You need to find where the -undefined suppress flag is set: often it
is buried in the configure script, sometimes it is in Makefile.in .
There are two possible replacements for this flag:
1) -undefined error which will let you compile with the default
-twolevel_namespace.
2)
I had a message from a friend of mine who was installing fink the old
fashioned way, bootstraping from source. He discovered that one of our
essential packages, debianutils, is out of date and the source is no
longer available at the Debian site. In fact, the modification dates
there indicate
I agree with your strategy Max: let's fix the problems and then release
0.4.1 relatively soon.
On the libmpeg problem, it looks to me like the version in unstable is OK
but the version in stable is the bad one. Has anybody tested this? We
should move the good version to stable as soon as we
Max, your new gdk-pixbuf package seems to have fixed all of the problems.
Thanks!
However, the current structure of this package does raise an important
question about the shared libraries policy. gdk-pixbuf has a bunch of
shared objects (foo.so), which are different from shared libraries.
If I
Max Horn [EMAIL PROTECTED] wrote:
At 7:22 Uhr -0400 25.04.2002, David R. Morrison wrote:
[snip]
So I would think that this package should be split into three pieces,
maybe the third one to be called gdk-pixbuf-loaders. It would be treated
like the binaries are treated in some other
Max Horn [EMAIL PROTECTED] wrote:
At 9:51 Uhr -0400 25.04.2002, David R. Morrison wrote:
Max Horn [EMAIL PROTECTED] wrote:
At 7:22 Uhr -0400 25.04.2002, David R. Morrison wrote:
[snip]
So I would think that this package should be split into three pieces,
maybe the third one
should one add all these libraries that are used in case they are present at
build time as an dependency?
Yes.
Sometimes you can avoid this with a flag to the ./configure script, however,
to force the package to compile without using the library. But if you
can't disable, and if it is used
Looks like a bug, alright.
Check out the openldap-ssl submission on the package submission tracker, and
see if that has fixed it. I'm just waiting for one final response from
the submitter before committing the new one to CVS.
-- Dave
___
Since Fink is written in perl and uses curl for downloads, you should be
able to install Fink on a linux box and run fink fetch foo or fink
fetch-all with no problem (if that's what you had in mind), assuming you
have uptodate versions of perl and curl there.
-- Dave
For those of you not watching the fink-commits list: I have added more
BuildDepends entries to many many packages in the past 12 hours. Don't
worry, it's all being done by a script now!
As before, the point is this: if package A depends on package B, and
package B depends on package C, there
As previously discussed, I now have put splitoff versions of the following
packages into the unstable tree:
graphics/fnlib-0.5-3.info
graphics/freetype-1.3.1-5.info
graphics/freetype-hinting-1.3.1-4.info
graphics/giflib-4.1.0-3.info
graphics/gimp-1.2.3-6.info
graphics/imagemagick-5.4.1-4.info
Speaking only for myself, of course: I think it's great whenever patches
go back upstream, no matter how they get there. Ideally, a packaging
system like fink should only need to adjust things like installation
directories, with the software itself just working due to good
configure scripts and
All of the libtool stuff has been sent back upstream to the libtool project
a long time ago, and we see the results of this sometimes, when other
upstream authors update the libtool that they are using. This includes
the 10.1 fix that you mention (-flat_namespace -undefined_suppress).
I don't
It's not a dumb question at all.
If you need to revise a package, and you are not a core fink developer,
please post the revised version to the package submission tracker, just
like you did with the original version.
-- Dave
In a discussion on #fink about the passwd problem, it was suggested that
we put an announcement about the problem in the News section on Fink's
homepage. Appended below is a diff file for the source of that page,
which would contain the announcement. Any suggestions about the wording,
or about
Can you check to see if you have the file
/sw/fink/dists/unstable/main/finkinfo/graphics/libungif-4.1.0b1-3.info
on your system? If you do, please remove it and try again.
