There have been several MacOS issues raised recently, and I'd like to
check that I haven't lost track of any of them, hence this email.
Could MacOS-interested people please review and comment on the
following?
(We're not there yet, because some of the following points still need
work, but the objective that I'm working towards is to be able to
prepare a release candidate for 1.8.4 that will work out of the box on
MacOS.)
So, I believe the issues are as follows.
- rl_get_keymap(_name) unresolved. We have a patch for this,
confirmed working.
- Pre-install dynamic_link errors. Status of this is unclear. We
have a proposed fix (use of DYLD_LIBRARY_PATH in
pre-inst-guile-env.in and pre-inst-guile.in), but it seems the
latest builds (by Roger) have not been going down the path that
previously hit these errors - perhaps because of the use of
--enable-shared=no. So I don't think we have any evidence yet that
the fix works.
- off64_t etc. Fix proposed, need to implement and test.
- Problem with rl_pending_input, mysteriously solved by
--enable-shared=no. My best guess is that this is caused by the
combination of
- having editline installed in /usr
- having readline installed in /usr/local
- the MacOS libtool not being able to encode an rpath in the built
guile-readline library (which I understand it does do on
GNU/Linux).
This means that even if you configure with
CPPFLAGS=/usr/local/include and LDFLAGS=/usr/local/lib - causing the
correct version of readline (i.e. the /usr/local one) to be picked
up at build time, at runtime guile will still pick up editline's
emulated readline library from /usr. And the emulated library
doesn't have rl_pending_input.
Does that sound right? I've pieced this together only from a few
third hand emails, so it could be way off base.
- Stack overflow. Doubling the "stack" option worked on Intel, but
not on Powerbook. Roger, could you also try reverting "stack" to
20000, and applying the attached stackchk.h patch (previously posted
here [1]), as that might also work on Powerbook, and would be a
nicer fix overall.
[1] http://lists.gnu.org/archive/html/bug-guile/2008-01/msg00013.html
- Duplicate symbol scm_cell, when building Autogen. Previously missed
this one, will investigate and follow up shortly.
Is that everything?
Regards,
Neil
--- libguile/stackchk.h 12 Feb 2006 13:42:51 -0000 1.20.2.1
+++ libguile/stackchk.h 29 Jan 2008 22:16:54 -0000
@@ -37,11 +37,11 @@
# if SCM_STACK_GROWS_UP
# define SCM_STACK_OVERFLOW_P(s)\
(SCM_STACK_PTR (s) \
- > (SCM_I_CURRENT_THREAD->base + SCM_STACK_LIMIT))
+ > (SCM_I_CURRENT_THREAD->base + ((SCM_STACK_LIMIT * sizeof(SCM)) /
sizeof(SCM_STACKITEM)))
# else
# define SCM_STACK_OVERFLOW_P(s)\
(SCM_STACK_PTR (s) \
- < (SCM_I_CURRENT_THREAD->base - SCM_STACK_LIMIT))
+ < (SCM_I_CURRENT_THREAD->base - ((SCM_STACK_LIMIT * sizeof(SCM)) /
sizeof(SCM_STACKITEM))))
# endif
# define SCM_CHECK_STACK\
{\
