IcedTea6 build failed for b41a73ca64e6 (--enable-nss)

Andrew Hughes ahughes at redhat.com
Thu Aug 16 16:45:20 PDT 2012



----- Original Message -----
> 
> 
> ----- Original Message -----
> > On Thu, 2012-08-16 at 07:02 -0400, Andrew Hughes wrote:
> > > > Do you know what the intention is, which file should be called
> > > > what
> > > > and be under which directory?
> > > 
> > > Something about your build is borked.
> > 
> > Could you try to minimize the usage of "your". For some reason it
> > irks
> > me seeing that usage when I merrily point out that one of the
> > standard
> > autobuilders is failing. The autobuilders are there precisely
> > because
> > nobody can test all the arches, setups and configuration options we
> > support. It is "our" build that is borked.
> 
> And the autobuilders I've setup, as well as my own builds, pass.
> Whatever setup is used here (which you didn't include in the original
> e-mail, hence why I'm stabbing in the dark) is not building a
> complete
> JDK.
> 

Sorry if I've been a little blunt on this, but it feels like I've been hounded
a bit on what is a very low priority issue here.  What's the rush?

This is especially true with this issue which turns out to be because
of a non-standard build option I asked you to change on the autobuilders
months ago.

> >From your other e-mail, it sounds like you're using the add-zero
> rule which, as mentioned before, is not supported.
> 
> > 
> > > It looks like your build has always been broken, but has only now
> > > come to light due to recent changes.
> > > Nothing new has been added.  The patching of java.security has
> > > merely been moved from before the build
> > > to afterwards, so that the build itself doesn't have an PKCS11
> > > provider configured with a missing
> > > configuration file.
> > 
> > Aha. Now I understand. That assumes that there is a full build
> > already
> > in place. That isn't always the case when we build zero
> > additionally.
> 
> No, this is a correct assumption.  The IcedTea target is meant to
> build a full
> JDK and this is what add-nss depends on.
> 
> > Then we "fake" a whole build and do a recursive make. See the
> > add-zero.stamp target:
> > 
> >     : # create directories which are expected by icedtea
> >     components,
> >     : # implicitly assumed to be created by the jdk build.
> > 
> 
> I don't agree with this fakery.  It makes the whole build much harder
> to maintain.

Looking at it briefly, the add-zero rule is clearly at fault.  If it just
wants to build HotSpot, it should be calling make hotspot, which would
just run the OpenJDK part of the build and not the add-x targets designed
for a full JDK.

However, my preferred solution would be to just remove these superfluous targets
altogether as they cause more trouble than they're worth.
-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/)
Fingerprint = EC5A 1F5E C0AD 1D15 8F1F  8F91 3B96 A578 248B DC07




More information about the distro-pkg-dev mailing list