[maemo-developers] [maemo-developers] Re: Tinymail using mmap() gets killed by the kernel of the device

[From: Philip Van Hoof]

On Fri, 2006-07-14 at 11:15 +0100, Ross Burton wrote:
> On Fri, 2006-07-14 at 11:41 +0200, Philip Van Hoof wrote:
> > When I run tinymail behind a camel with the mmap() patch, on the device,
> > it will get killed by the kernel (it's not a Segfault, it's a kill by
> > the kernel).
> > 
> > When I run tinymail (behind a camel with the mmap..) on my desktop or in
> > scratchbox, everything works perfectly.
> > 
> > The OOM killer?
> 
> Check dmesg, the kernel logs when it OOMs.

I don't remember a OOM message in dmesg. But at the terminal it prints
"Killed" (if it would have been a segfault, it would have printed
Segmentation fault).

I'm not at a location where I can get Internet access over wifi and I've
been using tinymail using the non-mmap patch (causing it to overwrite my
summary files with ones that can't be opened with the mmap code). So I
can't test at this moment. I will probably recheck the dmesg logs
tomorrow or this evening.

I do know that the mmap() and munmap() syscalls don't make the kill
happen. I know this because I mmap() before I know whether the summary
file has the right version or not. It's also detecting an incorrect
version, so it can at least read the first bytes (for sure). So it must
be something that happens while parsing the memory block.

I know I had to use some guint32's, maybe those are differently written
to the file on ARM?

Regretfully I don't have a debugger for the device (afaik) so I can't
check where or what's happening.

Running tinymail with gdb on SDK_ARMEL also isn't working :-(


-- 
Philip Van Hoof, software developer at x-tend 
home: me at pvanhoof dot be 
gnome: pvanhoof at gnome dot org 
work: vanhoof at x-tend dot be 
http://www.pvanhoof.be - http://www.x-tend.be