[Bug python/30489] New: gdb.Type.template_argument issue when using __int128 unsigned

[Bug python/30489] New: gdb.Type.template_argument issue when using __int128 unsigned

[werwolv98 at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30489

            Bug ID: 30489
           Summary: gdb.Type.template_argument issue when using __int128
                    unsigned
           Product: gdb
           Version: 13.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: python
          Assignee: unassigned at sourceware dot org
          Reporter: werwolv98 at gmail dot com
  Target Milestone: ---

Using the default libstdc++ pretty printer that ships with GCC 13.1, GDB fails
to parse the template types of `std::variant` (and others) as soon as one of
the template types used is a `__int128 unsigned`.

The pretty printer seems to use `Type.template_argument(n)` under the hood
which throws a `RuntimeError: syntax error` if one of the types is a unsigned
int128 type. This leads to GDB not printing the template types and acts as if
the variant contains no value.


```
std::variant<__uint128_t, int> a;
std::variant<__int128_t, int> b;
```

```
(gdb) print a
$1 = std::variant<> [no contained value]
(gdb) print b
$2 = std::variant<__int128, int> [index 0] = {0}

```

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

[Bug python/30489] gdb.Type.template_argument issue when using __int128 unsigned

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=30489

Tom Tromey <tromey at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-05-25
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
                 CC|                            |tromey at sourceware dot org

--- Comment #1 from Tom Tromey <tromey at sourceware dot org> ---
(In reply to WerWolv from comment #0)

> The pretty printer seems to use `Type.template_argument(n)` under the hood
> which throws a `RuntimeError: syntax error` if one of the types is a
> unsigned int128 type.

Can you post the Python stack trace for this?

Anyway I was able to reproduce the behavior, even with git master --
eliminating my theory that maybe the recent 128-bit work fixed this.

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

[Bug python/30489] gdb.Type.template_argument issue when using __int128 unsigned

[werwolv98 at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30489

--- Comment #2 from WerWolv <werwolv98 at gmail dot com> ---
(In reply to Tom Tromey from comment #1)
> (In reply to WerWolv from comment #0)
> 
> > The pretty printer seems to use `Type.template_argument(n)` under the hood
> > which throws a `RuntimeError: syntax error` if one of the types is a
> > unsigned int128 type.
> 
> Can you post the Python stack trace for this?
> 
> Anyway I was able to reproduce the behavior, even with git master --
> eliminating my theory that maybe the recent 128-bit work fixed this.

I wrote a small test script that simply tries to print the first template type
name as value of the type. The stacktrace for that is
```
Traceback (most recent call last):
  File "<string>", line 15, in to_string
  File "<string>", line 5, in get_template_arg_list
RuntimeError: syntax error
```

The only place I could find "syntax error" in the codebase was in the parser so
I assume it's comming from there somewhere.

With the original libstdc++ pretty printer you don't get any stacktrace at all
because they use a try-except in their get_template_arg_list as terminating
condition which fails on the first try so it just returns an empty list of
template types

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

[Bug python/30489] gdb.Type.template_argument issue when using __int128 unsigned

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=30489

--- Comment #3 from Tom Tromey <tromey at sourceware dot org> ---
Ok, thanks, I think I see the problem now.

The "syntax error" means that gdb is falling back on the
old-style name-parsing approach to extracting template
parameters.  This method shouldn't really be needed with
contemporary compilers, though.  So digging a bit deeper
& looking at the DWARF, the type of a variant in my test
program is:

 <2><a73>: Abbrev Number: 56 (DW_TAG_class_type)
    <a74>   DW_AT_name        : (indirect string, offset: 0x1216):
variant<__int128 unsigned, char const*>
[...]
 <3><c09>: Abbrev Number: 12 (DW_TAG_GNU_template_parameter_pack)
    <c0a>   DW_AT_name        : (indirect string, offset: 0x21b1): _Types
 <4><c0e>: Abbrev Number: 3 (DW_TAG_template_type_param)
    <c0f>   DW_AT_type        : <0x14a9>
 <4><c13>: Abbrev Number: 3 (DW_TAG_template_type_param)
    <c14>   DW_AT_type        : <0x1532>


I suspect what is going on is that the Python code
doesn't understand parameter packs; or that maybe these
aren't registered with the type in a usable way.

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

[Bug python/30489] gdb.Type.template_argument issue when using __int128 unsigned

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=30489

--- Comment #4 from Tom Tromey <tromey at sourceware dot org> ---
See bug #17272

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