Re: Recent commit in binutils-gdb

[Tom de Vries <tdevries@suse.de>]

On 30-04-2020 15:24, Andrew Burgess wrote:
> * Tom de Vries <tdevries@suse.de> [2020-04-30 07:49:56 +0200]:
> 
>> On 30-04-2020 06:10, Sterling Augustine via Gdb-patches wrote:
>>> The commit was proposed here:
>>>
>>> https://sourceware.org/pipermail/gdb/2020-March/000012.html
>>>
>>
>> With reproducer:
>> ...
>> $ cat b.cpp
>> namespace hide { struct Bar {  int y; }; }
>> struct Foo {  hide::Bar* b; };
>>
>> inline void x() { hide::Bar b; }
>>
>> static Foo aFoo;
>>
>> void * f() {
>>   return aFoo.b;
>> }
>>
>> int main() {
>>   f();
>>   return 0;
>> }
>> $ g++ -fdebug-types-section b.cpp -g
>> $ ./build/gdb/gdb -ex "b f" -ex run -ex "ptype hide::Bar" ./a.out
>> ...
>>
>> I can reproduce this problem before the recent commit 770479f223e "gdb:
>> Fix toplevel types with -fdebug-types-section", but not after.
>>
>> Also, when I revert commit 770479f223e, and leave this one in place, it
>> doesn't fix the problem. That seems to be because the proposed commit:
>> ...
>> @@ -10954,6 +10954,7 @@ dwarf2_cu::setup_type_unit_groups (struct
>> die_info *die)
>>                         compunit_language (cust),
>>                         0, cust));
>>
>> +      list_in_scope = get_builder ()->get_file_symbols ();
>>        auto &file_names = line_header->file_names ();
>>        for (i = 0; i < file_names.size (); ++i)
>>         {
>> ...
>> and the actual commit differ (note the different in context before):
>> ...
>> @@ -10942,6 +10942,7 @@ dwarf2_cu::setup_type_unit_groups (struct
>> die_info *die)
>>         = XOBNEWVEC (&COMPUNIT_OBJFILE (cust)->objfile_obstack,
>>                      struct symtab *, line_header->file_names_size ());
>>
>> +      list_in_scope = get_builder ()->get_file_symbols ();
>>        auto &file_names = line_header->file_names ();
>>        for (i = 0; i < file_names.size (); ++i)
>>         {
>> ...
>>
>> In summary, the problem was fixed by an earlier commit, and this commit
>> was misapplied. I think we should probably revert this one.
> 
> FWIW I agree.

Andrew, thanks for the input,  I've now reverted the patch.

Guys, my apologies for how this went.  A patch for this problem was
approved in Januari, but lingered around uncommitted, leading to
duplicated work.  We'll try to do better.

Thanks,
- Tom