* Use continuation->backing_store to determine if we return for the first
time from getcontext(). This avoids having to depend on non-POSIX semantics
of the Linux implementation.
* Avoid using __libc_ia64_register_backing_store_base by hardcoding the base
of the backing store. The kernel is changed to put the backing store at
this address. It's not entirely clean, but I want to avoid adding all sorts
of variables in our libc for this for now.
* Fix a genuine bug in the ia64 case: the ia64-specific fields are not
initialized. On FreeBSD this means they have garbage (= 0xd0). Call
memset() after allocation to make sure the contregs structure is properly
cleared. Since this may affect other architectures, this patch will be
ia64 only.
Submitted by: marcel
ports that depend on guile. As such I have tested most of them. Here
are my results:
--KNOWN TO WORK--
devel/gconf
games/gnomegames
games/gnomegames2
devel/g-wrap
x11-toolkits/guile-gtk
x11-toolkits/guile-gnome
math/guppi
finance/gnucash
deskutils/gnotime
devel/libgtop
devel/autogen
games/gnurobots
audio/snd
math/drgenius
net/serveez
science/libctl
x11-wm/scwm (works with patches to be sent to maintainer)
--KNOWN NOT TO WORK (maintainer is working on update)--
cad/geda
cad/geda-symcheck
cad/geda-utils
cad/gnetlist
cad/gschem
cad/gwave
cad/libgeda
--UNKNOWN--
mail/icqmail (marked IGNORE for other reasons)
lang/gnustep-guile (could not build for other reasons; will notify maintainer)
math/siag (did not test)
science/mpb (did not test)
PR: 43910
