#musicbrainz

/

      • deadchip
        no that's not a problem
      • i'm enough C/C++ literate myself to fix that
      • what i mean is this:
      • i'm trying to find out why this happens
      • luks
        hmm!
      • deadchip
        i guess they link libfftw statically AND dynamically
      • at least that's how it looks like
      • luks
        i've never seen this before
      • deadchip
        well they for sure do link it dynamically
      • luks: well and i assure you before we get into discussions that i must be doing something odd that i don't :P
      • (just a precaution :P)
      • they link against -lfftw3
      • tw3
      • and then they have:
      • luks
        yep, but why statically? :)
      • deadchip
        well AFLIB
      • luks: no idea
      • cause they're closed source people
      • they always develop like that
      • luks
        it links dynamically to fttw3 on here
      • -on
      • tru
        I had that problem on macosx.
      • deadchip
        luks:
      • AFLIB/.libs/libaflib.a(align.o):align.c:(.text+0x0): first defined here
      • tru
        I remcompiled fttw3 dynamic only.
      • deadchip
        yeah that's what i was about to do
      • but i first wanted to find out what the reason/bug is in their buildsystem
      • luks
        well, i think it's a problem with your setup
      • deadchip
        ahem
      • look what tru just said
      • it might be, "of course" a problem with both our setups >_>
      • "sure" o_O
      • luks
        no, i mean, do you want to link to fftw3 statically or dynamically?
      • deadchip
        well i'm trying to find the reason and submit them a patch
      • luks: it's not the point what _i_ want
      • that thing should not try to link boht
      • both*
      • luks
        well, it uses -lfftw3
      • deadchip
        yeah and somehow it also links to libfftw3.a
      • it's obvious
      • look at what i pasted above
      • luks: really i'm coding and buildsystem (at least autotools) and gcc/g++ literate enough to tell you that this is what this thing is trying to do
      • luks
        ah, i see
      • deadchip
        and i'm trying to find out why
      • luks
        well, this looks more like a libtool problem
      • deadchip
        it could be, though, a libfftw3 bug
      • luks
        it passes -lfftw3 to libtool
      • and libtool uses /usr/lib/libfftw3.a
      • deadchip
        but
      • luks: that would be no problem
      • if it wouldn't at the same time to link the solib as well
      • (or .dylib on osx then)
      • yeah i have it only built statically
      • i actually thought up to now that libfftw3 exists in both forms
      • so, the problem is, it doesn't know that it links it statically
      • basically, sort of, you're right
      • but they need to either:
      • luks
        stupid question, but do you have installed the dynamic lib?
      • deadchip
        - check for fftw3 being present as a dynamically linkable lib, and otherwise not build libofa
      • or
      • - adjust their buildsystem to link properly against libfftw3 as a static lib
      • luks: no i haven't but that is not the point
      • you see you're going to tell me
      • <luks> well, that is the problme
      • em*
      • but it's not
      • well in the _end_, right now, it is the problem
      • luks
        ideally it should link to static version as well, but obviously it doesn't
      • deadchip
        but the general problem is that their build system is erroneus
      • luks
        but i don't think that's libofa bug
      • deadchip
        yeah
      • sure it is
      • luks
        how?
      • deadchip
        -lblah doesn't specify whether it should be a dynamically linkable lib or a static lib
      • their build system needs to check for the presence of the dynamic libfftw3
      • luks
        what if i have a static fftw3 (that actually works) and want to it?
      • deadchip
        they can do that very easily with
      • AC_CHECK_LIB()
      • i think
      • since nm -D will not work with a static lib
      • deadchip tries
      • srotta
        Hmm. A quick question: shouldn't singles be (generally) named after their main track?
      • deadchip
        let's see
      • luks
        so you are saying that the build system should disallow linking to static version, right?
      • deadchip
        hmmm ok AC_CHECK_LIB does not use nm or anything
      • as the check passes
      • luks: either that, or it should accomodate for linking against the static version
      • mudcrow
        srotta: yes, usually
      • srotta
        I'm seeing quite many singles that have both A and B side in the release title.
      • deadchip
        right now, it is unaware that the lib might be present as a static lib
      • luks
        but what if i *want* link to the static version?
      • deadchip
        then you have to specify --with-fftw3-static to libofa's configure
      • which currently does not exist but they must add it
      • luks is not sure if that's a good idea
      • deadchip is quite sure that while it might be not neccessarily a "good idea", it's the only correct technical solution
      • look i don't code since 3 days ago ok? :P
      • luks
        well, if you have installed both versions libtool will prefer the dynamic one. if you have only the dynamic one, libtool will use it. if you have *correct* static lib, libtool will use it
      • where's the problem?
      • the only problem is if you have broken static version
      • deadchip
        i pasted you the problem
      • i don't understand why you don't see that libofa is just broken atm
      • ?
      • luks
        can you link anything agains that static lib?
      • deadchip
        what is the problem in seeing that?
      • are you basically saying that i _must_ have the dynamic lib installed in order to build libofa
      • and if i _don't_ have it
      • ?
      • luks
        but what i _want_ to use the static lib
      • srotta
        mudcrow: I also find it annoying, that for some reason there are several vinyl singles that are categorized as EP's :P
      • deadchip
        is that your statement?
      • luks
        ?
      • deadchip
        luks: ^
      • luks: ^
      • luks
        you get the error because your libfftw3 install it broken
      • deadchip
        there are a few other apps that i know of that allow of linking to another lib dynamically or statically and they are _aware_ of the fact that this other lib can come in both forms
      • and they all have a switch lib --with-somelib-static
      • and they do prefer the dynamic version usually, some do it the other way round
      • mudcrow
        srotta: a lot of singles are a mess, listed as all sorts of things.
      • deadchip
        but this is nothing unusual, and not weird and a "bad idea" in any way
      • luks: lol?
      • luks: how is it broken?
      • i simply did not specify to build the dynamic lib
      • luks
        can you link *anything* to that fftw3?
      • deadchip
        that does not qualify as broken
      • luks: even if i can't it does not qualify as broken if it's an option to build fftw3 like that
      • luks
        why should it require you to use the dynamic link?
      • *lin
      • bleh!
      • dynamic lib
      • deadchip
        luks:
      • [mderezynski [at core] ~]$
      • sure i can link something else against it
      • as i'm telling you THEIR (libofa's) setup is broken
      • not mine, not libfftw3, or anything else
      • Freso
        Whee! Almost done with the composers on Rubai! :D
      • luks
        deadchip: you don't use any fftw3 code there
      • deadchip
        luks: that doesn't matter
      • luks: if you are familiar with linking
      • luks: then you should know that this doesn't matter
      • luks
        echo 'int main() { fftw_alignment_of();}' > test.c && gcc test.c -lfftw3
      • deadchip
        i can illustrate you their problem like this:
      • luks:
      • wait
      • no no
      • here:
      • [mderezynski [at core] fftw]$ cat tu1.c
      • [mderezynski [at core] fftw]$
      • now:
      • [mderezynski [at core] fftw]$ cat tu2.c