meeting 2006-10-30

meeting 2006-10-30

[Andrew Cagney]

Testing and kernel problems: cmoller, roland
-> reproducer status?
-> other testing

Command line utilities: nprimje, pmuldoon
-> core files
-> name for fcrash?

Source Code Viewer: rmoseley, mcvet, scox
-> high step logic
-> multiple source files
-> PPC 64 backtrace
-> expressions
-> #include

Breakpoints: timoore, mjw

Java-GNOME update: cagney, pmuldoon, mcvet

Re: meeting 2006-10-30

[Stan Cox]

Testing and kernel problems: cmoller, roland

-toString versus string builder (which is a pattern which is useful for
letting
 a client present customized information)  
-testing assistance
 -It would be helpful to provide junit, dejagnu, dogtail tests
 -ibm providing tests
-utrace issues are all fixed
 -gdb uncovers a kernel bug, which is not a utrace bug
 -fix to be backported to FC6

action: Find working kernel for FC5,  (cagney, Nov-03) 
action: test for utrace issue (now fixed): (cmoller, Nov-03)


Command line utilities: nprimje, pmuldoon
 
fcore, fcrash, fstack
-One possibility is to make fcrash (which runs a program until it
 crashes) an option to fstack 
corefiles
-Tim exposing status registers
-Phil testing
-produces core file with appropriate sections
-corefile analysis
-testing is open ended 
-gdb tests by reading back in what it writes out
-elflint might be a possibility for verifying corefiles
fstack
-almost as fast as pstack 
 
action: fcore tests (pmuldoon, Nov-03)
action: expose all registers (timoore, Nov-03)

Source Code Viewer: rmoseley, mcvet, scox

CDT parser
-bugs: highlighting, multiline function declaration 
-parser is now creating the DOM

action: instruction step (mcvet)
action: multiple source files (inserting glue in correct places)
(rmoseley)
-only populate as needed when stack frame is clicked on

Breakpoints: timoore, mjw

Line numbering: can now get line/address correspondence
-Goal is be able to set breakpoint given file/line #
-hpd specifies breakpoint as pattern match
 -folks have tried to kludge gdb to do this
-inua predates frysk 
 
Java-GNOME update: cagney, pmuldoon, mcvet
-Problem fixed! 
-Unit test created!

Release 
-Java NIO is 32 bit.
 -one thing in inua is extended io; we would like to kill the code 
 -can do slicing without having to copy the data
 -(IDEA) project byte buffer onto memory mapped memory instead of
ptrace