Hi Daniel,
Am 27.08.2009 um 22:19 schrieb Daniel Pocock:
Dagobert Michelsen wrote:
Am 27.08.2009 um 17:52 schrieb Daniel Pocock:
Ok, I had a go at this with libConfuse
Could you let me know whether this type of thing should be
discussed on the board list, the maintainers list or the devel list?
Usually on maintainers@, but you have to post from your opencsw.org
account
on the list which you may or may not (Ihsan?) have yet.
I haven't got an opencsw.org account as far as I know - I've tried to
subscribe to that list using my main email address, [email protected]
-
I think it is waiting for someone to approve, because I haven't
received
the confirmation email yet
You can not subscribe yourself, only the list admin can do that (Ihsan
in
this case).
Anyhow, these are the issues I found:
The name confuse and libConfuse is used interchangeably:
- web site: http://www.nongnu.org/confuse/ "/libConfuse/ is a
configuration file parser library"
- tar file is called confuse-2.6.tar.gz, and it extracts to
confuse-2.6/*
- this is in configure.ac:
AC_INIT(libConfuse, 2.6, [email protected], confuse)
Please choose the name of what is in it. If there is only a library
in it
please use CSWlibconfigure for the package and libconfuse for the
catalog.
When you extract the file confuse-2.6.tar.gz, you get a directory
called confuse-2.6
Therefore, should I call it CSWconfuse (which works), or tweak
DISTNAME and WORKSRC so that the package name can be CSWlibconfuse?
As the project describes itself as "libConfuse is a configuration file
parser library"
please call it CSWlibconfuse and libconfuse and tweak the vars.
I also decided to add --disable-nls to configure, is this
considered good
practice, or is it preferred to make everything depend on gettext
and friends?
This depends on the package. If there are no negative effects the
existing
libs from the OpenCSW should be used to minimize size and to
maximize reuse.
Because confuse is used by Ganglia, and Ganglia is a performance
monitoring app,
I prefer to minimise dependencies. I will continue with --disable-
nls for now,
and maybe it can be enabled in a future release if people ask for it.
Ok, but as gettext is needed for almost anything you will most
certainly already
have it on the system.
Although confuse works on Solaris, the test suite doesn't work out
of the box,
so I also hacked the Makefile to skip the test phase - is this
acceptable for
now, or do I need to wait for upstream to merge in changes to the
test suite
before I can commit something to opencsw?
You can always commit the build descriptions. Before submitting you
must make sure
the package works. If you have failing tests you must at least
understand why they
fail and document that in the Makefile. Even better is reporting
this to upstream
and even better fixing the tests and providing a patch for upstream.
The test code fails to compile because it is not Solaris friendly.
It is easy to fix.
The library itself is working fine. I think that I will put some
conditional logic in
the Makefile so that it only skips the tests for confuse-2.6, and
when 2.7 arrives, it
will be necessary to re-assess the situation. I will also send the
proposed fixes
upstream for 2.7.
Great!
I'm now ready to commit my Makefile for this package, but the gar
guide doesn't
explain how to authenticate - will I be issued with an SVN
password, or do I need
to reconfigure my svn to work over ssh (using my key)?
No, the repository is hosted at SourceForge. I need your
SourceForge user name to
add you to the committers list. A description to the specific
commit procedure
for SourceForge is documented here:
<https://sourceforge.net/apps/trac/sourceforge/wiki/Subversion>
My Sourceforge username is d_pocock
You should be able to commit now.
Thanks for all the assistance with this.
Sure :-)
Best regards
-- Dago
_______________________________________________
maintainers mailing list
[email protected]
https://lists.opencsw.org/mailman/listinfo/maintainers
