Re: [LAD] Announcing PHASEX-0.14.96

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: William Weston <whw@...>
Cc: linux-audio-dev@lists.linuxaudio.org <linux-audio-dev@...>
Date: Monday, January 7, 2013 - 4:15 am

--f46d040715d5cd5f8804d2ab143f
Content-Type: text/plain; charset=ISO-8859-1

On Sun, Jan 6, 2013 at 6:57 PM, William Weston wrote:

>

just a note: i haven't really publically explained the very sad failure of
the weak linkage stuff.

we incorrectly assumed that the linux gcc+ld stack had the same
capabilities as their variants on OS X. OS X provides a *very nice* design
for weak linkage that is incredibly useful for precisely this sort of thing.

alas, it turned out that the work done on the apple variants was never
back-ported into mainstream gcc+ld, and as a result, the weak linkage test
approach that i (and a few others advocated) does not work on linux unless
code is specially linked in a way that is not sensible to recommend for
applications. basically, you can do weak linkage tests *between libraries*
(or other code linked with -fPIC, but not between libraries and
applications.

this is very sad, and the alternative (an approach first implemented by
native instruments, of all groups!) seems like the only way to handle this.

--p

--f46d040715d5cd5f8804d2ab143f
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

On Sun, J=
an 6, 2013 at 6:57 PM, William Weston <whw@linuxmail.org> wr=
ote:

Explicit checks for new JACK headers library calls have been added
to the autoconf system instead of relying on the weakjack linkage
alone. =A0Could you try it again?just a note: i h=
aven't really publically explained the very sad failure of the weak lin=
kage stuff.we incorrectly assumed that the linux gcc+ld stack had t=
he same capabilities as their variants on OS X. OS X provides a *very nice*=
design for weak linkage that is incredibly useful for precisely this sort =
of thing.
alas, it turned out that the work done on the apple variants was never =
back-ported into mainstream gcc+ld, and as a result, the weak linkage test =
approach that i (and a few others advocated) does not work on linux unless =
code is specially linked in a way that is not sensible to recommend for app=
lications. basically, you can do weak linkage tests *between libraries* (or=
other code linked with -fPIC, but not between libraries and applications.<=
br>
this is very sad, and the alternative (an approach first implemented by=
native instruments, of all groups!) seems like the only way to handle this=
.--p

--f46d040715d5cd5f8804d2ab143f--

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
Re: [LAD] Announcing PHASEX-0.14.96, William Weston, (Sun Jan 6, 11:57 pm)
Re: [LAD] Announcing PHASEX-0.14.96, Paul Davis, (Mon Jan 7, 4:15 am)