Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-22 Thread Ebrahim Mayat
On Sep 20, 2010, at 5:12 PM, Benjamin Reed wrote: This is a common issue with cmake projects. People write the cmake bits on linux and don't realize they're missing settings for proper ld initialization on other platforms. :) You generally need something along the lines of this in

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-22 Thread Benjamin Reed
On 9/22/10 10:06 AM, Ebrahim Mayat wrote: To recap the build and install scripts and the SplitOff field are: CompileScript: #!/bin/sh -ev /bin/mkdir build cd build %p/bin/cmake \ -DCMAKE_INSTALL_PREFIX=%i \ .. InstallScript: #!/bin/sh -ev cd build

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-22 Thread Ebrahim Mayat
On Sep 22, 2010, at 1:28 PM, Benjamin Reed wrote: %i would be wrong, you tell cmake the *real* install prefix, and then use DESTDIR to set it to the temporary root (%d). Otherwise, %i is the build directory: /sw/src/fink.build/package/foo/sw and then DESTDIR=%p makes it put another

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-22 Thread Daniel Macks
On Mon, 20 Sep 2010 23:23:50 -0400, Daniel Macks wrote: On Tue, 21 Sep 2010 06:26:48 0900, David R. Morrison wrote: Also, I'm wondering if we should maybe enhance that validator test so that if it finds a Shlibs with an incomplete pathname, it gives additional advice (such as

[Fink-devel] Shlibs: install_name and compatibility version

2010-09-20 Thread Ebrahim Mayat
Good Day When preparing a package update I get the following error: Validating .deb dir /sw/src/fink.build/root-fluidsynth-1.1.2-358... Package looks good! dpkg-deb -b root-fluidsynth-1.1.2-358 /sw/fink/10.5/local/main/ binary-darwin-powerpc dpkg-deb: building package `fluidsynth' in

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-20 Thread Alexander Hansen
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 9/20/10 10:15 AM, Ebrahim Mayat wrote: Good Day When preparing a package update I get the following error: Validating .deb dir /sw/src/fink.build/root-fluidsynth-1.1.2-358... Package looks good! dpkg-deb -b root-fluidsynth-1.1.2-358

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-20 Thread Ebrahim Mayat
On Sep 20, 2010, at 11:15 AM, Alexander Hansen wrote: $ otool -L /sw/src/fink.build/root-fluidsynth-shlibs-1.1.2-358/sw/ lib/libfluidsynth.1.dylib /sw/src/fink.build/root-fluidsynth-shlibs-1.1.2-358/sw/lib/ libfluidsynth.1.dylib: libfluidsynth.1.dylib (compatibility version 1.0.0,

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-20 Thread Benjamin Reed
On 9/20/10 5:03 PM, Ebrahim Mayat wrote: Since we are in the process of migrating to cmake from autotools, i would like to rectify this error. This is a common issue with cmake projects. People write the cmake bits on linux and don't realize they're missing settings for proper ld

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-20 Thread David R. Morrison
This sounds like something which should be added to the FAQ if its not there already. Also, I'm wondering if we should maybe enhance that validator test so that if it finds a Shlibs with an incomplete pathname, it gives additional advice (such as what Ben wrote in this email...) -- Dave

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-20 Thread Ebrahim Mayat
On Sep 20, 2010, at 5:12 PM, Benjamin Reed wrote: On 9/20/10 5:03 PM, Ebrahim Mayat wrote: Since we are in the process of migrating to cmake from autotools, i would like to rectify this error. This is a common issue with cmake projects. People write the cmake bits on linux and don't

Re: [Fink-devel] Shlibs: install_name and compatibility version

2010-09-20 Thread Daniel Macks
On Tue, 21 Sep 2010 06:26:48 0900, David R. Morrison wrote: Also, I'm wondering if we should maybe enhance that validator test so that if it finds a Shlibs with an incomplete pathname, it gives additional advice (such as what Ben wrote in this email...)   I think the actual bug