Debian Shibboleth install instructions
wferi at niif.hu
Thu May 15 22:12:18 UTC 2008
Russ Allbery <rra at debian.org> writes:
> Ferenc Wagner <wferi at niif.hu> writes:
>> How is this going?
> There's a first draft of general Git instructions at:
> I haven't had a chance to start working on the wiki page yet.
Hmm, what wiki page? Your draft is very useful, if only I could have
read that a week ago!
> (We had a major work project rollout at 5 this morning.)
That explains why you were around so early today...
>> I'm struggling with this stuff, although some pieces already fell into
>> place. What hurts most ATM is that git-buildpackage doesn't want to
>> work without --git-ignore-new:
> That's correct; you have to always use --git-ignore-new because we're not
> committing the new versions of config.guess and config.sub to the
> repository. We could do that instead, but I think it's messy; I prefer to
> depend on autotools-dev and update them automatically during the build,
> since then only binNMUs are required when Debian adds new architectures.
> Do we also have a patch for configure? If not, we don't need to run
> autoconf during the build, which will help with some of that.
I took the examples from your xml-security-c package. I don't know if
you really have to run autoconf before configuring.
>> # deleted: include/xsec/canon/XSECC14n20010315.hpp
>> # deleted: include/xsec/canon/XSECCanon.hpp
> Are these files shipped by upstream but deleted by make clean? If so,
> then this too will force always using --git-ignore-new.
Yes, they are. However, they aren't removed by make mostlyclean.
Extracting the tarball, configuring and making mostlyclean results in
$ tar -dzf xml-security-c-1.4.0.tar.gz | egrep -v "([UG]id|Mode) differs$"
xml-security-c-1.4.0/src/framework/XSECConfig.hpp: Mod time differs
xml-security-c-1.4.0/src/framework/XSECConfig.hpp: Size differs
and the diff is:
--- XSECConfig.hpp 2007-10-31 19:29:06.000000000 +0100
+++ /tmp/XSECConfig.hpp 2008-05-15 23:46:28.245662464 +0200
@@ -45 +45 @@
-#define XSEC_NO_XALAN 1
+/* #undef XSEC_NO_XALAN */
@@ -92 +92 @@
-/* #undef XSEC_XERCESPARSERLIAISON_REQS_DOMSUPPORT */
+#define XSEC_XERCESPARSERLIAISON_REQS_DOMSUPPORT 1
@@ -98 +98 @@
-/* #undef XSEC_SELECTNODELIST_REQS_NODEREFLIST */
+#define XSEC_SELECTNODELIST_REQS_NODEREFLIST 1
@@ -103 +103 @@
-/* #undef XSEC_XALAN_REQS_MEMORYMANAGER */
+#define XSEC_XALAN_REQS_MEMORYMANAGER 1
@@ -125 +125 @@
-/* #undef XSEC_OPENSSL_D2IX509_CONST_BUFFER */
+#define XSEC_OPENSSL_D2IX509_CONST_BUFFER 1
We aren't that far from getting back the useful feature of uncommited
change checking. Can't perhaps pristine-tar pull off a clever trick?
>> I'm also trying to find my way through pbuilder... Looks promising.
> It's extremely handy. I don't think I have any specific setup
> instructions anywhere for that. I recommend using it with cowdancer; it's
> a lot faster than the stock setup with a tarball. If you install
> cowdancer, the cowbuilder man page has details.
Thanks for the tip. Does overriding cleaner allow you not to have all
the build depends installed on the host machine? This is pretty
essential because I need both an Etch and a Sid build environment.
Changing them is still tough, I'm tempted to include .gbp.conf files
into the respective branches... :)
On performance, I plan to investigate parallel makes (-j4), which
could also make a big difference.
One more thing: could you please put links to the repositories on the
Alioth project page? git.debian.org loads rather slowly, and editing
the location bar isn't always practical either.
More information about the Pkg-shibboleth-devel