Re: [PATCH PR gdb/20057] Internal error on trying to set {char[]}$pc="string"

[Wei-min Pan <weimin.pan@oracle.com>]

On 1/24/2018 8:14 PM, Joel Brobecker wrote:
>> To support C99 VLA, function value_from_contents_and_address() was
>> modified to add a call to resolve_dynamic_type(), which in turn
>> calls resolve_dynamic_array() to resolve the dynamic array bounds
>> to static values. But the problem arises when function copy_type(),
>> called by resolve_dynamic_array(), expects the type to be copied
>> to have an associated objfile from which the new type is allocated,
>> or asserts. Since type char[] doesn't have an associated objfile
>> when the following gdb command:
>>
>> (gdb) set {char[]}$pc="hello"
>>
>> was issued, gdb asserts.
>>
>> The gdb_assert (TYPE_OBJFILE_OWNED (type)) line in copy_type() doesn't
>> look necessary or correct since space needed for the new type could be
>> allocated from either the type's objfile if it exists or gdbarch if
>> it doesn't, similar to what alloc_type_copy(), which is called after
>> gdb_assert() in copy_type(), does. Removing gdb_assert() fixes the
>> problem.
> I think removing the assert just shifts the issue elsewhere.
> Basically, you want the lifetime of the new type to match
> the lifetime of the object using it. The gdbarch structure
> has a lifetime that's different from objfiles.

Is there any reason why the gdbarch structure, which won't be freed until
the corresponding architecture is, needs to have a lifetime that matches
the objfiles?

> I happen to have hit the same issue as you, but from an Ada expression,
> and sent it a fix not long ago:
> https://www.sourceware.org/ml/gdb-patches/2018-01/msg00240.html
>
> Does it fix your problem too?
>
Yes, it does fix my problem of gdb asserting on the "set 
{char[]}$pc="hi"" command, as
reported in the PR, but still asserts on a slightly modified "set 
{unsigned char[]}$pc="hi"
command.

Thanks.