[Bug translator/6941] better prologue heuristics needed for userspace (non-regparm)

[Bug translator/6941] better prologue heuristics needed for userspace (non-regparm)

[fche at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=6941

--- Comment #2 from Frank Ch. Eigler <fche at redhat dot com> 2012-06-16 16:12:26 UTC ---
*** Bug 10762 has been marked as a duplicate of this bug. ***

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug translator/6941] better prologue heuristics needed for userspace (non-regparm)

[fche at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=6941

Frank Ch. Eigler <fche at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jan.kratochvil at redhat
                   |                            |dot com, mjw at redhat dot
                   |                            |com

--- Comment #3 from Frank Ch. Eigler <fche at redhat dot com> 2012-06-16 16:32:58 UTC ---
With gcc VTA, prologue searching at -O2 is becoming unnecessary.  However, at
-O0 it is becoming essential, for gcc does not run VTA processing at -O0, and
the dwarf location list data it emits is only accurate in the post-prologue
section.  So we need to figure out whether the given target executable was
built with -O0, and then perform proper prologue search heuristics.

To find out -O*, one can look at the DW_AT_producer of the
DW_TAG_compile_unit,; recent gcc emits the compiler option list in there.  We
can look for the absence of -O[123] and/or the presence of -O0, or
-fno-var-tracking.  It's a heuristic, but so there.  See gdb's
dwarf2read.c:process_full_comp_unit() to see its set of related heuristics.

See gdb's symtab.c:skip_prologue_sal() and skip_prologue_using_lineinfo() for
the actual prologue-skipping heuristics, which are similar to our own in
dwflpp.cxx:resolve_prologue_endings().

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug translator/6941] better prologue heuristics needed for userspace (non-regparm)

[jlebon at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=6941

Jonathan Lebon <jlebon at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |jlebon at redhat dot com
         Resolution|---                         |FIXED

--- Comment #4 from Jonathan Lebon <jlebon at redhat dot com> ---
> With gcc VTA, prologue searching at -O2 is becoming unnecessary.  However,
> at -O0 it is becoming essential, for gcc does not run VTA processing at -O0,
> and the dwarf location list data it emits is only accurate in the
> post-prologue section.  So we need to figure out whether the given target
> executable was built with -O0, and then perform proper prologue search
> heuristics.

The code now runs the prologue searching heuristics only for programs that were
not optimized (or for which there may be reasons to believe location info in
their prologues may be unreliable, such as in older GCC versions). See also the
related PR13420.

Commits 277a2b9 and b5b9879.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug translator/6941] better prologue heuristics needed for userspace (non-regparm)

[fche at redhat dot com]

------- Additional Comments From fche at redhat dot com  2008-10-07 14:36 -------
Created an attachment (id=2983)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=2983&action=view)
one possible heuristic

This patch implements one intuitive but insufficient heuristic
for this.  It's saved here for archiving only.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=6941

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.