-- Dave
___
Have big pipes? SourceForge.net is looking
On second thought, it might actually be possible for users without DevTools
to follow the fink rebuild passwd instructions. cc and make aren't
run in this case: the only commands invoked by the CompileScript and
InstallScript are sed, mkdir, and install. Shouldn't those already be
present, even
A few days ago I put splitoff versions of several packages into unstable,
including some packages which conflict with each other: giflib vs. libungif,
freetype vs. freetype-hinting, and imagemagick vs. imagemagick-nox.
A few users reported trouble to me, from which I learned a few things:
1)
Yup. Got that one two. New versions of both are in CVS.
-- Dave
___
Have big pipes? SourceForge.net is looking for download mirrors. We supply
the hardware. You get the recognition. Email Us: [EMAIL PROTECTED]
=?ISO-8859-1?Q?Jean-Fran=E7ois_Mertens?= [EMAIL PROTECTED] wrote:
And amaya-gtk depends on libungif _ which is NOT provided by gifib
(and this is a main package)
This is something the recent discussion didn't clarify for me:
for long, libungif provided giflib, and not the other way around,
I think we need a policy for how to handle packages which are abandoned by
their maintainers. I propose this:
1) Change the maintainer field to
Maintainer: The Fink Project [EMAIL PROTECTED]
We can add formerly maintained by... to DescPackaging if it seems
appropriate.
2)
I've run into trouble figuring out an upgrade strategy for pairs of packages
when converting them to splitoff form. There are two pairs already in
unstable: freetype/freetype-hinting and giflib/libungif. There is a third
one in the shared-libraries/splitoff module: gnome-vfs/gnome-vfs-ssl.
I
David R. Morrison wrote:
I've just fixed a bug in my system-tetex package which made it impossible to
install unless the user already had the Developer Tools installed. (The
post-install script was calling ranlib.) The new version is in the unstable
tree, but I would like to suggest
Of course i suppose we'd have to update every package
Not necessarily. If the Source-MD5 field is present, fink should check
against it; if not, fink would let the download through. It then would
be up to the maintainers of packages to add this if they wanted.
-- Dave
Here is the latest list of splitoff packages, which are now in the unstable
tree and will move to stable in a few days:
languages/guile-1.4-4
languages/librep-0.14-6
languages/ruby-1.6.6-2
languages/tcltk-8.3.3-7
libs/expat-1.95.1-3
libs/libghttp-1.0.9-3
libs/libole2-0.2.4-2
Hi Martin. All that is being added is a BuildDepends, so I don't want to
bump version numbers. The lesstif package is being built the same way,
but now being subdivided into splitoffs.
For testing purposes, I only put the new lesstif into unstable for a few
days, so I could only add
Hi all. I was just compiling a package from unstable, and I got
make: gcc: Command not found
How could this happen, you may wonder? My guess is that the package maintainer
has the April Dev Tools installed, and so gcc is present on his system
(invoking gcc2).
This is just a heads up to watch
I'm corresponding with a user who has his /bin/sh symlinked to bash.
It is very instructive to see what problems he is having, as we prepare
for 10.2. He has had the 0.0.0 versioning problem that Dave V. found
a libtool fix for, and he has had another problem: for some reason,
on his system,
I'm trying to debug a problem for a user, and I believe I have isolted it
to the fragment of a Makefile reproduced below. Is it possible that
this behaves differently under zsh and bash? (I don't know my shell
scripting very well...) I wondered if the test -f could be the culprit.
THanks,
Mark:
Just to put this all in a bit of perspective: I understand that the Debian
project has around 700 developers. Fink has 20 core developers, plus
another 20-30 people who have made substantial contributions.
Apple offered some support to fink late last summer in the form of providing
Mark,
As you have correctly pointed out, fink's evolution package is unusable.
Max has removed it so that others will not have to go through an experience
similar to yours.
Most of fink's packages are in fact usable. We're sorry that you had a
poor initial experience. If you have trouble with
Fink's passwd package says the following to users during installation:
Or
you can say no here and add the users and groups manually (e.g. on your
central NetInfo server). If you don't know what all of this is about,
just say yes.
