This is the mail archive of the
frysk@sources.redhat.com
mailing list for the frysk project.
minutes 2007-01-30
- From: Stan Cox <scox at redhat dot com>
- To: Frysk List <frysk at sourceware dot org>
- Date: Wed, 31 Jan 2007 17:26:32 -0500
- Subject: minutes 2007-01-30
Testing
-All f* commands should have a man page and each command should pass its
tests (nurdin)
Breakpoints
-test with breakpoints and setjmp/longjmp works with FC5 2.6.17 kernel
-crash if stepping into and past a breakpoint
pc is on a trap instruction (breakpoint)
do a single step with no signal (signal 0)
-if signal is delivered then it works
get a sigtrap
pc is at trap + 1
do signal delivery
do a step then Crash
-related to signal delivery problem?
High Level Breakpoints (timoore)
-Inline functions
-given function/line should return multiple addresses
-.debug_info describe where an inlining occurred
-.debug_line describes everywhere that has valid code for that line
-gives you range of addresses for that line
-can describe with a single address or a range
-use case multiple invocations of small inline function on the same
line
-timoore in Westford next week
Core File Analysis (pmuldoon)
-Look at abstract hosts class, call it "deadhost", and make it work
for deadprocs from core files
-Should there be a builder which would iterate over all the sections
of the corefile?
Window and session manager
-continue refactoring (mcvet)
Accessibility
-requires using native code (sami)
-necessary if event viewer will use dogtail testing
-perhaps the event viewer can be unit tested outside of frysk
Symbol Table Lookup (nurdin)
-Changing stack frame code to use libelf instead of libunwind's home
grown version for elf symbol searching