On Thu, 16 Oct 2025 at 13:22, Tomasz Kaminski <[email protected]> wrote:
>
>
>
> On Thu, Oct 16, 2025 at 2:13 PM Jonathan Wakely <[email protected]>
wrote:
>>
>> On Thu, 16 Oct 2025 at 12:53, Tomasz Kaminski <[email protected]>
wrote:
>> >
>> >
>> >
>> > On Thu, Oct 16, 2025 at 1:24 PM Jonathan Wakely <[email protected]>
wrote:
>> >>
>> >> With this change locations that don't have a source file look like
this:
>> >>
>> >>   18# __libc_start_call_main at [0x7fd6568f6574]
>> >>   19# __libc_start_main@GLIBC_2.2.5 at [0x7fd6568f6627]
>> >>   20# _start at [0x4006a4]
>> >>
>> >> Instead of:
>> >>
>> >>   18# __libc_start_call_main at :0
>> >>   19# __libc_start_main@GLIBC_2.2.5 at :0
>> >>   20# _start at :0
>> >>
>> >> For a non-empty stacktrace_entry, if the function name returned by
>> >> description() is an empty string we now print "<unknown>" for the
>> >> function name. For an empty (e.g. default constructed)
stacktrace_entry
>> >> the entire string representation is now "<unknown>" instead of an
empty
>> >> string.
>> >>
>> >> Instead of printing "<unknown>" for the function name, we could set
that
>> >> string in the stacktrace_entry::_Info object, so that description()
>> >> returns "<unknown>" and then operator<< wouldn't need to handle an
empty
>> >> description() string. However, returning an empty string from that
>> >> function seems simpler for users to detect, rather than having to
parse
>> >> "<unknown>".
>> >>
>> >> We could also choose a different string for an empty stacktrace_entry,
>> >> maybe "<none>" or "<invalid>", but "<unknown>" seems OK for now.
>> >>
>> >> libstdc++-v3/ChangeLog:
>> >>
>> >>         * include/std/stacktrace
>> >>         (operator<<(ostream&, const stacktrace_entry&)): Improve
output
>> >>         when description() or source_file() returns an empty string,
>> >>         or the stacktrace_entry is invalid.
>> >>         (operator<<(ostream&, const basic_stacktrace<A>&)): Use
>> >>         size_type of the correct specialization.
>> >> ---
>> >>
>> >> v2: Restore fmtflags as pointed out by Nathan. Adjust control flow in
>> >> operator<< as suggested by Tomasz.
>> >>
>> >> Tested x86_64-linux.
>> >>
>> >>  libstdc++-v3/include/std/stacktrace | 29
+++++++++++++++++++++++++----
>> >>  1 file changed, 25 insertions(+), 4 deletions(-)
>> >>
>> >> diff --git a/libstdc++-v3/include/std/stacktrace
b/libstdc++-v3/include/std/stacktrace
>> >> index b9e260e19f89..28ffda76328f 100644
>> >> --- a/libstdc++-v3/include/std/stacktrace
>> >> +++ b/libstdc++-v3/include/std/stacktrace
>> >> @@ -685,11 +685,31 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
>> >>    {
>> >>      string __desc, __file;
>> >>      int __line;
>> >> -    if (__f._M_get_info(&__desc, &__file, &__line))
>> >> +    if (__f._M_get_info(&__desc, &__file, &__line)) [[likely]]
>> >>        {
>> >> -       __os.width(4);
>> >> -       __os << __desc << " at " << __file << ':' << __line;
>> >> +       if (__desc.empty()) [[unlikely]]
>> >> +         __os << "<unknown>";
>> >> +       else
>> >> +         __os << __desc;
>> >> +       __os << string_view(" at ");
>> >> +       if (!__file.empty()) [[likely]]
>> >> +         return __os << __file << ':' << __line;
>> >> +       // else fall through and write native_handle() as the
location.
>> >>        }
>> >> +
>> >> +    if (__f) [[likely]]
>> >> +      {
>> >> +       struct _Flag_guard
>> >> +       {
>> >> +         ios_base& _M_ios;
>> >> +         ios_base::fmtflags _M_f;
>> >> +         ~_Flag_guard() { _M_ios.setf(_M_f); }
>> >> +       };
>> >> +       _Flag_guard _ = { __os, __os.setf(ios::hex, ios::basefield) };
>> >> +       __os << "[0x" << __f.native_handle() << ']';
>> >> +      }
>> >> +    else
>> >> +      __os << "<unknown>";
>> >>      return __os;
>> >
>> > I know that I ask for another change ag, but if (!__f) we will not
ever get info for it, so I would prefer:
>> > if (!__f) [[unlikelly]
>> >   return __os << "<unknown>";
>> >
>> > string __desc, __file;
>> > int __line;
>> > if (__f._M_get_info(&__desc, &__file, &__line)) [[likely]]
>> > {
>> >   // unchanged
>> > }
>> >
>> >  struct _Flag_guard
>> >   {
>> >       ios_base& _M_ios;
>> >       ios_base::fmtflags _M_f;
>> >       ~_Flag_guard() { _M_ios.setf(_M_f); }
>> >     };
>> >     _Flag_guard _ = { __os, __os.setf(ios::hex, ios::basefield) };
>> >     __os << "[0x" << __f.native_handle() << ']';
>> >    return __os;
>> >
>> > This way it is clear, that we never produce <unknown> at <unknown>
>> >
>>
>> OK, that makes sense.
>>
>> Ignoring the details of the logic in the function for now, the output
>> currently looks like:
>>
>>  15# myfunc(int) at /tmp/bt.cc:48
>>  16# myfunc(int) at /tmp/bt.cc:48
>>  17# main at /tmp/bt.cc:61
>>  18# __libc_start_call_main at [0x7f56578d3574]
>>  19# __libc_start_main@GLIBC_2.2.5 at [0x7f56578d3627]
>>  20# _start at [0x400684]
>>
>> We could output the hex address even when we have a filename, e.g.
>>
>>  15# myfunc(int) at /tmp/bt.cc:48 [0x4008b7]
>>  16# myfunc(int) at /tmp/bt.cc:48 [0x4008b7]
>>  17# main at /tmp/bt.cc:61 [0x40091a]
>>  18# __libc_start_call_main [0x7f0682923574]
>>  19# __libc_start_main@GLIBC_2.2.5 [0x7f0682923627]
>>  20# _start [0x400684]
>
> Maybe we could put the hex address at the front, and then pad it
>  15# [0x4008b7] myfunc(int) at /tmp/bt.cc:48
>  16# [0x4008b7] myfunc(int) at /tmp/bt.cc:48
>  17# [0x40091a] main at /tmp/bt.cc:61
>  18# [0x7f0682923574] __libc_start_call_main
>  19# [0x7f0682923627] __libc_start_main@GLIBC_2.2.5
>  20# [0x400684] _start
>
> And then maybe pad the address.

