Re: GDB Tracepoints for the Linux Kernel

[Jim Blandy <jimb at codesourcery dot com>]

Nicholas Mc Guire <der.herr@hofr.at> writes:
> the only thing I imediatly found to fuss about is the gdb.patch - would
> be nice to have this against 6.6 instead of cvs - 6.6 actually fails in
> two file:

Okay, that's not suprising, but good to know.  I wrote that entire
patch on Saturday and Sunday; I've had zero time to even read it over
again, let alone think about porting it to other GDB versions.
Looking over it now I see some pretty bizarre stuff.

The plan is to have unmodified FSF and patched trees in the Trac
Subversion repository.  Then it would be easy to make nightly
snapshots.

That said, I guess there's no reason not to just put the whole tarball
up for download.

> KGDB tracepoints are not quite there yet - but we do hope to get them into
> KGDB once they work - its nice for remote debugging on the embedded 
> targets I guess.

I put in a plug for you and your KGDB stuff at FOSDEM.  I appreciated
the bug fixes in your patch.  I was very tight on time; 45 minutes
saved not looking for a GDB bug because you'd already done so was a
big help.

You should take a look at the changes I made to ax-gdb.c, ax-gdb.h,
ax-general.c, and dwarf2loc.c, and to collect_symbol and its callers
in tracepoint.c.  Unpatched GDB can't find most variables in Linux for
tracepoints, because it can't handle DW_AT_frame_base.

> Aside from that if you have interesting work we always have students 
> looking for some cool stuff to do - let me know.

Well, you'll probably want them to focus on your stuff, but the TODO
file is free for all.  :)