Bug 205751

Summary: java/openjdk8: Patch breaks CCACHE
Product: Ports & Packages Reporter: Christian Ullrich <chris>
Component: Individual Port(s)Assignee: Greg Lewis <glewis>
Status: Closed FIXED    
Severity: Affects Some People CC: glewis
Priority: --- Keywords: easy, needs-qa, patch, patch-ready
Version: LatestFlags: bugzilla: maintainer-feedback? (java)
koobs: merge-quarterly?
Hardware: Any   
OS: Any   
URL: https://svnweb.freebsd.org/changeset/ports/381360
Attachments:
Description Flags
Patch to java/openjdk8/files/patch-bsd
none
Revised patch to java/openjdk8/files/patch-bsd none

Description Christian Ullrich 2015-12-31 07:59:01 UTC
Created attachment 164891 [details]
Patch to java/openjdk8/files/patch-bsd

There is a bug in one of the FreeBSD patches to the java/openjdk8 port, introduced in r381360, that causes the ccache check to always fail. I think the variable reference is swallowed whole by m4, but it's autoconf, so who knows what is really going on.

Attached is a patch to the patch.
Comment 1 Christian Ullrich 2015-12-31 08:09:47 UTC
Created attachment 164892 [details]
Revised patch to java/openjdk8/files/patch-bsd
Comment 2 Kubilay Kocak freebsd_committer freebsd_triage 2015-12-31 14:17:28 UTC
Thanks you Chris :)

CC original committer
Comment 3 Greg Lewis freebsd_committer freebsd_triage 2016-01-01 22:54:14 UTC
patch-bsd is autogenerated, so your patch will just get overwritten next time it gets generated.

My preference is to commit your patch to the main OpenJDK repository and then regenerate patch-bsd.  I'd really encourage that, but it will require that you have signed the contributor agreement for OpenJDK.

The alternative is to generate a separate patch just for the FreeBSD ports repository.
Comment 4 Christian Ullrich 2016-01-01 23:20:25 UTC
I had not signed the CA before; I have now, but instead of waiting for however long it takes Oracle to process the thing, how about this: Forget my patch and just consider this hypothetical situation:

- There is a bug in the steaming heap of autoconf entropy underlying the
  configure script.
- It manifests by losing a reference to a positional argument when the
  script is generated.
- m4 has a way to write dollar signs, @S|@, that circumvents this issue.

Now, under our hypothetical scenario, do you think you might be able to fix such a bug yourself? I would be very surprised, of course, if it turned out that something like this actually existed and could be fixed in this manner.
Comment 5 Christian Ullrich 2016-01-12 08:45:58 UTC
FWIW, the OCA has gone through now.
Comment 6 commit-hook freebsd_committer freebsd_triage 2016-01-21 05:26:24 UTC
A commit references this bug:

Author: glewis
Date: Thu Jan 21 05:25:44 UTC 2016
New revision: 406824
URL: https://svnweb.freebsd.org/changeset/ports/406824

Log:
  . Unreak CCACHE

  PR:		205751
  Submitted by:	chris@chrullrich.net

Changes:
  head/java/openjdk8/files/patch-common-autoconf-build-performance.m4
  head/java/openjdk8/files/patch-common-autoconf-generated-configure.sh
Comment 7 Greg Lewis freebsd_committer freebsd_triage 2016-01-21 05:27:55 UTC
Thanks Chris!  I've committed your changes to the bsd-port repository for openjdk8 and to the FreeBSD ports repo (as separate patches there until I next regenerate patch-bsd).