[gentoo-dev] mozilla non-deterministic behavior.

[gentoo-dev] mozilla non-deterministic behavior.

[Ferris McCormick]

Let me apologize in advance for sending this to the wrong places
rather that submitting a bug report.  I don't know where the bug is,
and I an double-posting because I am not sure this isn't SPARC specific.


Very Short description:
-----------------------
  1.  If you give mozilla complete garbage to display,
      sometimes it fills the screen with garbage, and sometimes
      it exits silently.  This is with the SAME garbage ---
      
http://gentoo.oregonstate.edu/experimental/sparc/livecd/sparc64/
	gentoo-sparc64-1.4_rc4-08Sep2003-cd1.iso
      will do.
  2.  Same with mozilla-firebird, except it provides "Illegal Instruction"
      as a clue (to buffer overrun?)

Longer description:
-------------------
  As above, plus the following:
  3.  Systems involved are sparc U2(lacewing), U60(antaresia) with 
identical kernels 2.4.21-r1(SMP)
  4.  mozilla is " [  I] net-www/mozilla-1.4-r3 (0)" (current for sparc)
  5.  mozilla-firebird is "[M~I] net-www/mozilla-firebird-0.7 (0)"
      otherwise running pretty much flawlessly.
  6.  xfree on U2 is "[  I] x11-base/xfree-4.3.0-r3 (0)" (current for 
      sparc)
  7.  U60 has two root partitions.  One is an exact clone of the other (as 
      of 13 November) except for xfree-4.3.0-r2.  Behavior is 
      non-deterministic with either partition mounted on "/".
  8.  All xfree versions have xfree patched (as per comments #32, 33 to 
      bug 19776 because for me xfree is otherwise unusable.)
  9.  Except for today's invitation to update glibc, both systems are
      completely current and rock solid.

Observations:
-------------
  10. On a third system (U2-SMP, terciopelo), problem is less likely to 
      occur (maybe
      never, but I haven't sat at it displaying garbage over and over).
      It started out as a clone of the original U2, and is current,
      but I cannot guarantee that its software configuration is
      identical to the original U2 (lacewing).
  11. The U2(lacewing) and U60(antaresia) are not identically configured
      (hardware or software packages.)
  12. Empirically, -root- has a slightly better chance of success than
      I as me do.
  13. Empirically, it seems to help if you try to list a '.bz2' file 
      first; both mozillas filter that out as something they can't deal
      with and ask for help, but then seem more likely to display
      garbage for the '.iso' file.
  14. The "Illegal instruction" makes this look like a buffer overrun or
      otherwise an attempt to execute data or a program which isn't
      loaded yet.
  15. Note that mozilla looks to be pretty agressively threaded.
  16. For me, gdb has problems with mozilla.
  17. Oh, before you ask:  The mozilla ebuilds downgrade the CFLAGS for
      mozilla to "-mcpu=v8 -mtune=ultrasparc" so there should be no
      sparc64 issues here nor similar "-cpu=v9" questions.
  18. No, I don't use any font server.
  19. I doubt that this is related to xfree-4.3.0-r3's cryptic "If
      you have font problems with openoffice, regenerate XF86Config"
      advice, because:
   a. xfree-4.3.0-r2 seems to exhibit the same behavior;
   b. And I couldn't act on that advice anyway, because it doesn't
      give a clue how one should change XF86Config.
----------------

By no means do I reject the possibility that this is a problem of my
own creation.  But if it is such a thing, I do not see how it does not
stem from poor interaction among otherwise benign packages.  (I say this
so that I am more likely to see what I've screwed up immediately after
sending this, so that I'll have to come back and admit my mistakes :-) )

I suppose I am asking for suggestions on how you (collectively) would
like me to proceed.  I am not really in a position to track down what look
to me like (possibly sparc only) timing problems between mozilla and ?? 
which cause mozilla silently to give up.

Sorry for going on so long,
Regards,
Ferris

--
Ferris McCormick (P44646, MI) <fmccor@inforead.com>
Phone: (703) 392-0303
Fax:   (703) 392-0401


--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] mozilla non-deterministic behavior.

[Felipe Ghellar]

I have tried the test case you suggested, about 10 times in a row, and 
Mozilla exited silently in all of them. This was with mozilla-1.4-r3 on 
an Athlon XP.

