a Travis CI documentation page.
* Try something different for setting the environment.Guy Harris2019-02-081-20/+40
| | | | This is what some Travis documentation suggests; see if that works.
* Try building with Xenial as well as Trusty, and doing fewer macOS builds.Guy Harris2019-02-081-12/+86
| | | | | | | | | | | Try building with both Xenial and Trusty on Ubuntu, so we can try testing builds with DPDK, which is only available on Xenial. Don't build with both gcc and clang on macOS for now; according to the Trusty documentation, gcc is just an alias for clang, and to really compile with *GCC*, you'd have to set CC in the environment to a specific GCC version, and you'd have to use "brew install" to get versions other than 4.9.
* We'd have to use Xenial to test DPDK.Guy Harris2019-01-061-1/+4
|
* Try adding the DPDK developer package to test the DPDK support.Guy Harris2019-01-061-0/+1
|
* update configure and cmakejingleyang2018-12-241-0/+1
|
* return to default .travis.ymljingleyang2018-12-181-20/+10
|
* add cache for travisjingleyang2018-12-171-3/+5
|
* reduce travis.ymljingleyang2018-12-171-4/+2
|
* make -s all for dpdk compilejingleyang2018-12-171-5/+5
|
* enable DPDK for travisjingleyang2018-12-171-2/+5
|