Strange Java package build failures (Error in <unknown> scriptlet ?)

[Fabio Valentini]

Hi everybody,

Since ~2 days ago, the rawhide koji buildroot has exhibited some
*really weird* issues when building Java packages with maven (when
building packages locally with "mock -r fedora-rawhide-x86_64
--enablerepo local foo.src.rpm").

During installation of the build dependencies, some random (?) Java
package will have a scriptlet failure like this: "Error in <unknown>
scriptlet in rpm package mockito" (the package that this error occurs
in is not always the same). This is always the last scriptlet run
before the "Verifying" stage of "dnf install".

Then, later, during execution of %build, the following errors show up,
which make the it fail:

/usr/share/maven/bin/mvn: Failed to set JAVACMD
The JAVA_HOME environment variable is not defined correctly
This environment variable is needed to run this program
NB: JAVA_HOME should point to a JDK not a JRE

Sometimes, scrubbing the mock buildroot makes the next build succeed,
sometimes it doesn't. I tried entering the buildroot with "mock shell"
but it didn't yield any useful information.

I see that the scripts setting up JAVA_HOME etc. try to find a java +
javac executable, but they're all there, and the alternatives setup
also looks sane (no broken links in /etc/alternatives).

The only Java related updates I see in the buildroot for the "critical
time frame" are builds of java-11-openjdk, but I'm not sure how they
could have introduced such inconsistently buggy buildroot behaviour :(

Does anybody have an idea what might be the problem here? It's really
annoying to be unable to build Java packages locally most of the time
(curiously, I haven't seen this issue happen in any koji builds yet).

Fabio

[Ondrej Dubaj]

On Fri, Aug 14, 2020 at 7:07 PM Fabio Valentini 
wrote:
> Hi everybody,
>
> Since ~2 days ago, the rawhide koji buildroot has exhibited some
> *really weird* issues when building Java packages with maven (when
> building packages locally with "mock -r fedora-rawhide-x86_64
> --enablerepo local foo.src.rpm").
>
> During installation of the build dependencies, some random (?) Java
> package will have a scriptlet failure like this: "Error in <unknown>
> scriptlet in rpm package mockito" (the package that this error occurs
> in is not always the same). This is always the last scriptlet run
> before the "Verifying" stage of "dnf install".
>
> Then, later, during execution of %build, the following errors show up,
> which make the it fail:
> /usr/share/maven/bin/mvn: Failed to set JAVACMD
> The JAVA_HOME environment variable is not defined correctly
> This environment variable is needed to run this program
> NB: JAVA_HOME should point to a JDK not a JRE

Experiencing the same problem. Haven;t found any usefull info about this
> Sometimes, scrubbing the mock buildroot makes the next build succeed,
> sometimes it doesn't. I tried entering the buildroot with "mock shell"
> but it didn't yield any useful information.
>
> I see that the scripts setting up JAVA_HOME etc. try to find a java +
> javac executable, but they're all there, and the alternatives setup
> also looks sane (no broken links in /etc/alternatives).
>
> The only Java related updates I see in the buildroot for the "critical
> time frame" are builds of java-11-openjdk, but I'm not sure how they
> could have introduced such inconsistently buggy buildroot behaviour :(
>
> Does anybody have an idea what might be the problem here? It's really
> annoying to be unable to build Java packages locally most of the time
> (curiously, I haven't seen this issue happen in any koji builds yet).
>
> Fabio
> _______________________________________________
> devel mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct:
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail]

[Fabio Valentini]

On Mon, Aug 17, 2020 at 7:14 AM Ondrej Dubaj  wrote:
> Experiencing the same problem. Haven;t found any usefull info about this

We have since identified the root cause of this problem:
https://bugzilla.redhat.com/show_bug.cgi?id=1869030

If you're doing local mock builds, you can work around the problem by
running mock with "--isolation=simple".

For COPR builds, it's not that simple (no pun intended), but I built
un-broken glibc in a COPR that you can add for rawhide x86_64 builds:
https://copr.fedorainfracloud.org/coprs/decathorpe/unbroken-glibc-2.33/
Just adding "copr://decathorpe/unbroken-glibc-2.33" as an external
repo should fix things for COPR builds until glibc is fixed in rawhide
(or until systemd is patched on all fedora releases and on COPR
servers etc.)

Fabio