Open robert-schmidtke opened 5 years ago
I realized someone might be worried about the authenticity of the builds, so here is what you can do. I'm going to use https://github.com/paciofs/multichain/releases/tag/2.0-dev-20190123T153408 as an example.
$ wget https://github.com/paciofs/multichain/releases/download/2.0-dev-20190123T153408/multichain-2.0-dev-42e424a-osx.tar.gz
$ shasum -a 256 multichain-2.0-dev-42e424a-osx.tar.gz | cut -f 1 -d ' '
93a5db98282672ba6a83c5bfdf0b93e9edc2aed49f9c54446670a10d6ee3751f
Hopefully this convinces you that I have not uploaded a forged version from my laptop. The build log, script and checkout is all public, so you can trace every step. Finally you can compare the branches across repositories: https://github.com/MultiChain/multichain/compare/2.0-dev...paciofs:2.0-dev and you should find only differences related to building MultiChain on Travis.
@robert-schmidtke I was wondering if you ever managed to get Travis CI working on master/2.0-release? I tried copying your configurations file but the second stage always fail to me.
+cd ./v8
+export RELEASE=out.gn/x64.release
+RELEASE=out.gn/x64.release
+ninja -C out.gn/x64.release antlr4 icu inspector torque v8_libbase v8_libplatform v8_libsampler
ninja: Entering directory `out.gn/x64.release'
[6/672] CXX obj/v8_libbase/file-utils.o
FAILED: obj/v8_libbase/file-utils.o
....[CASCADE OF ERRORS]....
6 errors generated.
[8/672] CXX obj/v8_libbase/ieee754.o
ninja: build stopped: subcommand failed.
The command "./.travis/build-v8.sh antlr4 icu inspector torque v8_libbase v8_libplatform v8_libsampler" failed and exited with 1 during .
I saw in your Travis CI that yours is working, but they skipped building v8 because it was cached. I wonder if something has changed, even though you checkout a specific branch? Perhaps do you have any clues on that?
In addition to what I described in the initial post, I also added the 2.0-release branch (see https://github.com/paciofs/multichain/releases). I have not tried building the master branch. I checked out the version of V8 as suggested in https://github.com/MultiChain/multichain/blob/master/V8_mac.md and then built it accordingly. In the meantime, the developers have provided a prebuilt version of V8 for Mac: https://github.com/MultiChain/multichain-binaries/raw/master/macos-v8.tar.gz Maybe you could try that instead? I don't think something has changed in V8, the last time it was built in Travis is here: https://travis-ci.org/paciofs/multichain/builds/502041701 Maybe you can compare the logs/environment/configuration?
I have not tried building the master branch
Apparently it should be no issue, as the difference between the 2.0-release and master right now are just the travis files.
Mine fails in the second stage for some reason that I can't explain. I checked your logs line by line and they match exactly until... they don't.
Summary:
ninja: Entering directory `out.gn/x64.release'
[6/672] CXX obj/v8_libbase/file-utils.o
FAILED: obj/v8_libbase/file-utils.o
6 errors generated.
[7/672] CXX obj/v8_libbase/logging.o
FAILED: obj/v8_libbase/logging.o
6 errors generated.
[8/672] CXX obj/v8_libbase/ieee754.o
ninja: build stopped: subcommand failed.
The command "./.travis/build-v8.sh antlr4 icu inspector torque v8_libbase v8_libplatform v8_libsampler" failed and exited with 1 during .
Which is absolutely bonkers because the repository wast a match by match with yours, with the exception of pointing to my own multichain repository instead of here, but the error was before all that.
Absolutely strange to me this... I'm trying again, let's see.
You could try checking out a specific version of depot_tools, as suggested here: https://github.com/MultiChain/multichain/blob/master/V8_mac.md I do not have this step in my scripts. It appears that there is some difference between the checkouts, as your log says to build 672 targets, while mine only builds 667. V8 and the other sources should be the same. I'll kick off a build for master myself.
Nice catch with suggestion! Running build right now
Ok so my build of master fails with the same error as yours. Let me know whether checking out depot_tools properly works.
Failed [477/672] CXX obj/v8_libbase/file-utils.o
I'm trying again but without the --depth=1.
If it doesn't work then I'll probably be using the compiled V8 already.
@robert-schmidtke
You can delete the build caches in Travis.
No luck from my side. Tried with the checkout, without depth and deleting cache.
So the thing that appears to do the trick was to ignore the extra semicolon warning as well (first stage is succeeding now): https://github.com/paciofs/multichain/commit/a8025cb916f74bd20d7cc0697c6dee155d717a68 Cannot be bothered figuring out what was going wrong here.
Yup: https://travis-ci.org/paciofs/multichain/builds/523862865
Hi everyone,
this is not an issue as such, but since I haven't found a mailing list, here it goes. I have forked MultiChain and added Travis CI scripts for the master, 1.0.x-dev, 1.0.x-release, 2.0-dev and 2.0-release branches that build macOS High Sierra binaries (macOS 10.13.6 as of today) and release them to GitHub, whenever I merge your
upstream
changes and push them. The builds are "versioned" with the latestupstream
branch merge commit date (UTC). They do not include a properREADME.txt
, which you distribute with your pre-compiled packages, as I could not find them in your repository (I only include aREADME
with a reference to the proper upstream tree). However they do contain themultichain-cli
,multichain-util
,multichaind
andmultichaind-cold
binaries.If you find this useful, I can open a PR, otherwise I just wanted to put it out here, in case someone does not want to go through the trouble of compiling. I have too many different versions of
boost
andopenssl
installed (viabrew
, self-built or provided binaries) on my machine to compile the thing properly (i.e. weird pointer corruption errors due to linking of different versions, failed assertions and whatnot). The binaries mentioned above run fine on my macOS Mojave 10.14.3 as well, and the very simple test case included in the.travis.yml
scripts succeeds on High Sierra 10.13.6 on Travis.As a disclaimer I'll have to say that I need MultiChain as part of a project we are working on, and as such I cannot say I'll be able to maintain this fork, at least on a regular basis (or at all).
Cheers and keep up the great work! Robert