9:16 p.m.
Hey all --
I've had quite a bit of trouble getting 21.5.29 working in our
environment. After comparing paths.h between 21.5.28 and 21.5.29, I
finally figured out why:
-#define PATH_LOADSEARCH ".../lib/xemacs-21.5-b28/lisp"
+#define PATH_LOADSEARCH ".../share/xemacs-21.5-b29/lisp"
-#define PATH_SITE ".../lib/xemacs/site-lisp"
+#define PATH_SITE ".../share/xemacs/site-lisp"
-#define PATH_DATA ".../lib/xemacs-21.5-b28/etc"
+#define PATH_DATA ".../share/xemacs-21.5-b29/etc"
-#define PATH_INFO ".../share/info"
+#define PATH_INFO ".../share/xemacs-21.5-b29/info"
and a similar change in setup-paths.el.
Those of us with, ahem, complex installations tend to do some
post-processing on the install tree, and this is broken by arbitrarily
moving stuff around. I did eventually find an inconspicuous changelog
entry dating to 2007, and from there, a mail thread on the issue:
http://thread.gmane.org/gmane.emacs.xemacs.beta/25477/focus=25511
but this is a silly thing to have wasted so much time on.
In the future, I'd appreciate it if this sort of thing were
highlighted right up top in the release notes, along with other things
that are important to your friendly neighborhood local maintainers...
--
Ron Isaacson
Morgan Stanley
ron.isaacson(a)morganstanley.com / (212) 762-1051
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
6:57 a.m.
Ron Isaacson writes:
I've had quite a bit of trouble getting 21.5.29 working in our
environment. After comparing paths.h between 21.5.28 and 21.5.29, I
finally figured out why:
xemacs -debug-paths
might short-circuit that a bit.
In the future, I'd appreciate it if this sort of thing were
highlighted right up top in the release notes, along with other things
that are important to your friendly neighborhood local maintainers...
I'll do what I can, but with two years between beta releases, I can't
really say I'm surprised something like that snuck through. I'm not
even sure that the actual change wasn't in 21.5.28 or even earlier;
the logs in 21.5.29 look like bugfixes to an earlier change.
You also shouldn't take the fact of a release or its absence too
seriously; compared to any given commit, the only extra QA that a
release gets in 21.5 is a guarantee that it builds, runs, and shows no
new regressions on my machines (Mac OS X with default GCC and Gentoo
Linux with GCC 4.2 currently) in my default configuration (with Mule,
X11, modules, debug, and error checking, and maybe a couple other
minor features).
AFAIK, few of the active developers have the resources to regularly do
installs, so beta testing would be very helpful for this kind of
thing.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
11:21 p.m.
All grumbling aside, I'm still struggling with this... I don't think
the Makefile was properly updated to reflect the change from lib to
share. The install-arch-dep target does:
if test "${prefix}" != "${exec_prefix}"; then \
$(MAKEPATH) ${exec_prefix}/lib/${instvardir}; \
for dir in \
lib/${inststaticdir} \
lib/${instvardir}/etc \
lib/${instvardir}/info \
lib/${instvardir}/lisp; do \
if test ! -d ${exec_prefix}/$${dir}; then \
$(LN_S) ${prefix}/$${dir} ${exec_prefix}/$${dir}; fi; \
done; \
fi
However, at startup time, it seems to be looking for these directories
in share/${instvardir}, not lib. Patch attached.
<stephen(a)xemacs.org> wrote:
Ron Isaacson writes:
> I've had quite a bit of trouble getting 21.5.29 working in our
> environment. After comparing paths.h between 21.5.28 and 21.5.29, I
> finally figured out why:
xemacs -debug-paths
might short-circuit that a bit.
> In the future, I'd appreciate it if this sort of thing were
> highlighted right up top in the release notes, along with other things
> that are important to your friendly neighborhood local maintainers...
I'll do what I can, but with two years between beta releases, I can't
really say I'm surprised something like that snuck through. I'm not
even sure that the actual change wasn't in 21.5.28 or even earlier;
the logs in 21.5.29 look like bugfixes to an earlier change.
You also shouldn't take the fact of a release or its absence too
seriously; compared to any given commit, the only extra QA that a
release gets in 21.5 is a guarantee that it builds, runs, and shows no
new regressions on my machines (Mac OS X with default GCC and Gentoo
Linux with GCC 4.2 currently) in my default configuration (with Mule,
X11, modules, debug, and error checking, and maybe a couple other
minor features).
AFAIK, few of the active developers have the resources to regularly do
installs, so beta testing would be very helpful for this kind of
thing.
--
Ron Isaacson
Morgan Stanley
ron.isaacson(a)morganstanley.com / (212) 762-1051
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
3:02 p.m.
Ron Isaacson <Ron.Isaacson(a)morganstanley.com> writes:
All grumbling aside, I'm still struggling with this... I
don't think
the Makefile was properly updated to reflect the change from lib to
share.
That's right - these symlinks are downright silly. The real reason is
that the path-searching has bugs in situations where prefix !=
exec-prefix. I've just pushed a bunch of changes that, in conjunction,
fix the problem.
Thanks for the report!
--
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
7:59 p.m.
<sperber(a)deinprogramm.de> wrote:
Ron Isaacson <Ron.Isaacson(a)morganstanley.com> writes:
> All grumbling aside, I'm still struggling with this... I don't think
> the Makefile was properly updated to reflect the change from lib to
> share.
That's right - these symlinks are downright silly. The real reason is
that the path-searching has bugs in situations where prefix !=
exec-prefix. I've just pushed a bunch of changes that, in conjunction,
fix the problem.
Great, thanks. I haven't tried another build yet, but with startup.el
looking in emacs-data-roots instead of emacs-roots, it should find
these installed in the PREFIX location instead of EXEC_PREFIX, and the
symlinks should indeed be unnecessary.
I'll let you know when 21.5.30 comes out. :-)
--
Ron Isaacson
Morgan Stanley
ron.isaacson(a)morganstanley.com / (212) 762-1051
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
5377
days inactive
5395
days old
4 comments
3 participants
participants (3)
-
Michael Sperber -
Ron Isaacson -
Stephen J. Turnbull