linker optimisation causes crashes on armel with -Wl,-O1

Marking public even though this hasn't been retraced (yet). I believe the retracer for armel is currently down, but hopefully we will get a trace on it soon.

what does the daily tag means there?

Process thing that our team uses to tag bugs that were found in the course of testing daily snapshots

Could you try to get a valgrind log of the issue? What image are you testing? Is that on real hardware or virtual boxes?

setting to incomplete until we get details on the crash

This is real hardware, and unfortunately we don't have valgrind for armel

This is blocking the use of X on both imx51, and dove at the moment. Would appreciate if we could raise this at least to "high"

weird, i dont see it on imx51

sorry but the bug has virtually no useful detail and you are the only one having the issue, any change the box you are using could have issues? nautilus crash should not break desktop login. you can also try moving the nautilus.desktop entry to see if GNOME starts without this one...

ok, after a dist-upgrade i can confirm it.
during upgrade i got the packageset from http://paste.ubuntu.com/363941/

(not including nautilus though)
i'll try to downgrade package by package and see when it goes away

after downgrading the gvfs packages to 1.5.1-0ubuntu2, I am able to get past this

more specifically libgvfscommon0 seems to be the culprit

With libgvfscommon0 1.5.1-0ubuntu2 installed, it works. However if I upgrade to 1.5.2-0ubuntu1, that's when I start to see the problem with nautilus.

seems to be caused by gvfs update. setting that to high

didn't have much luck getting a useful backtrace, even with all -dbgsym I could think of.

I spent quite a bit of time on Friday trying to bisect mainline gvfs to see if I could determine where this broke. At the end of the exercise I ended up with current 1.5.2 upstream + the few patches in the package and I was still unable to reproduce the bug. As a sanity check, I reinstalled the current (known broken) libgvfscommon 1.5.2-0ubuntu1 package and still could not reproduce the problem.

I'm now working from a live image, where I can reproduce the problem again. I tried reverse applying the patch that Alexander posted on this bug, and I can still reproduce the problem. Still investigating.

Narrowed it down to -O1. If I remove -O1 from the LDFLAGS in debian/rules, then it works (defaults to -O2).

LDFLAGS or CFLAGS? LDFLAGS default to no optimization (-O0) not O2.

This was all I changed from debian/rules:
-LDFLAGS += -Wl,-z,defs -Wl,-O1 -Wl,--as-needed
+LDFLAGS += -Wl,-z,defs -Wl,--as-needed

I noticed something scroll by saying that it was defaulting to -O2

On 02.02.2010 19:07, Paul Larson wrote:
> This was all I changed from debian/rules:
> -LDFLAGS += -Wl,-z,defs -Wl,-O1 -Wl,--as-needed
> +LDFLAGS += -Wl,-z,defs -Wl,--as-needed
>
> I noticed something scroll by saying that it was defaulting to -O2

that would be surprising, -Wl,-O2 doesn't exist.

I've uploaded this and am hence lowering importance, but this needs to be filed with binutils upstream.
nautilus (1:2.29.2-0ubuntu2) lucid; urgency=low

  * Only add -Wl,-O1 to LDFLAGS if DEB_HOST_ARCH != armel; thanks Paul Larson;
    LP #512959.

 -- Loïc Minier <email address hidden> Tue, 02 Feb 2010 19:45:44 +0100

No, I didn't change that in nautilus, it was in gvfs

Marking the nautilus task invalid since this is really happening in gvfs/binutils

Revert the nautilus change and redid the change for gvfs; sorry, was focused on the bug title and that matched the nautilus rules, my bad.

StacktraceTop:
 raise () from /lib/libc.so.6
 ?? () from /lib/libc.so.6
 ?? () from /lib/libc.so.6

StacktraceTop:
 raise () from /lib/libc.so.6
 ?? () from /lib/libc.so.6
 ?? () from /lib/libc.so.6

StacktraceTop:
 raise () from /lib/libc.so.6
 ?? () from /lib/libc.so.6
 ?? () from /lib/libc.so.6

Bug in armel binutils; keeping gvfs task open to revert the workaround once it gets fixed, but that's not release critical.

we should look if there is a simple way to reproduce it.

High priority for investigation. Possibly change importance once understood.

Is this bug still valid? If not, it will close automatically.

[Expired for Linaro Binutils because there has been no activity for 60 days.]

[Expired for binutils (Ubuntu) because there has been no activity for 60 days.]

[Expired for gvfs (Ubuntu) because there has been no activity for 60 days.]