Re: [PATCH][Preprocessor]patch to fix PR 90581

[Qing Zhao <qing.zhao@oracle.com>]

> On May 30, 2019, at 3:46 PM, David Malcolm <dmalcolm@redhat.com> wrote:
> 
> On Thu, 2019-05-30 at 11:23 -0500, Qing Zhao wrote:
>> Hi,
>> 
>> PR 90581 (provide an option to adjust the maximum depth of nested
>> #include)
>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90581
>> 
>> is to add a new cpp option -fmax-inlcude-depth to set the maximum
>> depth of nested #include.
>> 
>> '-fmax-include-depth=DEPTH'
>>     Set the maximum depth of the nested include.  The default value
>> is
>>     200.
>> 
>> Please check the attached patch.
>> I have done bootstrap and regression test on X86, no any issue.
>> 
>> thanks a lot.
>> 
>> Qing.
>> 
> Thanks for working on this.  It's looking promising, but I agree that a
> param would be better than an option.

I just checked both gcc source code and gcc documentation. 
looks like that 

—param name=value 

is an option for “Options that Control Optimization”.

Since this new option is for preprocessor, I think that it might not be fit in?

let me know if I miss anything here.

> 
> One idea that occurred to me looking at the patch...
> 
>> index 3ee8bc4..480c282 100644
>> --- a/libcpp/directives.c
>> +++ b/libcpp/directives.c
>> @@ -831,7 +831,7 @@ do_include_common (cpp_reader *pfile, enum include_type type)
>>     }
>> 
>>   /* Prevent #include recursion.  */
>> -  if (pfile->line_table->depth >= CPP_STACK_MAX)
>> +  if (pfile->line_table->depth >= CPP_OPTION (pfile, max_include_depth))
>>     cpp_error (pfile, CPP_DL_ERROR, "#include nested too deeply");
> 
> ...a nice usability tweak here would be to give a hint about the new
> param, to give the user an idea on how to increase the limit.

Yes, I agree.
will fix this.

Thanks.

Qing
> 
> Maybe something like:
> 
> cpp_error (pfile, CPP_DL_ERROR,
> 	  "%<#include%> nested too deeply (depth %i);"
> 	  " use %<--param max-include-depth=LIMIT%> to support deeper nesting",
> 	  pfile->line_table->depth);
> 
> (though probably that would be better as a followup "note")
> 
> Dave