Re: zsh (4.2.0-10)

Re: zsh (4.2.0-10)

[Clint Adams]

> According the manpage, I sent a kill -s SIGCONT. Did that, this is what
> happened:

Technically it should be "kill -s CONT".

> ../../Test/B04read.ztst: starting.
> make[2]: *** [check] Terminated
> make[1]: *** [check] Terminated
> Build killed with signal 15 after 150 minutes of inactivity
> Build killed with signal 9 after 5 minutes of inactivity
> ******************************************************************************

Probably sbuild sent the SIGTERM long before you sent the SIGCONT, and
the build didn't react until you did.

Previously, this had only been seen using sbuild on ultrasparc, and
recently Kurt Roeckx reported a problem using sbuild on amd64.  I assume
you're using sbuild on m68k.  Yes?

Re: zsh (4.2.0-10)

[Clint Adams]

> Oups... I tried it locally, ran top in one shell, SIGSTOP (and also STOP)
> stop the process like ctrl-z. With SIGCONT or CONT it somehow starts again,
> but not for very long. Is this really supposed to work?

As I recall, if you kill the right process with a SIGTERM or something,
then send SIGCONT to the remaining processes, it will proceed to the
other tests and complete the build successfully.

> Yes. I was wondering, too. The previous build worked out fine, it was built
> on a different box. I also had zsh-beta a couple of times and zsh_4.2.0-7 on
> my mac without problems, so maybe its something fancy with the Amiga? I got
> an ICE while building gcc-snapshot on the Amiga where a mac crashed only
> much later. Should I try to build it on another box? Currently we have some
> CPU cycles to kill on m68k, I'll just try t build it on my mac.

I find it curious that only sbuild exposes this problem, and now that
it's manifesting on multiple architectures, I have even less of an idea
of what the problem could be.

I wonder if a simple perl script which merely did the same signal handling as
sbuild and ran zsh's make check would give the same result.

Re: zsh (4.2.0-10)

[Clint Adams]

> I wonder if a simple perl script which merely did the same signal handling as
> sbuild and ran zsh's make check would give the same result.

It doesn't, nor does "sudo chroot /path/to/chroot sudo -u clint -H sh -c
'cd /tmp/zsh && make check".  Maybe it has something to do with the
pipes for the logs.

Re: zsh (4.2.0-10)

[Goswin von Brederlow]

"Christian T. Steigies" <cts@debian.org> writes:

> On Thu, May 13, 2004 at 12:06:04PM -0400, Clint Adams wrote:
>> > Oups... I tried it locally, ran top in one shell, SIGSTOP (and also STOP)
>> > stop the process like ctrl-z. With SIGCONT or CONT it somehow starts again,
>> > but not for very long. Is this really supposed to work?
>> 
>> As I recall, if you kill the right process with a SIGTERM or something,
>> then send SIGCONT to the remaining processes, it will proceed to the
>> other tests and complete the build successfully.
>> 
>> > Yes. I was wondering, too. The previous build worked out fine, it was built
>> > on a different box. I also had zsh-beta a couple of times and zsh_4.2.0-7 on
>> > my mac without problems, so maybe its something fancy with the Amiga? I got
>> > an ICE while building gcc-snapshot on the Amiga where a mac crashed only
>> > much later. Should I try to build it on another box? Currently we have some
>> > CPU cycles to kill on m68k, I'll just try t build it on my mac.
>> 
>> I find it curious that only sbuild exposes this problem, and now that
>> it's manifesting on multiple architectures, I have even less of an idea
>> of what the problem could be.
>> 
>> I wonder if a simple perl script which merely did the same signal handling as
>> sbuild and ran zsh's make check would give the same result.
>
> I don't think it's sbuild's fault, after all it only installes/deinstalls
> packages and then calls dpkg-buildpackage.
> But the interesting thing is, it now built fine on my mac. The significant
> differences between those machines are, the amiga has an 68060 processor,
> while the mac has an 68040. The amiga runs linux-2.4.26, while the mac runs
> linux-2.2.26.
>
> Either I had too many cosmic rays from work in my pocket, or we have a
> problem hiding in 060 or linux-2.4 support on m68k. Does somebody want to
> build gcc-snapshot on anon 060 machine?

See BTS. With kernel 2.2.10 on my Amiga 68060 it fails with a normal
error real early. No ICE or any.

> Or maybe I am completely wrong, an easy test would be if somebody tried to
> build zsh on another 060 machine to see if it hangs during the tests too.
>
> Christian

Zsh build fine on Amiga: debuild -B -us -uc -m"Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de>"

A4000T 68040@50MHz 96Mb ram
Linux a4000 2.2.10 #1 Mon Nov 27 20:50:14 CET 2000 m68k GNU/Linux

sid chroot with proc mounted and small /dev

mrvn@a4000:~% ls /mnt/chroot/sid/dev
MAKEDEV@  dsp1      loop4   midi1       port   ram14  ram9       smpte3
agpgart   dsp2      loop5   midi2       ptmx   ram15  random     sndstat
audio     dsp3      loop6   midi3       pts/   ram16  rmidi0     tty
audio1    full      loop7   mixer       ram@   ram2   rmidi1     tty0
audio2    initctl|  mem     mixer1      ram0   ram3   rmidi2     urandom
audio3    kmem      midi0   mixer2      ram1   ram4   rmidi3     zero
audioctl  loop0     midi00  mixer3      ram10  ram5   sequencer
console   loop1     midi01  mpu401data  ram11  ram6   smpte0
core@     loop2     midi02  mpu401stat  ram12  ram7   smpte1
dsp       loop3     midi03  null        ram13  ram8   smpte2

mrvn@a4000:~% gcc --version
gcc (GCC) 3.3.3 (Debian 20040429)

mrvn@a4000:~% ld --version
GNU ld version 2.14.90.0.7 20031029 Debian GNU/Linux

mrvn@a4000:~% as --version 
GNU assembler 2.14.90.0.7 20031029 Debian GNU/Linux

mrvn@a4000:~/zsh% cat *changes
Format: 1.7
Date: Wed, 12 May 2004 09:22:36 -0400
Source: zsh
Binary: zsh zsh-static zsh-doc
Architecture: m68k
Version: 4.2.0-10
Distribution: unstable
Urgency: medium
Maintainer: Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de>
Changed-By: Clint Adams <schizo@debian.org>
Description:
 zsh        - A shell with lots of features
 zsh-static - A shell with lots of features (static link)
Changes:
 zsh (4.2.0-10) unstable; urgency=medium
 . 
   * ZW#19921: fix for ZW#19920.
   * ZW#19877: fix for ZW#19869.
   * ZW#19878: fix for ZW#19877.
   * ZW#19893: fix the potential for an infinite loop in history locking.
Files:
 14ad54d8f01446c4b5bea411cb059591 1922336 shells optional zsh_4.2.0-10_m68k.deb
 21f97d44023fee096daa0652418bc191 552574 shells optional zsh-static_4.2.0-10_m68k.deb



Now we need someone with 2.4 + 040 and amiga + 2.4 to test.

MfG
        Goswin

Re: zsh (4.2.0-10)

[Clint Adams]

> Zsh build fine on Amiga: debuild -B -us -uc -m"Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de>"
> 
> A4000T 68040@50MHz 96Mb ram
> Linux a4000 2.2.10 #1 Mon Nov 27 20:50:14 CET 2000 m68k GNU/Linux

Could you try with sbuild?
I also would be interested in knowing about any hung "make check"
instances where sbuild wasn't involved.

> Now we need someone with 2.4 + 040 and amiga + 2.4 to test.