6:06 p.m.
I can’t access the XEmacs tracker. The error message is:
http://tracker.xemacs.org/XEmacs/its
Service Temporarily Unavailable
The server is temporarily unable to service your request due to maintenance downtime or
capacity problems. Please try again later.
Apache/2.2.3 (Debian) mod_python/3.2.10 Python/2.4.4 proxy_html/2.5 Server at
tracker.xemacs.org Port 80
Is it really only a temporary problem or something serious?
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
I � Unicode
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
4:02 a.m.
Mike FABIAN writes:
http://tracker.xemacs.org/XEmacs/its
Service Temporarily Unavailable
Is it really only a temporary problem or something serious?
It's back up now.
The university had a 2-day power out for scheduled maintenance, and I
forgot to bring the system back up.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
5:44 p.m.
New subject: Problems with byte compiled startup files
"Stephen J. Turnbull" <stephen(a)xemacs.org> さんは書きました:
Mike FABIAN writes:
> http://tracker.xemacs.org/XEmacs/its
>
> Service Temporarily Unavailable
> Is it really only a temporary problem or something serious?
It's back up now.
The university had a 2-day power out for scheduled maintenance, and I
forgot to bring the system back up.
Thank you very much!
I’ve added
http://tracker.xemacs.org/XEmacs/its/issue405
which was reported as
http://bugzilla.novell.com/show_bug.cgi?id=432404
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
I � Unicode
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
3:38 p.m.
New subject: Problems with byte compiled startup files
Mike FABIAN writes:
http://tracker.xemacs.org/XEmacs/its/issue405
which was reported as
http://bugzilla.novell.com/show_bug.cgi?id=432404
I can't replicate with your example init file (does the custom.el file
actually exist? I don't have one in my test set up since you didn't
mention its contents) on any of several platforms (PPC and Intel Mac
with Mac OS X 10.4 and 10.5 respectively, Gentoo Linux/AMD Opteron,
NetBSD 3/Intel Pentium III), and I have always (since I started using
XEmacs in 1996) byte-compiled my init file. I think most of the
developers do. I've never heard of a problem with byte-compiled init
files, whether init.el or application dotfiles.
My tests are with Mercurial tip or close to it, though. I will try
with a vanilla 21.5.28 tomorrow, I guess, but I don't think any
changes have been made to the byte-compiler or byte-code interpreter
recently (Aidan seems to have been fiddling with it in Dec 2007).
One possibility is that you've got an evil C compiler. I can't help
with that because XEmacs currently compiles warning-free with gcc,
except for the warnings about "not a prototype" in getopt.h, the
attempt to fool the compiler into accessing memory in a debug
function, and a deprecation warning in a Mac OS X header file. That
is with the default GCC 4.0.1 (Apple build) and GCC 4.3.2, and with
g++ of GCC 4.3.2 (except that g++ complains about implicit conversions
from const char[] to char* a few hundred times, but that seems
unlikely to be related to this).
Another possibility is that it's related to 64-bit uncleanliness, but
that's something I know nothing about, and my Opteron box is in 64-bit
mode.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
3:46 p.m.
New subject: Problems with byte compiled startup files
"Stephen J. Turnbull" <stephen(a)xemacs.org> さんは書きました:
Mike FABIAN writes:
> http://tracker.xemacs.org/XEmacs/its/issue405
>
> which was reported as
>
> http://bugzilla.novell.com/show_bug.cgi?id=432404
I can't replicate with your example init file (does the custom.el file
actually exist?
No. I forgot to mention that.
I don't have one in my test set up since you didn't
mention its contents) on any of several platforms (PPC and Intel Mac
with Mac OS X 10.4 and 10.5 respectively, Gentoo Linux/AMD Opteron,
NetBSD 3/Intel Pentium III), and I have always (since I started using
XEmacs in 1996) byte-compiled my init file. I think most of the
developers do. I've never heard of a problem with byte-compiled init
files, whether init.el or application dotfiles.
My tests are with Mercurial tip or close to it, though. I will try
with a vanilla 21.5.28 tomorrow, I guess, but I don't think any
changes have been made to the byte-compiler or byte-code interpreter
recently (Aidan seems to have been fiddling with it in Dec 2007).
One possibility is that you've got an evil C compiler. I can't help
with that because XEmacs currently compiles warning-free with gcc,
except for the warnings about "not a prototype" in getopt.h, the
attempt to fool the compiler into accessing memory in a debug
function, and a deprecation warning in a Mac OS X header file. That
is with the default GCC 4.0.1 (Apple build) and GCC 4.3.2, and with
g++ of GCC 4.3.2 (except that g++ complains about implicit conversions
from const char[] to char* a few hundred times, but that seems
unlikely to be related to this).
Another possibility is that it's related to 64-bit uncleanliness, but
that's something I know nothing about, and my Opteron box is in 64-bit
mode.
I am using AMD 64bit as well.
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
I � Unicode
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
3:57 p.m.
New subject: Problems with byte compiled startup files
Mike FABIAN <mfabian(a)suse.de> さんは書きました:
> One possibility is that you've got an evil C compiler. I
can't help
> with that because XEmacs currently compiles warning-free with gcc,
> except for the warnings about "not a prototype" in getopt.h, the
> attempt to fool the compiler into accessing memory in a debug
> function, and a deprecation warning in a Mac OS X header file. That
> is with the default GCC 4.0.1 (Apple build) and GCC 4.3.2, and with
> g++ of GCC 4.3.2 (except that g++ complains about implicit conversions
> from const char[] to char* a few hundred times, but that seems
> unlikely to be related to this).
>
> Another possibility is that it's related to 64-bit uncleanliness, but
> that's something I know nothing about, and my Opteron box is in 64-bit
> mode.
