On 10/3/2019 11:53 AM, Wei-min Pan wrote:
>
> On 10/3/2019 11:31 AM, Simon Marchi wrote:
>> On 2019-10-03 2:21 p.m., Wei-min Pan wrote:
>>> Let's use an example (checking omitted):
>>>
>>> We're replacing:
>>>    name = ctf_type_aname_raw (fp, tid);
>>>    TYPE_NAME (type) = obstack_strdup (&of->objfile_obstack, name);
>>>    free (name);
>>>
>>> with
>>>    gdb::unique_xmalloc_ptr<char> name (ctf_type_aname_raw (fp, tid));
>>>    TYPE_NAME (type) = obstack_strdup (&of->objfile_obstack, name.get ()); 
>>>
>>>
>>> The allocated copy from ctf_type_aname_raw is not freed. Or did I 
>>> miss something?
>>>
>>> Weimin
>> Yes, the second snippet frees the copy returned by ctf_type_aname_raw.
>>
>> It might just be that you are not familiar with the concept of 
>> std::unique_ptr in C++:
>>
>>    https://en.cppreference.com/w/cpp/memory/unique_ptr
>>
>> It's a wrapper around a simple pointer that automatically calls a 
>> deleter function
>> when it goes out of scope.  gdb::unique_xmalloc_ptr is a 
>> specialization of std::unique_ptr
>> that uses xfree as the deleter function.
>>
>> Here's a similar use:
>>
>> https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=blob;f=gdb/fbsd-tdep.c;h=9422e3c1a7e1a65c76b088f42bb5986bd13a089f;hb=HEAD#l1530
>>
>> `fbsd_core_vnode_path` return a pointer to an allocated C String, 
>> which `cwd` wraps.  When
>> it goes out of scope, `cwd` automatically calls `xfree` with the 
>> pointer.
>
> Oh, thanks for the explanation, reference and example. Will switch to 
> using gdb::unique_xmalloc_ptr<> instead.

Hi Simon,

I've switched to using gdb::unique_xmalloc_ptr, please see the attached 
patch, and ran
the CTF test suite successfully. Will submit a revised patch upstream 
shortly.

Thanks again for your reviews and help.

Weimin