[Android-tools-devel] Bug#927676: android-platform-build: FTBFS randomly (jh_build returned exit code 123)

[Santiago Vila]

Package: src:android-platform-build
Version: 1:8.1.0+r23-2
Severity: serious
Tags: ftbfs

Dear maintainer:

I tried to build this package in buster but it failed:

--------------------------------------------------------------------------------
[...]
 debian/rules build-indep
dh build-indep --with javahelper
   dh_update_autotools_config -i
   dh_autoreconf -i
   jh_linkjars -i
   debian/rules override_dh_auto_build-indep
make[1]: Entering directory '/<<BUILDDIR>>/android-platform-build-8.1.0+r23'
jh_build --javacopts="-source 7" --no-javadoc --main=com.android.signapk.SignApk signapk.jar tools/signapk/
warning: [options] bootstrap class path not set in conjunction with -source 7
1 warning
javadoc: error - No public or protected classes found to document.
1 error
jh_build: find tools/signapk/ -name '*.java' -and -type f -print0 | xargs -s 512000 -0 /usr/lib/jvm/default-java/bin/javadoc -locale en_US -classpath /usr/share/java/bcprov.jar:/usr/share/java/bcpkix.jar:/usr/share/java/apksig.jar:debian/_jh_build.signapk -d debian/_jh_build.javadoc/api -quiet -encoding ISO8859-1 -source 1.7 -notimestamp  returned exit code 123
make[1]: *** [debian/rules:25: signapk.jar] Error 2
make[1]: Leaving directory '/<<BUILDDIR>>/android-platform-build-8.1.0+r23'
make: *** [debian/rules:31: build-indep] Error 2
dpkg-buildpackage: error: debian/rules build-indep subprocess returned exit status 2
--------------------------------------------------------------------------------

This happens randomly but the failure rate measured here is around 50%,
not something that could be called "build from source". I've tried in
lots of different systems, with different amounts of memory, with 1 CPU
and 2 CPUs, self-hosted with KVM and commercial VPS, and the
randomness always happens.

I've put a bunch of build logs here:

https://people.debian.org/~sanvila/build-logs/android-platform-build/

Unfortunately, I don't have a "recipe" to reproduce it as such, except
that I would expect this randomness to be "intrinsic" to the pàckage
and the tools used by the package, given that I can reproduce it
everywhere.

In either case, if you need a test machine to reproduce the randomness
please contact me privately and I will gladly offer ssh access.

If this is really a bug in one of the build-depends, please use
reassign and affects, so that this is still visible in the BTS web
page for this package.

BTW: I suspect of javahelper (CC to Niels because of this), and in
fact I tried this but didn't work: There are two "find" commands in
jh_build at lines 231 and 233. I tried inserting a "sort -z" between
the "find" and the "xargs" but the failures were still random, so
apparently it must be something else.

(This is the trick used by Cyril Brulebois when debugging the clojure
packages, but unfortunately does not seem to work here).

Thanks.