[Bug tapsets/18649] int_arg() misbehaves on x86[_64] for 32-bit uprobe in binary having debuginfo

["mcermak at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=18649

--- Comment #8 from Martin Cermak <mcermak at redhat dot com> ---
So, SYSTEMTAP_DEBUGINFO doesn't seem to work for built-in debuginfo. So I
attepmted to "trace" how -P works and disabled prologue searching in the
source:

=======
diff --git a/tapsets.cxx b/tapsets.cxx
index 2a50fbd..463dd09 100644
--- a/tapsets.cxx
+++ b/tapsets.cxx
@@ -1719,7 +1719,7 @@ query_addr(Dwarf_Addr addr, dwarf_query *q)
     {
       if (!dw.die_entrypc(fnscope, &addr))
         return;
-      if (dwarf_tag(fnscope) == DW_TAG_subprogram &&
+      if (false && dwarf_tag(fnscope) == DW_TAG_subprogram && // XXXYYY
           (q->sess.prologue_searching ||
            (q->has_process && !q->dw.has_valid_locs()))) // PR 6871 && PR 6941
         {
@@ -2210,7 +2210,7 @@ query_cu (Dwarf_Die * cudie, dwarf_query * q)
       if (rc != DWARF_CB_OK)
         q->query_done = true;

-      if (!q->filtered_functions.empty() &&
+      if (false && !q->filtered_functions.empty() && // XXXYYY
           !q->has_statement_str && // PR 2608
            (q->sess.prologue_searching ||
             (q->has_process && !q->dw.has_valid_locs()))) // PR 6871 && PR
6941
=======

This seems to help. So would it make sense if I'd try to add that tri-state
long option as suggested in Comment #5, that would drive prologue searching at
these two places?

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