bug#620: #620 - 23.0.60; Bootstrapping with the GNUstep port impossible - Emacs bug report logs

bug#620: #620 - 23.0.60; Bootstrapping with the GNUstep port impossible - Emacs bug report logs

[Adrian Robert]

Hi,

Someone else said they can't reproduce it, and you yourself report  
later (http://emacsbugs.donarmstrong.com/cgi-bin/bugreport.cgi? 
bug=984) a bug when emacs is being run on GNUstep, so is it safe to  
close this report?

As far as the loading-every-file-when-bootstrapping, I think this is a  
general bug with CANNOT_DUMP handling -- so that should probably be  
reported separately.

thanks

bug#620: #620 - 23.0.60; Bootstrapping with the GNUstep port impossible - Emacs bug report logs

[Yavor Doganov]

Adrian Robert wrote:
> 
> Someone else said they can't reproduce it,

Maybe on Muck OS.  On GNU/Linux, I can reproduce it always on several
hosts.  With CVS 2008-10-03 19:00 UTC:

Loading bindings (source)...
Loading files...
Loading cus-face...
Loading faces...
Loading /home/yavor/src/emacs/lisp/minibuffer.el (source)...
Fatal error (11)make[2]: *** [compile-onefile] Ошибка сегментирования
make[2]: Leaving directory `/home/yavor/src/emacs/lisp'
make[1]: *** [/home/yavor/src/emacs/src/../lisp/scroll-bar.elc] Ошибка 2

It happens on random files.

> and you yourself report later a bug when emacs is being run on
> GNUstep,

I build it this way:

./configure --with-x=no
make bootstrap
./configure --with-ns
make
make install

But it is only usable with `Emacs -nw'.

> so is it safe to close this report?

No.

> As far as the loading-every-file-when-bootstrapping, I think this is
> a general bug with CANNOT_DUMP handling -- so that should probably
> be reported separately.

I see.  This is a practical problem, as my Pentium II was building
Emacs for about a day until the failure.  Even on a powerful machine,
it takes hours.

bug#620: #620 - 23.0.60; Bootstrapping with the GNUstep port impossible - Emacs bug report logs

[Glenn Morris]

Yavor Doganov wrote:

> Adrian Robert wrote:
>> 
>> Someone else said they can't reproduce it,
>
> Maybe on Muck OS.  On GNU/Linux, I can reproduce it always on several
> hosts.

I have the same problem, also on random files, on 32-bit GNU/Linux.

On x86_64, I can't even get that far, since it segfaults trying to
make the autoloads, but I think that may be a different problem, so I
will file a separate bug for that.

>> As far as the loading-every-file-when-bootstrapping, I think this is
>> a general bug with CANNOT_DUMP handling -- so that should probably
>> be reported separately.

That's just what CANNOT_DUMP means. The files that would normally be
dumped have to be loaded every time.

> I see.  This is a practical problem, as my Pentium II was building
> Emacs for about a day until the failure.  Even on a powerful machine,
> it takes hours.

I can build it fairly quickly, if someone can suggest what debugging
info would be useful.

bug#620: #620 - 23.0.60; Bootstrapping with the GNUstep port impossible - Emacs bug report logs

[Glenn Morris]

[-- Attachment #1: Type: text/plain, Size: 331 bytes --]


I managed to get bootstrap-emacs to run under gdb while it compiles
files. So far, the segfaults are always in alloc.c, mostly in
mark_object at line 5540 or 5434, but once in mark_interval at 1508.

I have attached 4 backtraces from one `make bootstrap' (running under
gdb meant make carried on after a file failed to compile).


[-- Attachment #2: backtrace 1 --]
[-- Type: application/octet-stream, Size: 11257 bytes --]

[-- Attachment #3: backtrace 2 --]
[-- Type: application/octet-stream, Size: 4251 bytes --]

[-- Attachment #4: backtrace 3 --]
[-- Type: application/octet-stream, Size: 4221 bytes --]

[-- Attachment #5: backtrace 4 --]
[-- Type: application/octet-stream, Size: 4230 bytes --]

bug#620: #620 - 23.0.60; Bootstrapping with the GNUstep port impossible - Emacs bug report logs

[Yavor Doganov]

clone 620 -1
retitle -1 Segfaults in alloc.c within gdb on gNewSense
thanks

(Please merge back the bug if it turns out to be the same issue.)

Glenn Morris wrote:
> 
> So far, the segfaults are always in alloc.c, mostly in mark_object
> at line 5540 or 5434, but once in mark_interval at 1508.

Oddly enough, these are almost equivalent to the problems I see
(unrelated to bootstrapping) under gNewSense DeltaH (Base 1.14.1, GUI
0.12.0).  I was about to report the problem separately, after more
investigation.

On gNewSense with GNOME (with the default Metacity WM), I basically
experience the same behaviour as described in #984.

Under Window Maker and the GNUstep art backend, Emacs starts, the
frame is drawn and I can perform some basic operations.  (There are
other major issues, but let's leave them out for now.)

If I open xterm and start Emacs from within gdb, it always crashes
(under both GNUstep backends) with a backtrace very similar (if not
equivalent) to the ones Glenn posted.

Now the odd: If under wmaker I start Emacs 22.3, and from there start
Emacs.app either with M-x gdb or M-x shell; gdb ..., I get the same
backtrace as in #984.

I should probably mention that gnustep-back has fairly complicated WM
management code, and support for all window managers in the wild is
not guaranteed.  However, applications should not crash if they're run
under a poorly supported WM (Window Maker is the best supported, and
Metacity is claimed to be supported).  Personally, I was surprised to
see Emacs.app crashing under GNOME while running under wmaker.  All
other GNUstep apps run fine under GNOME.  It is either a subtle bug in
Emacs.app, or a bug in GNUstep that is exposed by a complex app such
as Emacs.

Hopefully I'll followup with more details to the cloned bug as soon as
I figure out the logic behind all of this weirdness.

Processed: Re: bug#620: #620 - 23.0.60; Bootstrapping with the GNUstep port impossible - Emacs bug report logs

[Emacs bug Tracking System]

Processing commands for control@emacsbugs.donarmstrong.com:

> clone 620 -1
bug#620: 23.0.60; Bootstrapping with the GNUstep port impossible
bug 620 cloned as bug 1083.

> retitle -1 Segfaults in alloc.c within gdb on gNewSense
bug#1083: 23.0.60; Bootstrapping with the GNUstep port impossible
Warning: Unknown package 'ns'
Changed bug title to `Segfaults in alloc.c within gdb on gNewSense' from `23.0.60; Bootstrapping with the GNUstep port impossible'.

> thanks
Stopping processing here.

Please contact me if you need assistance.

Don Armstrong
(administrator, Emacs bugs database)