Now the problem with this is, if someone has added the
There are lots of different approaches to dealing with the Mac-text vs.
Unix-text issue. You actually wanted mac2unix, I think, or something like
that...
If you are not going to use emacs or vi, there is a free Mac text editor,
BBEdit Lite, which handles the Mac/Dos/Unix text conversion very
Ben Hines [EMAIL PROTECTED] wrote:
Do you have a script that will let us determine all the builddepends
we should be putting in our packages? (ie, all the ones you have been
adding to people's info files. :)
-Ben
Hi Ben. I'm answering to the fink-devel list because other people may want
All you need to do is fink selfupdate-cvs. We updated all of the
affected packages to work around this problem.
-- Dave
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas --
mathias meyer wrote:
the same error alo occurs when trying to build libxml2-2.4.21-2 and
bonobo-1.0.20-1 with gcc3. i juat startet a little logfile of successful
gcc3 compilations of fink packages, so if this is of interest i can post
it on the list. havan't compiled that many packages,
I have added a new file to the shared-libraries CVS module, called
BuildDepends.list .
If you are creating a new package and you want to be sure to cover all the
bases with BuildDepends, here's what you do:
1) Figure out everything that the package depends on (this includes
I'd like to draw everyone's attention to a new section in the docs:
http://fink.sourceforge.net/doc/porting/preparing.php#gcc3
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas --
Lloyd, thanks again for your work on updating the X11 section of the fink
documentation. I have now incorporated your changes into the web site,
and made a few additional updates.
I'd like to encourage other Fink developers to check it out
http://fink.sourceforge.net/doc/x11/index.php
to see
Hi. Because of the changes made at sourceforge, doing a fink install from
source at the moment is very difficult, perhaps impossible. However,
a binary install works fine. What you need to do is, immediately after
installing, run fink selfupdate-cvs. This updates the package location
Max, I wonder if we should in the future make two kinds of binary distribution:
the Mac OS X binary (an Apple Installer .pkg, as now), and the Darwin binary
(which presumbaly would be a .deb for the Darwin Debian system). I don't
know how many people are actually trying to run Fink on Darwin,
Sorry, I think I got confused between system-free86 and free86-base, etc.
Your list is probably fine.
-- Dave
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas --
The Source-MD5 tools have revealed that a number of TeX-related packages are
changing with time. (I downloaded sources at different times on my two
machines, and I'm getting checksum errors.) I'm planning to fix this by
putting a definitive copy of the sources on fink's website (unless they
are
Hello all.
As part of my work on the shared-libraries project, I now have a fink
installation with essentially every fink package installed, stable and unstable
(except of course for conflicting packages). This makes it easy to see
where there are violations of fink's FHS (filesystem hierarchy
Over the last few days, the following shared-library versions have been
added to the unstable tree. The ones with a star will be moved to stable
later this week, after some time has passed for testing.
-- Dave
*crypto/gnome-vfs-ssl-1.0.3-5.info
crypto/gnome-vfs-ssl-1.0.5-3.info
Jeremy Higgs [EMAIL PROTECTED] wrote:
Hi!
I wasn't quite sure who to email about this, but since you're doing much of
the shared-libraries converting, I thought it would be more appropriate to
email you.
I got this while doing a 'fink update-all' a couple of minutes ago:
fink needs
Thanks. I'll add emacs21 to the list in the porting docs of packages which
are known to not yet compile under Jaguar.
However, note that ncurses is an essential package in Fink, so you should
have gotten curses support even though the library moved in OS X.
I'm puzzled that you did not.
--
How about %p/share/dx?
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm
___
Fink-devel mailing
Kyle,
The process of upgrading existing packages to fit into a good shared
libraries system is almost over. All of the dependency problems that have
arisen in the past few weeks have been side-effects of the upgrade, and all
of them have been quickly resolved. Once the upgrade is over, there
Kyle,
I agree that we need to improve fink to respond better in those circumstances.
