Re: frysk mtg minutes 20060829

[Wu Zhou <woodzltc at cn dot ibm dot com>]

Elena Zannoni wrote:
> Stepan: Enable the ppc64 build of Frysk in RHEL5 and Fedora. Add the
> ppc64 java-gnome packages to RHEL5 and Fedora.

Thanks for that.  We are updating our ppc64 box to test this.
> Libunwind Alex: adding remote features: shouldn't be too hard, but not
> being able to see how to offer them.  If you are remote, the user is
> supposed to tell libunwind how to use the information. Possibly export
> some functions that the users can use as callback.  Cagney says that
> the problem has been resolved for ia64. Ia64 uses itanium unwind info,
> which is in the remote process.  Callback to read from remote memory
> were put in place when JeffJ made gdb use libunwind.  Also need for
> callbacks to free remote memory that it allocates.  It now just
> allocates local memory. But this memory will need to be freed
> explicitly once it is allocated remotely.  Alex says that you need to
> cache information (didn't catch that, sorry).

Is there any code available right now? If yes, we would be very happy to have a look at that first.

> Mosberger is out on vacation until the end of august, which it may
> mean he is out until after labor day :-(

That is unfortunate.

BTW, when is your labor day?  We want to have a rough grasp about when most of you are on vacation. 
 At China, we will have a one-week vacation from Oct, 1 to Oct, 10 (that is our national days). 
Most of us will take a break in that period.

> utrace: Roland nothing to report.  Chris: still not able to track down that
> 0xff problem. Will use heavy hammer of adding log messages.

any detail about this 0xff problem?

> Mark: making progress. Trying to cleanup and writing tests. Gave up
> on rawhide, using fc5 as devel platform.

We met some problem on rawhide too.  On ppc64, FryskGui just reports SEGV and doesn't show anything 
else.
> Tim: register window cleanup for floating point registers. Interface
> issues because you need to be able to switch between different ways to
> read the registers (mmx regs).
>
> Andrew: done some workarounds in the code for kernel bugs.
> Mike: done fix for double parent/ghost
>
> Stan: can get dwarf info for current frame. Clean up the patch and
> post it. Then use libunwind to be able to get the dwarf info for other
> frames is next.  Also writing tests for this part of the code.
> Extending on the Java interface for libdwarf that Adam started.
> Not sure how the gui is going to use this code though.
>
> Phil: rhel5, kernel bugs, ibm bug with ppc64 (due to VNC)

Thanks for the excellent problem determination work, Phil!  This isn't a blocker any more.  But I'd 
like to open a bug against the vnc server now.  Which is the proper way to handle this?  on the 
upstream site, or on the sourceware bugzilla?

> Mike: glibc reported bug. Monitor refactoring. It seems to be a real
> memory corruption bug.
>
> Rick: misc bugfixes. eggtray icon fixes.
> working on documentation as well.
>
> Nurdin: evenviewer: accessibility, just figure out where a bug was.
> Then we can write dogtail tests.  Suggest dumping WIP on a cvs branch,
> just to be safe.
>
> frysk bugzilla list: on sware: seems like its going to happen soon.
>
> please get going on the fc6t3 builds before the weekend!!

We will try fc6 test3 once it is out.  If it is stable enough, we will switch our development to that.

Regards
- Wu Zhou