|| ||"Justin M. Forbes" <jmforbes-AT-rpath.com>|
|| ||rPUA-2006-0107-1 conary conary-build conary-policy
|| ||Fri, 16 Jun 2006 17:41:01 -0400|
rPath Update Advisory: 2006-0107-1
Products: rPath Linux 1
Conary 1.0.20 is a maintenance release.
o jvmdir, javadir, javadocdir, and thisjavadocdir have been added
to the default set of macros.
o An r.JavaDoc() build action has been added. It functions
exactly like the r.Doc() build action, except it copies into
o When the r.addArchive() source action attempts to guess the
main directory in which to build, it now recognizes when
sources have been added in an order that defeats its algorithm,
and provides a helpful error message. (CNY-557)
o A --tag-script parameter has been added to the rollback
o A bug in clone behavior that caused Conary to try cloning
excessive troves has been fixed. If you were cloning "uphill"
from branch /a/b to branch /a, and a buildreq was satisfied by
a trove on /a//c, Conary would try to clone the buildreq to /a
as well. (CNY-499)
o A bug in the r.Ant() and r.JavaCompile() build actions which
caused the CLASSPATH variable to get mangled has been fixed.
o A bug in r.ClassPath() that caused a traceback has been fixed.
o A bug that caused the change password tab to be displayed when
browsing the repository via the web as an anonymous user has
o The web service code no longer permits the anonymous user to
view the change password (/conary/chPassForm) form.
o The r.addPatch() source action no longer hangs when presented
with large patches, which fixes a bug introduced in Conary 1.0.19.
Conary Policy: The conary-policy package has been updated, and
conary-policy 1.0.6 has been released concurrently.
o CheckSonames no longer suggests calling r.Ldconfig, because the
NormalizeLibrarySymlinks policy will automatically take care of
every case where CheckSonames could have suggested calling
o CheckSonames will warn about shared libraries that have a
SONAME that matches the name of an actual file instead of a
symlink only once, and it will print a more helpful message as
information rather than as a warning.
o A bug that caused the new interpreter checking behavior in
EnforceBuildRequirements to traceback in many situations has
to post comments)