Re: [PATCH v5 2/3] Implement "set cwd" command on GDB

[Pedro Alves <palves@redhat.com>]

On 10/03/2017 05:39 PM, Sergio Durigan Junior wrote:

>> But what if inferior_cwd _is_ NULL, when ...
>>
>>
>>>    memset (&si, 0, sizeof (si));
>>>    si.cb = sizeof (si);
>>>  
>>> @@ -2514,6 +2527,10 @@ windows_create_inferior (struct target_ops *ops, const char *exec_file,
>>>        flags |= DEBUG_PROCESS;
>>>      }
>>>  
>>> +  if (cygwin_conv_path (CCP_POSIX_TO_WIN_W, inferior_cwd,
>>> +			infcwd, strlen (inferior_cwd)) < 0)
>>> +    error (_("Error converting inferior cwd: %d"), errno);
>>
>> ... we get here?  It looks to me like this conversion should
>> skipped here then, and ...
> 
> You're right, this should be skipped.
> 
>>> +
>>>  #ifdef __USEWIDE
>>>    args = (cygwin_buf_t *) alloca ((wcslen (toexec) + wcslen (cygallargs) + 2)
>>>  				  * sizeof (wchar_t));
>>> @@ -2574,7 +2591,7 @@ windows_create_inferior (struct target_ops *ops, const char *exec_file,
>>>  		       TRUE,	/* inherit handles */
>>>  		       flags,	/* start flags */
>>>  		       w32_env,	/* environment */
>>> -		       NULL,	/* current directory */
>>> +		       infcwd,	/* current directory */
>>
>> ... here still pass NULL.
> 
> Here we pass NULL because that was the old behaviour when we did not
> care about the inferior's cwd.  This is the same approach that we use on
> fork_inferior: if the user hasn't provided any paths via "set cwd", then
> we don't do anything.

Exactly, I'm saying that you _should_ pass NULL iff
inferior_cwd is NULL.  That's what msdn's documentation of
CreateProcess says you should pass in order to have the
process inherit its parent's cwd.

Thanks,
Pedro Alves