This page has a partial list of issues which should be addressed before the 7.0 release. [NAME] at the beginning of each item is the name of a point of contact who will help trying to get the issue fixed faster.
- [Tom Tromey] varobj support for Python pretty-printing:
- Test and document using Python Eggs to distribute and activate pretty-printers.
Optional action items:
- Python support: Support for new parameters, read/write/search inferior memory, symbol handling (tentative).
- If type GC work goes in, change Python's Type object to memoize results.
- Multiprocess support
- (merge the work done in the multiprocess branch)
PR/9174: gdb can't handle PIE
To be reviewed before release (can be reviewed after branching):
Breakpoint commands not executed during script evaluation at startup:
Caz Yokoyama's kgdb patch.
I don't think the most recent one was ever reviewed. http://permalink.gmane.org/gmane.comp.gdb.patches/50989
Watchpoint on an unloaded shared library
gdb.objc/objcdecode.exp test error
Fix breakpoints when several source files have the same name
- Jonas Maebe's calling convention patch, pinged twice.
- Nathan Froyd's internal error patch, pinged once.
- Jan Kratochvil's watchpoint series.
- Sami Wagiaalla's namespace patches.
- Keith's recent dwarf2/c++ patches.
Done:
PR/9747: Quitting during pagination request causes trouble during "next" command.
- Python support:
- Changing some internals to allow printing of strings with embedded NUL characters.
- Support for new commands, convenience functions, inferior frame handling.
Rewrite base varobj/MI patches according to the spec.
- Process Record and Replay
- (apparently generating a lot of interest)
- [EliZ] Add missing configure options to the README (ping Eli about it).
- Inlining support.
[ThiagoB] PR/9723: gdb breakpoints silently fail on PIE binaries
- Solution would be to add a warning stating that PIE binaries are not supported, when GDB detects that it's handling one of those.
- Assert in frame.c:get_frame_arch.
- This assert was inserted to catch any calls to get_frame_arch with a NULL frame. Added in our sources on 2009-01-15. If too close to branch time, we might want to remove it in the branch sources, to reduce the risk.
See: http://www.sourceware.org/ml/gdb-patches/2009-01/msg00285.html (and followup messages). This became OBE since the change was made quite a while ago (Jan 2009) and thus has accumulated sufficient field testing. But this is OBE also because Ulrich re-implemented this function using frame_unwind_arch.
- This assert was inserted to catch any calls to get_frame_arch with a NULL frame. Added in our sources on 2009-01-15. If too close to branch time, we might want to remove it in the branch sources, to reduce the risk.
Build failure on cygwin with "--enable-targets=all --enable-64-bits-bfd". (http://sourceware.org/ml/gdb-patches/2009-08/msg00559.html)
- [ThiagoB] Python support:
Rename python source code and testcase files to avoid 8.3 limitations. Patch: http://www.sourceware.org/ml/gdb-patches/2009-09/msg00149.html. Checked in.
PR/9711: Quadratic slowdown for "where" command.
PR/9786: Typing "info frame" immediately after connecting to a remote target causes an assertion error on x86 GNU/Linux (32 bit).
There is a patch (RFC) that actually fixes this: http://www.sourceware.org/ml/gdb-patches/2009-08/msg00524.html (checked in)
GDB SEGV: http://www.sourceware.org/ml/gdb/2009-09/msg00078.html
- "Speed up find_pc_section" breakage. More info to be added later.
"Darwin breakage": http://www.sourceware.org/ml/gdb/2009-08/msg00056.html
- Actually a consequence of the "Speed up find_pc_section" breakage (see other Action Item above).
Assertion now solved, but lots of complaints still emitted; workaround suggested but may not be liked enough to make it to the tree (see this sub-thread).
=> Need to decide whether the workaround should be applied or not (Update 2009-09-15: No interest so far, so rejected for now).
- Actually a consequence of the "Speed up find_pc_section" breakage (see other Action Item above).
- The catch syscall patches.
- Use external editor in 'commands' command
- PRecord queries should not be defaulted to 'n'.
- This causes trouble when the query is auto-answered, for instance when GDB is driven by Eclipse in GDB/MI mode.