The problem has been with us ever since users have been prompted for a
choice (from late October or early November). For example, if a user
attempts to install bundle-tetex, he/she is repeatedly asked to choose
It was brought up today on #fink that in the shared libraries project, there
has not been total consistency in how we are using the -bin and -dev
variants.
I admit to being somewhat at fault here. Although we clearly spelled out
in the policy discussions that -bin should be used when the
In many cases, it is caused by building the library with bash as your
shell. See the comments about bash on fink's porting page.
-- Dave
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las
fink selfupdate-cvs will fix your problem.
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm
___
Hi Justin. The lesstif choice is a temporary thing... lesstif-dev is going
away, and once the new lesstif has been tested and I can move it to stable,
lesstif-dev will go away completely.
The other one, choosing between kdebase3 and kdebase3-ssl, is a well-known
shortcoming of fink: it asks you
Torrey,
Thanks for the heads up about forthcoming changes.
Actually, the fink team had a chance to figure out the lesstif problems
in advance of the 4.2 release, because we did have fink packages which
were tracking development versions of XFree86. We had identified that
there was an
Hi Erik. You haven't told us what problem you're having!
We've run into a number of different libtool problems, and have done our
best to address them. We send patches back to the libtool team but I'm
not sure if all of them have been accepted.
You might have a look at
You might have to edit ltmain.sh by hand, in order to apply Dave's patch...
-- another Dave
(msg resent by request)
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas -
Hi Bill. maxima is under GPL, right? Let's make a tarball of the source
and put it on fink's site... the package can download from there.
-- Dave
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28
Everyone please note that the change which Ben has requested will require
a bump in revision number, in order to force users to update to the new
version. (He needs this so that he can put openmotif into fink.)
-- Dave
___
Don't
Sounds like a good idea to me.
-- Dave
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas -
http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink
All of you kind folks who are testing gcc3 and bash for fink compiles, or
who have been seeded, should please update your fink installation to the
latest pre-release version. I have updated fink's ltconfig file using
Ben Hines' install_name patch, and this will immediately fix a number of
fink
Justin, I've never heard of this. What are you talking about?
-- Dave
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas -
http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink
I'm trying to build autogen now, and it looks pretty easy.
By the way, are you aware that sourceforge has two Mac OS X boxes in their
compile farm? :-)
-- Dave
___
Don't miss the 2002 Sprint PCS Application Developer's Conference
mathias meyer [EMAIL PROTECTED] wrote:
david
To use the pre-release version of fink, you need to check out the fink
module from CVS, cd into that directory, and run ./inject.pl.
do you mean the
cd tempdir/fink
cvs -z3 update -d
./inject.pl
procedure which gives me now
Thanks for the report. We knew about the apt problem, and a number of
others, and we're working on getting Fink ready for Jaguar. You might
want to look at the web page
http://fink.sourceforge.net/doc/porting/preparing.php
-- Dave
The bash versioning bug showed up in two essential packages, libiconv and
gettext. Since Max is away, I went ahead and fixed these (in unstable),
and then Masanori completed the fix in the case of libiconv. (I missed the
fact that there were two different configure scripts.)
We'll leave it up
You asked why Fink Commander cannot be installed using Fink.
Here's a brief summary: the things which Fink installs are designed to be
put into one location in the filesystem, and to stay there. Most Mac OS X
applications, including Fink Commander, are designed so that they can be
put anywhere
Regarding your report for failure for the newest readline package:
the package should compile on 10.1 with gcc2 (or the old developer tools),
or on 10.2 with gcc3. It won't compile on 10.1 with gcc3.
-- Dave
___
Don't miss the
readline won't compile with gcc3 on 10.1 because of the way the patch was
made. An extra library to link to has to be specified when the gcc3
compiler is being used to build shared libraries, in this instance,
and the (non-libtool) custom build-system of readline is specifying those
libraries in
The experimental CVS module sounds good to me too.
