Re: Bug#394418: mono_1.1.18-3 (arm/unstable): FTBFS: SIGSEGV
On 11/16/06 Paolo Molaro wrote:
> > He has netwinders, cats, etc machines available. All of them are arm
> > v4l, which seemed to me to be the relevant thing here.
>
> I have contacted Bdale already, I'll mail Vince shortly anyway:
> having two boxes likely increases the chance to hit the bug if it's
> not already fixed (I'll use mono 1.2.1 to build and I made a few arm
> fixes the last few days).
A small update.
Vince gave me access to a Simtec-OSIRIS (as reported in /proc/cpuinfo).
On it I built 3 times in a row a svn snapshot of what is soon to be released
as mono 1.2.1. It built successfully each time (on this fairly fast
machine this involves about 2 hours of jitting and garbage collecting).
I then built the 1.1.18-x version that apt-get source downloaded
and that one showed the FilenotFound exception bug. I applied the fix
I implemented in mono 1.2.1 and it built successfully. Right now
it is running a second build of the same sources: it hasn't finished
yet, but it has passed the point which used to cause the bug.
Bdale gave me access to netwinder: the build is not yet finished as
this is a much slower box with 64 MB of ram vs 128, but it has passed
the point of failure. I'll let the build finish and then do another
build inside the unstable chroot, but I'm confident it will build there
as well (on Vince's box the builds where on a sid chroot).
So the summary is that either using mono 1.2.1 or applying the patch to
earlier versions solves all the building issues for mono as well as the
packages which manifested themselves as a FileNotfound exception: the
bug was in the exception handling code.
The only still unresolved issue is the SEGV which showed up on elara:
http://buildd.debian.org/fetch.cgi?pkg=mono&arch=arm&ver=1.1.18-3&stamp=1161343362
If I am given access to elara I can try to reproduce that, but it didn't
happen in jitted code and it didn't happen in the 6 builds I ran today
on other boxes: if it's a bug in the runtime it would be common to all
the architectures and unless it has been fixed in mono 1.2.1 already, it
doesn't look common or reproducible.
The patch is here if anyone is interested:
http://lists.ximian.com/pipermail/mono-patches/2006-November/082367.html
but I suggest to go ahead with mono 1.2.1 which should be released on
Monday (we're preparing rc tarballs today).
Thanks.
lupus
--
-----------------------------------------------------------------
lupus@debian.org debian/rules
lupus@ximian.com Monkeys do it better
Reply to: