Bug 236091 - java/openjdk8: crash when building FreeBSD docs on powerpc64
Summary: java/openjdk8: crash when building FreeBSD docs on powerpc64
Status: New
Alias: None
Product: Ports & Packages
Classification: Unclassified
Component: Individual Port(s) (show other bugs)
Version: Latest
Hardware: Any Any
: --- Affects Only Me
Assignee: freebsd-java mailing list
URL:
Keywords:
: 236349 (view as bug list)
Depends on:
Blocks:
 
Reported: 2019-02-27 22:04 UTC by Steve Wills
Modified: 2019-11-06 19:36 UTC (History)
2 users (show)

See Also:
bugzilla: maintainer-feedback? (java)


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Steve Wills freebsd_committer 2019-02-27 22:04:45 UTC
See this email:

https://lists.freebsd.org/pipermail/freebsd-pkg-fallout/Week-of-Mon-20190225/1137487.html

and please let me know if more information or access for troubleshooting or debugging is needed.
Comment 1 Greg Lewis freebsd_committer 2019-08-26 06:00:25 UTC
*** Bug 236349 has been marked as a duplicate of this bug. ***
Comment 2 Greg Lewis freebsd_committer 2019-11-06 18:25:03 UTC
Is this still happening?  I can't find a package build with the latest openjdk8 on powerpc64 12.x log to check.
Comment 3 Piotr Kubaj freebsd_committer 2019-11-06 18:44:01 UTC
(In reply to Greg Lewis from comment #2)
Yes it still happens. Check e.g. latest build on http://pylon.nyi.freebsd.org/build.html?mastername=head-powerpc64-default&build=p515672_s354110:
28	es-freebsd-doc-53465,1	misc/freebsd-doc-es	build	1	new_compiler_error	00:04:04
29	de-freebsd-doc-53465,1	misc/freebsd-doc-de	build	0	new_compiler_error	00:03:59
30	it-freebsd-doc-53465,1	misc/freebsd-doc-it	build	0	new_compiler_error	01:25:46
33	pl-freebsd-doc-53465,1	misc/freebsd-doc-pl	build	0	new_compiler_error	00:04:03
34	el-freebsd-doc-53465,1	misc/freebsd-doc-el	build	0	new_compiler_error	00:04:02
37	ru-freebsd-doc-53465,1	misc/freebsd-doc-ru	build	0	new_compiler_error	00:04:47
39	en-freebsd-doc-53465,1	misc/freebsd-doc-en	build	0	new_compiler_error	00:04:11
53	tr-freebsd-doc-53465,1	misc/freebsd-doc-tr	build	0	new_compiler_error	00:07:01
54	fr-freebsd-doc-53465,1	misc/freebsd-doc-fr	build	0	new_compiler_error	00:04:58
59	mn-freebsd-doc-53465,1	misc/freebsd-doc-mn	build	0	new_compiler_error	04:12:04
63	nl-freebsd-doc-53465,1	misc/freebsd-doc-nl	build	0	new_compiler_error	00:04:30
65	da-freebsd-doc-53465,1	misc/freebsd-doc-da	build	0	new_compiler_error	00:04:34
66	bn-freebsd-doc-53465,1	misc/freebsd-doc-bn	build	0	new_compiler_error	00:04:38

Interestingly, this seems to be an issue that pops up randomly. There are doc packages that built:
12656	ko-freebsd-doc-53465,1	misc/freebsd-doc-ko	success	00:06:14
12739	ja-freebsd-doc-53465,1	misc/freebsd-doc-ja	success	01:39:03
13921	zh_tw-freebsd-doc-53465,1	misc/freebsd-doc-zh_tw	success	02:07:21
14338	hu-freebsd-doc-53465,1	misc/freebsd-doc-hu	success	03:57:45
14870	pt-freebsd-doc-53465,1	misc/freebsd-doc-pt	success	03:09:45
15020	zh_cn-freebsd-doc-53465,1	misc/freebsd-doc-zh_cn	success	04:18:27
Comment 4 Greg Lewis freebsd_committer 2019-11-06 19:36:38 UTC
Looking at

http://pylon.nyi.freebsd.org/data/head-powerpc64-default/p515672_s354110/logs/errors/en-freebsd-doc-53465,1.log

I see that the JVM no longer crashes.  It instead throws StackOverflowError.  In other words, the crash was a manifestation of bug 222146.  Now that has been fixed it looks like the underlying cause is an infinite recursion bug in textproc/fop.

I can see from glancing at TernaryTree.java that this is possible since insert() does recursively call itself.  I've done no in depth analysis to understand why it would only happen for certain languages or on certain architectures (although powerpc64 being big endian might be a factor).

Instead of an in depth debugging session, it might be worthwhile to try updating textproc/fop to 2.4 and seeing if that fixes it.  You could also potentially see if running with openjdk11 rather than openjdk8 made a difference.

At the moment this seems like the point of investigation is textproc/fop rather than java/openjdk8 though.