-- Dave
---
Sponsored by:
ThinkGeek at http://www.ThinkGeek.com/
___
Fink-devel mailing list
[EMAIL PROTECTED]
Thanks for taking the trouble to make reports -- it really helps fink.
There are two ways you can do this: either use the bug tracker which is
linked from fink's homepage (with a separate item for each bug, please),
or else email the package maintainer directly. You can find the maintainer's
Yes, the former maintainer who originally brought this package to fink is
taking an extended break from the internet, and you would be welcome to
take the package over. The way to do that would be to write a revised
version of the .info and .patch files for the package, and then submit
your
I've just committed revised versions of dia, expat, libghttp, php, and xfce
which address bash and/or gcc3 issues.
(These are not my packages.)
I would like to encourage all fink developers to take a look at the lists
prepared by JF Mertens of packages which do NOT compile under gcc3. Even
if
On Jul 3,2002 12:17:03 +0200, Karim-Pierre Maalej [EMAIL PROTECTED] wrote :
Hi everybody,
I am just beginning trying to port Unix software on Mac OS X. The
first step is to port the software on Darwin, still using X11 and Motif
libs. I have experienced problems, and as I am a user of
On Jul 5,2002 13:43:06 -0600, Jeff Whitaker [EMAIL PROTECTED] wrote :
On Fri, 5 Jul 2002, Benjamin Reed wrote:
Jeff Whitaker [[EMAIL PROTECTED]] wrote:
If it doesn't actually need openssl (library-wise or otherwise) then it
shouldn't depend on it. The problem that Martin uncovered is a
On Jul 8,2002 07:45:28 -0700, Vartan Katchikian [EMAIL PROTECTED] wrote :
Dear Fink Admins,
We are the best seller french written Mac magazine SVM Mac, and I'm
the How-To's section editor. We have a cover CD and we would like to
know if you authorize us to put the Fink Installer and a few
From: David R. Morrison [EMAIL PROTECTED]
To: [EMAIL PROTECTED]
Subject: distributing Fink (was Re: Dr Dobb's Journal)
Cc: [EMAIL PROTECTED]
Rosalyn Lum [EMAIL PROTECTED] wrote:
Hi,
DDJ is working on a Lightweight Language CD and would like to possibly
include the MAC version of Python
I have solved the problems with the new version of readline, in a new
revision, readline-2.3-1.
The underlying problem is that the wrong -install_name is being given when the
library is being built. That can't be fixed until the upstream maintainers
have been consulted.
When some users have a
-- Forwarded message --
Date: Thu, 11 Jul 2002 16:52:58 -0700
From: Avi Rappoport [EMAIL PROTECTED]
To: [EMAIL PROTECTED]
Subject: Thanks very much for fink
I'm starting to run command line apps, mainly search engines, on Mac OS X and I really
am happy to have found Fink. I
If you need to use fink's copy of the sourcefile for you package (which fink
would have for all packages that were stable at the time of the last release),
the URL for accessing this has changed. You formerly used something like
Source: http://fink.sourceforge.net/bindist/source/%n-%v.tar.gz
On Sep 20,2002 00:58:41 +1000, Michael Stillwell [EMAIL PROTECTED] wrote :
What is the fink group's policy on dependencies?
[snip]
I edited the .info file and built the package with /none/ of these
dependencies; it seems to work fine. Why do I need to install python
(and X11!) to get an
When your packages are ready, you submit them for consideration by using
the package submission tracker linked from fink's webpage.
-- Dave
---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
Thanks for diagnosing this problem: I had noticed it affecting fvwm2 at some
point, but was never able to accurately determine what was going on.
I have created modified versions of the swi-prolog and wmmail packages
which do not depend on libxpm, and they both build fine on 10.1 and 10.2;
I
Ben Hines has made a nice list of packages which have not yet been moved to
10.2. Perhaps this is a good time for fink developers to make a list packages
which should *not* be moved.
1) manconf should not be moved (I believe this is the upshot of previous
discussions)
2) presumably,
Yes, 0.5.0pre.cvs sounds good to me...