That looks OK for 32-bit targets:

 15# [0x080487f4] myfunc(int) at /tmp/bt.cc:48
 16# [0x080487f4] myfunc(int) at /tmp/bt.cc:48
 17# [0x0804885c] main at /tmp/bt.cc:61
 18# [0xf7944060] __libc_start_call_main
 19# [0xf7944137] __libc_start_main@GLIBC_2.0
 20# [0x08048587] _start

but the padding makes it look very noisy for 64-bit:

 15# [0x00000000004008c7] myfunc(int) at /tmp/bt.cc:48
 16# [0x00000000004008c7] myfunc(int) at /tmp/bt.cc:48
 17# [0x000000000040092a] main at /tmp/bt.cc:61
 18# [0x00007f8dd65e4574] __libc_start_call_main
 19# [0x00007f8dd65e4627] __libc_start_main@GLIBC_2.2.5
 20# [0x0000000000400694] _start

That's assuming we pad with zeros. With spaces (and right-aligned) it looks
a bit strange IMHO:

 15#           [0x400917] myfunc(int) at /tmp/bt.cc:48
 16#           [0x400917] myfunc(int) at /tmp/bt.cc:48
 17#           [0x40097a] main at /tmp/bt.cc:61
 18#     [0x7f900a411574] __libc_start_call_main
 19#     [0x7f900a411627] __libc_start_main@GLIBC_2.2.5
 20#           [0x4006e4] _start


I think I'll push the version that adds the address at the end, unpadded,
and we can revisit it later if we want to.

Reply via email to