openssh on NeXTstep

[Michael Weiser]

On Tue, 23 Sep 2003, Ben Lindstrom wrote:

> > That's a pity - I've got two NeXTstep m68k boxes here which run quite
> I have a nice 25mhz Slab sitting on a desk at home.. Access has never been
> an issue.  It is willingness to baby the platform.  At 25mhz a build
> takes around 15 - 20 minutes (assuming all goes well).  And it seems to
> get worse. =)
I'm used to that and willing to do testing if that'd help you continue 
supporting the port.

> > What is the actual purpose of mysignal()? Is it safe to just remove that
> > define? I've diff'ed openssh 3.7.1p1 against 3.6.1p2 and found that in
> No we can't remove it.  It actually is there for a very good reason.
> Partly to solve issues with different singal() versions.  mysignal() is
> suppose to be an extremely strict version of signal().  In a few spots of
> our code we have needed it.  Between 3.6.1 and now we decided to use
> mysignal() by default for every signal() in our tree.
Okay. 

> It would be nice to know what signal() location it is causing the
> behavior.  Maybe a backtrace from gdb and server side -ddd reports.
It's the very first call from seed_rng called in main. The debug session
with backtraces is below. Funnily enough it doesn't segfault any more but 
recall mysignal!? Is perhaps the #define wreaking havoc with the actual 
signal call in bsd-misc.h so that it recursively calls itself until it 
crashes?

(gdb) break mysignal
Reading in symbols for bsd-misc.c...done.
Breakpoint 1 at 0x33f46: bsd-misc.c:206.
(gdb) cd openbsd-compat
Working directory /private/tmp/openssh-3.7.1p2/openbsd-compat.
(gdb) run -p 2222
Starting program: /private/tmp/openssh-3.7.1p2/sshd -p 2222

Breakpoint 1, mysignal (sig=?, act=void) at bsd-misc.c:206
206     {
(gdb) n
mysignal (sig=?, act=void) at bsd-misc.c:226
226             return (signal(sig, act));
(gdb) 

Breakpoint 1, mysignal (sig=?, act=void) at bsd-misc.c:206
206     {
(gdb) bt
Reading in symbols for entropy.c...done.
Reading in symbols for sshd.c...done.
#0  mysignal (sig=?, act=void) at bsd-misc.c:206
#1  0x33f58 in mysignal (sig=?, act=void) at bsd-misc.c:226
#2  0x24590 in seed_rng () at entropy.c:78
#3  0x4f82 in main (ac=?, av=0x33820c) at sshd.c:951
(gdb) n
mysignal (sig=?, act=void) at bsd-misc.c:226
226             return (signal(sig, act));
(gdb) 

Breakpoint 1, mysignal (sig=?, act=void) at bsd-misc.c:206
206     {
(gdb) bt
#0  mysignal (sig=?, act=void) at bsd-misc.c:206
#1  0x33f58 in mysignal (sig=?, act=void) at bsd-misc.c:226
#2  0x33f58 in mysignal (sig=?, act=void) at bsd-misc.c:226
#3  0x24590 in seed_rng () at entropy.c:78
#4  0x4f82 in main (ac=?, av=0x33820c) at sshd.c:951
(gdb) 

> It also be nice to see a copy of your config.
Attached. I had to manually remove HAVE_MMAP and HAVE_TCSENDBREAK because 
they were misdetected. Otherwise it's as configure wrote it.

> I am planning on getting a new "NeXT" laptop in a few months.. Oddly
> enough they mispelt the name of it.. It has "Apple" written on it.
Same for me. ;)
-- 
bye, Micha
-------------- next part --------------
A non-text attachment was scrubbed...
Name: config.h.gz
Type: application/x-gzip
Size: 7092 bytes
Desc: 
Url : http://lists.mindrot.org/pipermail/openssh-unix-dev/attachments/20030924/bed34fa5/attachment.bin