[Bug gdb/19361] Invalid comparison functions

[Bug gdb/19361] Invalid comparison functions

[ssbssa at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=19361

Hannes Domani <ssbssa at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ssbssa at sourceware dot org

--- Comment #2 from Hannes Domani <ssbssa at sourceware dot org> ---
Both functions were changed to std::sort, but I'm not sure if the symmetry
requirements were fixed with this.

compare_processes:
https://sourceware.org/git/?p=binutils-gdb.git;a=commitdiff;h=b129dcac882da87767cb863ad8b2572f90e9e5bc

qsort_cmp:
https://sourceware.org/git/?p=binutils-gdb.git;a=commitdiff;h=39ef2f6256737db92f5d60fa201fe0b301bb8100

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug gdb/19361] Invalid comparison functions

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=19361

--- Comment #3 from Tom Tromey <tromey at sourceware dot org> ---
The linux-osdata.c code looks ok now.

It's hard to say if the objfiles.c code is totally correct or not.
Will std::sort compare an object with itself?
Anyway it would be fine to apply an updated version of that patch.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug gdb/19361] Invalid comparison functions

[tetra2005 at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=19361

--- Comment #4 from Yuri Gribov <tetra2005 at gmail dot com> ---
(In reply to Tom Tromey from comment #3)
> The linux-osdata.c code looks ok now.

Yes, looks good except maybe for ...

> Will std::sort compare an object with itself?

... as I've heard rumors that std::sort _can_ compare object to itself (never
seen a proof though).

> It's hard to say if the objfiles.c code is totally correct or not.

I think the code in objfiles.c is still violating ordering axioms. E.g. if
  objfile1->separate_debug_objfile == objfile2
then sort_cmp(sect1, sect2) and sort_cmp(sect2, sect1) will both return false
which is a contradiction. Some implementations of sort return erratic results
in such circumstances.

Sadly I can't rerun my tool on gdb now as it's limited to qsort (std::sort
checking requires instrumentation in compiler...).

-- 
You are receiving this mail because:
You are on the CC list for the bug.