---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
Fink-devel mailing list
[EMAIL PROTECTED]
Max Horn [EMAIL PROTECTED] wrote:
At 14:44 Uhr -0700 06.10.2002, Ben Hines wrote:
On Sunday, October 6, 2002, at 01:16 PM, David R. Morrison wrote:
Conflicts: neon-ssl, neon, neon19, neon19-ssl, neon21, neon21-ssl, neon22
Replaces: neon-ssl, neon, neon19, neon19-ssl, neon21, neon21-ssl
I'm working on the next phase of the shlibs project, and I'd like to get the
opinion of other Fink developers on one aspect of this.
The system will automatically examine the libraries which have been linked
to by a particular binary, and keep track of which packages provide those
libraries.
Let me suggest that if we are going to go for option 1, that the labels
on the pdb web page make this clear, i.e.,
In 0.4.1-stable (10.1):
In current-stable (10.2):
In current-unstable (10.2):
I also think that when we release 0.5.0, for some transition period we
should include four lines
The time has come to decide the fate of the fink package for openssl in 10.2.
It would be good to get this settled before too many things move to the
stable tree.
My vote is to remove the package, and to eliminate many other packages
from the crypto tree which depend on it. In many cases, those
I'm forwarding this to fink-devel because I'm not sure what the right
approach is.
tetex-base says Provides: dialog and dialog says Conflicts: tetex-base
The situation is that dialog is a very small package which someone might
want independently; tetex-base is a large package which includes
Dear Fink Devlopers,
I have implemented the next phase of the shlibs project, which I'll discuss
in this message.
In this phase, in addition to making sure that all shared libraries are
kept in separate fink packages which can remain installed long after
other things have been upgraded, you will
Ben Hines has asked me in private email why we need the %n (= %v-%r)
part of the new Shlibs field.
Although I can't give an example of this at the moment, I can explain
some future scenarios where this will be used.
Suppose that your package builds two shared libraries, libfoo.1.dylib
and
I can provide one more clue about this situation: I made the binary
installer for 0.4.1, and my user id is 2011. I'm not at all clear
how the permissions on your system could have been changed, however.
-- Dave
---
This sf.net email is
Martin,
Should I make a new copy of the 0.4.1 installer which has permissions 775
on the root directory instead of 700? If so, should I give it a new
name?
-- Dave
---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
Dear Fink Developers,
As you may have seen in a few recent messages, there is a serious problem
with the fink 0.4.1 binary installer which occurs if a user installs using
the installer twice. The permissions on the root directory of the machine
are altered, and the machine no longer boots!
I've
Martin Costabel [EMAIL PROTECTED] wrote:
David R. Morrison wrote:
[]
What my fix does is to alter the .pax.gz file with the contents of the
installation so that the permission on . is 1755 in the directory
instead of 0700.
What did you do with the bom file? Not only does it still
jeff whitaker [EMAIL PROTECTED] wrote:
Hi folks: I'm getting ready to add a python-2.2.2 package to CVS. Some
time back, I was asked to add a two-level namespace python shared library
(/sw/lib/python2.2/config/libpython2.2.dylib) because koffice apparently
wouldn't build without it. Now
Jason,
You are following the instructions on the jag-bootstrap page, and yet one
of your earlier messages showed that you were in a fink-0.4.1-full directory?
Where did that come from? You shouldn't have fink-0.4.1 on your system
at all if you are following those instructions...
-- Dave
I've tried with both source installers I could find on the Fink page.
Well, one is intended for 10.1 and one for 10.2, so that could be part of
the problem.
If you ran the bootstrap.sh found in the fink-0.4.1-full tarball, then you
installed an old, 10.1, version of fink and you need to do the
Martin Costabel [EMAIL PROTECTED] wrote:
David R. Morrison wrote:
Maybe somebody
else can make a new 0.4.1 installer? I'll be happy to talk you through
the process...
In case this is still needed, I made one. It looks OK to me, but since I
had never touched PackageMaker before
101 - 200 of 1174 matches
Mail list logo