Felipe

Ferris McCormick wrote:
> Let me apologize in advance for sending this to the wrong places
> rather that submitting a bug report.  I don't know where the bug is,
> and I an double-posting because I am not sure this isn't SPARC specific.
> 
> 
> Very Short description:
> -----------------------
>   1.  If you give mozilla complete garbage to display,
>       sometimes it fills the screen with garbage, and sometimes
>       it exits silently.  This is with the SAME garbage ---
>       
> http://gentoo.oregonstate.edu/experimental/sparc/livecd/sparc64/
> 	gentoo-sparc64-1.4_rc4-08Sep2003-cd1.iso
>       will do.
>   2.  Same with mozilla-firebird, except it provides "Illegal Instruction"
>       as a clue (to buffer overrun?)
> 
> Longer description:
> -------------------
>   As above, plus the following:
>   3.  Systems involved are sparc U2(lacewing), U60(antaresia) with 
> identical kernels 2.4.21-r1(SMP)
>   4.  mozilla is " [  I] net-www/mozilla-1.4-r3 (0)" (current for sparc)
>   5.  mozilla-firebird is "[M~I] net-www/mozilla-firebird-0.7 (0)"
>       otherwise running pretty much flawlessly.
>   6.  xfree on U2 is "[  I] x11-base/xfree-4.3.0-r3 (0)" (current for 
>       sparc)
>   7.  U60 has two root partitions.  One is an exact clone of the other (as 
>       of 13 November) except for xfree-4.3.0-r2.  Behavior is 
>       non-deterministic with either partition mounted on "/".
>   8.  All xfree versions have xfree patched (as per comments #32, 33 to 
>       bug 19776 because for me xfree is otherwise unusable.)
>   9.  Except for today's invitation to update glibc, both systems are
>       completely current and rock solid.
> 
> Observations:
> -------------
>   10. On a third system (U2-SMP, terciopelo), problem is less likely to 
>       occur (maybe
>       never, but I haven't sat at it displaying garbage over and over).
>       It started out as a clone of the original U2, and is current,
>       but I cannot guarantee that its software configuration is
>       identical to the original U2 (lacewing).
>   11. The U2(lacewing) and U60(antaresia) are not identically configured
>       (hardware or software packages.)
>   12. Empirically, -root- has a slightly better chance of success than
>       I as me do.
>   13. Empirically, it seems to help if you try to list a '.bz2' file 
>       first; both mozillas filter that out as something they can't deal
>       with and ask for help, but then seem more likely to display
>       garbage for the '.iso' file.
>   14. The "Illegal instruction" makes this look like a buffer overrun or
>       otherwise an attempt to execute data or a program which isn't
>       loaded yet.
>   15. Note that mozilla looks to be pretty agressively threaded.
>   16. For me, gdb has problems with mozilla.
>   17. Oh, before you ask:  The mozilla ebuilds downgrade the CFLAGS for
>       mozilla to "-mcpu=v8 -mtune=ultrasparc" so there should be no
>       sparc64 issues here nor similar "-cpu=v9" questions.
>   18. No, I don't use any font server.
>   19. I doubt that this is related to xfree-4.3.0-r3's cryptic "If
>       you have font problems with openoffice, regenerate XF86Config"
>       advice, because:
>    a. xfree-4.3.0-r2 seems to exhibit the same behavior;
>    b. And I couldn't act on that advice anyway, because it doesn't
>       give a clue how one should change XF86Config.
> ----------------
> 
> By no means do I reject the possibility that this is a problem of my
> own creation.  But if it is such a thing, I do not see how it does not
> stem from poor interaction among otherwise benign packages.  (I say this
> so that I am more likely to see what I've screwed up immediately after
> sending this, so that I'll have to come back and admit my mistakes :-) )
> 
> I suppose I am asking for suggestions on how you (collectively) would
> like me to proceed.  I am not really in a position to track down what look
> to me like (possibly sparc only) timing problems between mozilla and ?? 
> which cause mozilla silently to give up.
> 
> Sorry for going on so long,
> Regards,
> Ferris
> 
> --
> Ferris McCormick (P44646, MI) <fmccor@inforead.com>
> Phone: (703) 392-0303
> Fax:   (703) 392-0401
> 
> 
> --
> gentoo-dev@gentoo.org mailing list
> 


--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] mozilla non-deterministic behavior.

[Ferris McCormick]

Thanks for the feed-back.  I have taken the liberty of attaching
extracts of your responses to (my new) bug report 33750 as
evidence of failure someplace besides at my desk.

For your information, in case you are curious:  Yes, I try to
display garbage on purpose.  Last spring, a similar failure was
a result of incompatible freefont packages on my system,
so, every now and then, I exercise mozilla with junk.  In
this case, xfree-4.3.0-r2 --> xfree-4.3.0-r3 seemed like a
good time.  (But apparently, failure is independent from
xf version.)
--
Ferris McCormick (P44646, MI) <fmccor@inforead.com>
Phone: (703) 392-0303
Fax:   (703) 392-0401


--
gentoo-dev@gentoo.org mailing list