Repository navigation
test failures on Android #1430
Description
Activity
Tomorrow I am going to run the tests and check if I can patch some tests. I am seeing that the build of iojs is not working on
armv7ldevices, I am getting this onarmv7ldevices:assertion "(err) == (0)" failed: file "../src/node.cc", line 3481, function "void node::PlatformInit()" [1] + Stopped (signal) ./iojs@shigeki could you give the result of
uname -aon the device you tested? I want to identify on whatarmv7is working and on which one is not.Added to it, I want to work on fix the way
configureis detecting the macros of compiler binaries from Android NDK, but maybe that should be discussed in other issues after solve this first.I suspect that's because of the pthread_sigmask() in PlatformInit; IIRC, in Android it was broken until Jellybean. Can you try this patch?
diff --git a/src/node.cc b/src/node.cc index 67cf140..f32a1a0 100644 --- a/src/node.cc +++ b/src/node.cc @@ -3463,7 +3463,13 @@ inline void PlatformInit() { sigset_t sigmask; sigemptyset(&sigmask); sigaddset(&sigmask, SIGUSR1); - const int err = pthread_sigmask(SIG_SETMASK, &sigmask, nullptr); + int err = pthread_sigmask(SIG_SETMASK, &sigmask, nullptr); + +#ifdef __ANDROID__ + // Android < 4.1 has a broken pthread_sigmask() implementation. + if (err == EINVAL) + err = sigprocmask(SIG_SETMASK, &sigmask, nullptr); +#endif // Make sure file descriptors 0-2 are valid before we start logging anything. for (int fd = STDIN_FILENO; fd <= STDERR_FILENO; fd += 1) {
@Gioyik I never had such an error on my Android (Lolipop) so probably it comes from the issue that is pointed out by Ben. My device's output is
root@mako:/ # uname -a Linux localhost 3.4.0-perf-g16e203d #1 SMP PREEMPT Wed Nov 19 05:47:52 UTC 2014 armv7l GNU/Linux root@mako:/ # grep -e ro.product.cpu -e ro.build.version /system/build.prop
and build.prop is in https://gist.github.com/shigeki/9a5db4c9ea0e3da978fd .
I made tentative fixes to pass all the tests on my Android in https://github.com/shigeki/io.js/tree/android-fix .
Please test it if it works on your device.Is it not-onerous to get an update on this? I don't suppose it's been quietly fixed in the last nine months or something?
I don't have an android device to run tests on Android now. So I close this.
As commented in #1396 (comment) , several tests fails on Android.
Some of them can be tentatively fixed by a small patch but 18 failures still remains.
The patch and full test results( --mode=release message parallel sequential) are show in
https://gist.github.com/shigeki/fd25f15657f058ceccea and the list of failures are
$ grep 'not ok' test_out.txt not ok 16 - test-child-process-default-options.js not ok 20 - test-child-process-env.js not ok 23 - test-child-process-exec-env.js not ok 59 - test-cluster-bind-privileged-port.js not ok 87 - test-cluster-shared-handle-bind-privileged-port.js not ok 202 - test-fs-access.js not ok 229 - test-fs-readfile-pipe.js not ok 230 - test-fs-readfile-pipe-large.js not ok 233 - test-fs-readfilesync-pipe-large.js not ok 262 - test-http-304.js not ok 307 - test-http-curl-chunk-problem.js not ok 321 - test-http-full-response.js not ok 415 - test-https-simple.js not ok 477 - test-net-pipe-connect-errors.js not ok 514 - test-process-config.js not ok 577 - test-stdio-closed.js not ok 779 - test-child-process-execsync.js not ok 785 - test-fs-watch.js4 failures are because process.getuid is
undefinedon Android and some of them can be fixed by changing to invoke external commands. It needs more investigations.CC: @Gioyik