[Bug ada/28180] FAIL: gdb.ada/interface.exp: print s (timeout)

["vries at gcc dot gnu.org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=28180

--- Comment #3 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #0)
> With current master I noticed this timeout:
> ...
> (gdb) PASS: gdb.ada/interface.exp: print r
> print s^M
> Multiple matches for s^M
> [0] cancel^M
> [1] s at
> /home/vries/gdb_versions/devel/binutils-gdb.git/gdb/testsuite/gdb.ada/
> interface/foo.adb:20^M
> [2] s at
> /home/vries/gdb_versions/devel/binutils-gdb.git/gdb/testsuite/gdb.ada/
> interface/foo.adb:?^M
> > FAIL: gdb.ada/interface.exp: print s (timeout)
> ...

Doing a bit of investigation, we see:
...
Breakpoint 1, foo () at
/home/vries/gdb_versions/devel/binutils-gdb.git/gdb/testsuite/gdb.ada/interface/foo.adb:22
22         Do_Nothing (R);  -- STOP
(gdb) info locals
r = (x => 1, y => 2, w => 3, h => 4)
<G6b_LAST> = 8
<foo__G6b___U> = 8
s = (x => 1, y => 2, w => 3, h => 4)
s = 0x7fffffffd890
...

So both "s" are local variables.

That makes it easy to find in the dwarf.

We have:
...
 <2><1204>: Abbrev Number: 31 (DW_TAG_variable)
    <1205>   DW_AT_name        : (indirect string, offset: 0x13e6): s.14
    <1209>   DW_AT_type        : <0x1213>
    <120d>   DW_AT_artificial  : 1
    <120d>   DW_AT_location    : 5 byte block: 91 e0 7d 23 18   (DW_OP_fbreg:
-288; DW_OP_plus_uconst: 24)
...
and:
...
 <3><1249>: Abbrev Number: 33 (DW_TAG_variable) 
    <124a>   DW_AT_name        : s
    <124c>   DW_AT_decl_file   : 1
    <124d>   DW_AT_decl_line   : 20
    <124e>   DW_AT_type        : <0x146d>
    <1252>   DW_AT_location    : 6 byte block: 91 e0 7d 23 18 6        
(DW_OP_fbreg: -288; DW_OP_plus_uconst: 24; DW_OP_deref)
...

I suppose the first is the unexpected one.  It has DW_AT_artificial, but is
still put into the symbol table because it has a DW_AT_name.

This patch:
...
diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
index acabee3315f..44327ecbd9a 100644
--- a/gdb/dwarf2/read.c
+++ b/gdb/dwarf2/read.c
@@ -21585,6 +21585,9 @@ new_symbol (struct die_info *die, struct type *type,
struct dwarf2_
cu *cu,
   if (name)
     {
       int suppress_add = 0;
+      attr = dwarf2_attr (die, DW_AT_artificial, cu);
+      if (attr != nullptr)
+       suppress_add = 1;

       if (space)
        sym = space;
...
gets us:
...
$ gdb -q -batch  outputs/gdb.ada/interface/foo -ex "b foo.adb:22" -ex r -ex "p
s"
Breakpoint 1 at 0x404455: file foo.adb, line 22.

Breakpoint 1, foo () at foo.adb:22
22         Do_Nothing (R);  -- STOP
warning: Unknown upper bound, using 1.
$1 = (x => 1, y => 2, w => 3, h => 4)
...
but with the warning because foo__G6b___U (which is another artificial
variable) cannot be evaluated.

Hmm, so ada uses named artifical variables that need to be in the symbol table
in order to evaluate them, but they should be ignored in terms of gdb commands
that investigate source constructs like "p s".  That seems to be the root
cause.

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