Re: 1.1.2 is broken if you have closefrom()
On June 8, 2006 6:02:44 PM -0400 Alan DeKok <aland@nitros9.org> wrote:
Frank Cusack <fcusack@fcusack.com> wrote:
1. Can I just check this into the 1.1.0 branch and then upintegrate to HEAD (later)?
Ok. I'm not sure the fix is required for head, as that is already using a newer libtool, which should work.
It should be in head also (read on).
I'd like to see the diff, or have a description of it first, though. Working with libtool & autoconf is pretty insane.
Yeah. So what I did is upgrade to libtool and ltdl 1.5.22. The current libtool and ltdl is GNU + changes. This is difficult. I rearranged our autoconf stuff (ever so slightly :-)) to allow drop-in of the untouched originals. That's why I asked about the change the ltdl.c -- it'd be better to change our code than to require changes in ltdl.c each time we upgrade. I doubt upgrading libtool and ltdl has any real software effect, other than allowing R= to work. But being able to use unmodified sources does have a good maintenance effect.
3. CVS and branching is pitiful. CVS is also otherwise not very good. I suggest p4 (it's free for open source use), but barring that, svn.
Hmm... subversion doesn't do multiple repositories, and it requires everyting but the kitchen sink to build. I'm using perforce at work, and it's fine (if a little odd). For open source VCS's, I'd prefer mercurial.
Don't know mercurial, and don't know svn well enough to argue for it (except vs CVS). I will say that I'm surprised that a long list of dependencies is an issue -- what system today (that we care about) doesn't have either binary svn or an easy way (yum, apt-get, etc) to get dependencies? I'd be willing to host a p4 server on a trial basis, and possibly even long term. -frank
Frank Cusack wrote:
On June 8, 2006 6:02:44 PM -0400 Alan DeKok <aland@nitros9.org> wrote:
I'd like to see the diff, or have a description of it first, though. Working with libtool & autoconf is pretty insane.
Yeah. So what I did is upgrade to libtool and ltdl 1.5.22. The current libtool and ltdl is GNU + changes.
I'm also wary of a new version of libtool and ltdl. I think we should just pull the libtool stuff from CVS head. This version has been tested for a while, and it should solve the problem with the $R variable, too. I also note the autotools versions have to be compatible: libtool 1.4 comes with autoconf 2.13, and libtool 1.5 comes with autoconf 2.50. Upgrade one but not the other may lead to unpredictable results. And finally, the autotools upgrade won't change the behaviour of the resulting binary, therefore I don't mind calling the next version 1.1.3. -- Nicolas Baradakis
On June 9, 2006 1:56:16 PM +0200 Nicolas Baradakis <nbk@sitadelle.com> wrote:
Frank Cusack wrote:
On June 8, 2006 6:02:44 PM -0400 Alan DeKok <aland@nitros9.org> wrote:
I'd like to see the diff, or have a description of it first, though. Working with libtool & autoconf is pretty insane.
Yeah. So what I did is upgrade to libtool and ltdl 1.5.22. The current libtool and ltdl is GNU + changes.
I'm also wary of a new version of libtool and ltdl. I think we should just pull the libtool stuff from CVS head. This version has been tested for a while, and it should solve the problem with the $R variable, too.
I also note the autotools versions have to be compatible: libtool 1.4 comes with autoconf 2.13, and libtool 1.5 comes with autoconf 2.50. Upgrade one but not the other may lead to unpredictable results.
Well, even 1.1 as-is actually requires autoconf-2.50. I was going to put add that to configure.in as part of the patch. The current libtool-freeradius integration is pretty sloppy, it wasn't smooth sailing trying to upgrade. We need to fix that. The libtool just changes the build process, so I'm VERY confident of its correctness. I only updated ltdl because the new libtool requires it. I'm also highly confident of ltdl correctness. I guess the only other thing I can say here is that the difference between 1.1.2 and 1.1.3 would be the non-FR software, if folks run into problems we can just tell them to use 1.1.2 instead ... with no worries of any actual FR differences. -frank
Frank Cusack wrote:
On June 9, 2006 1:56:16 PM +0200 Nicolas Baradakis <nbk@sitadelle.com> wrote:
I also note the autotools versions have to be compatible: libtool 1.4 comes with autoconf 2.13, and libtool 1.5 comes with autoconf 2.50. Upgrade one but not the other may lead to unpredictable results.
Well, even 1.1 as-is actually requires autoconf-2.50. I was going to put add that to configure.in as part of the patch.
I don't understand: the current branch 1.1 with libtool 1.4 has always used autoconf 2.13.
The current libtool-freeradius integration is pretty sloppy, it wasn't smooth sailing trying to upgrade. We need to fix that. The libtool just changes the build process, so I'm VERY confident of its correctness. I only updated ltdl because the new libtool requires it. I'm also highly confident of ltdl correctness.
libtool isn't exactely what I'd call an "exact science". I'm more confident in the patched version we have tested for a while in CVS head than a new uptream release.
I guess the only other thing I can say here is that the difference between 1.1.2 and 1.1.3 would be the non-FR software, if folks run into problems we can just tell them to use 1.1.2 instead ... with no worries of any actual FR differences.
I completely agree with that. -- Nicolas Baradakis
participants (2)
-
Frank Cusack -
Nicolas Baradakis