Anything I could test to find out whether it is related to the
C compiler or 64-bit uncleanliness?
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
I � Unicode
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
7:22 a.m.
New subject: Problems with byte compiled startup files
Mike FABIAN writes:
Mike FABIAN <mfabian(a)suse.de> さんは書きました:
>> One possibility is that you've got an evil C compiler. I can't help
>> with that because XEmacs currently compiles warning-free with gcc,
>> except for the warnings about "not a prototype" in getopt.h, the
>> attempt to fool the compiler into accessing memory in a debug
>> function, and a deprecation warning in a Mac OS X header file. That
>> is with the default GCC 4.0.1 (Apple build) and GCC 4.3.2, and with
>> g++ of GCC 4.3.2 (except that g++ complains about implicit conversions
>> from const char[] to char* a few hundred times, but that seems
>> unlikely to be related to this).
>>
>> Another possibility is that it's related to 64-bit uncleanliness, but
>> that's something I know nothing about, and my Opteron box is in 64-bit
>> mode.
Oh, BTW, could you attach Installation and the compiled simple init.el
to the bug report?
Also, do you apply any local patches to the XEmacs sources? Copies
(or pointers to them) might help.
Anything I could test to find out whether it is related to the
C compiler or 64-bit uncleanliness?
As I say, I don't know much about 64-bit or AMD vs. Intel issues; you
might ask somebody around you, including the Emacs people since they
have similar problems to ours with respect to our evil use of unions
and/or casting for Lisp objects.
For the C compiler, you could try building with Intel's compiler or a
BSD compiler, also older versions of GCC.
Get a build log and look for warnings other than the getopt.h
prototype warning (that's some sort of backwards compatibility thing
and doesn't actually affect the xemacs binary AFAIK) and "operation
has no effect" warning in emacs.c (that is a memory debugging tool).
Build with --with-xemacs-compiler=g++, --with-union-type, and
--with-debug for the same version of GCC and look for warnings. You
should filter with "grep -v 'deprecated conversion'" because of the
const char[] vs. char* issue, which I don't think is related, and even
if it were there are a million of those, of which exactly zero come
from src/bytecode.c.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
11:46 a.m.
New subject: Problems with byte compiled startup files
Stephen J. Turnbull writes:
I can't replicate with your example init file (does the custom.el
file
actually exist?
[...]
My tests are with Mercurial tip or close to it, though. I will try
with a vanilla 21.5.28 tomorrow
Nope, no problem with that on Mac OS X 10.4.11/iBook G4.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
10:57 a.m.
New subject: Problems with byte compiled startup files
"Stephen J. Turnbull" <stephen(a)xemacs.org> さんは書きました:
Oh, BTW, could you attach Installation and the compiled simple
init.el
to the bug report?
done.
Also, do you apply any local patches to the XEmacs sources? Copies
(or pointers to them) might help.
I also attached that information to
http://tracker.xemacs.org/XEmacs/its/issue405
There are a few patches but I don’t believe any of them has anything
to do with byte code.
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
I � Unicode
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
12:30 p.m.
New subject: Problems with byte compiled startup files
"Stephen J. Turnbull" <stephen(a)xemacs.org> さんは書きました:
> Anything I could test to find out whether it is related to the
> C compiler or 64-bit uncleanliness?
As I say, I don't know much about 64-bit or AMD vs. Intel issues; you
might ask somebody around you, including the Emacs people since they
have similar problems to ours with respect to our evil use of unions
and/or casting for Lisp objects.
I found now that it indeed happens only on 64-bit.
For the C compiler, you could try building with Intel's compiler
or a
BSD compiler, also older versions of GCC.
Get a build log and look for warnings other than the getopt.h
prototype warning (that's some sort of backwards compatibility thing
and doesn't actually affect the xemacs binary AFAIK) and "operation
has no effect" warning in emacs.c (that is a memory debugging tool).
I cannot see any warnings which seem to be related. bytecode.c compiles
without warnings.
Build with --with-xemacs-compiler=g++, --with-union-type, and
--with-debug for the same version of GCC and look for warnings. You
should filter with "grep -v 'deprecated conversion'" because of the
const char[] vs. char* issue, which I don't think is related, and even
if it were there are a million of those, of which exactly zero come
from src/bytecode.c.
Even if I compile with these options I cannot see warnings from bytecode.c.
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
I � Unicode
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
4:19 p.m.
New subject: Problems with byte compiled startup files
Mike FABIAN writes:
I found now that it indeed happens only on 64-bit.
I cannot see any warnings which seem to be related. bytecode.c
compiles
without warnings.
Comparing the 64-bit/32-bit attachments, pretty clearly it's not
directly a 64-bit problem as such. The 32-bit-compiled file is OK,
but for some reason the 64-bit-compiled file is being saved in UTF-8.
Which is *not* supposed to happen, ever. Byte-compiled files are
supposed to be saved in binary or a special escape-quoted encoding.
Why 64-bit-ness would affect this, I don't know. These are both the
same SuSE sources, right?
I don't have a patch, yet, since I don't know why this is happening.
buffer-file-coding-system is forced to raw-text or escape-quoted in
byte-compile-insert-header (lisp/bytecomp.el). Apparently something
is overriding that. You could try putting a Lisp debugger breakpoint
on write-region before byte-compiling, then checking the value of that
variable.
Is latin-unity active? I notice that -vanilla seems to help. Do you
have anything in the site-start or default.el files? (Those don't
seem to be available directly on build.opensuse.org.) I can't think
of anything else offhand that would override buffer-file-coding-system.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta
5666
days inactive
5672
days old
10 comments
3 participants
participants (3)
-
Mike FABIAN -
Stephen J. Turnbull -
Stephen J. Turnbull