On Fri, Jul 3, 2015 at 2:39 PM, Ulf Magnusson <ulfalizer.lkml@xxxxxxxxx> wrote:
> On Fri, Jul 3, 2015 at 1:40 PM, Ulf Magnusson <ulfalizer.lkml@xxxxxxxxx> wrote:
>> On Fri, Jul 3, 2015 at 1:33 PM, Andreas Ruprecht
>> <andreas.ruprecht@xxxxxx> wrote:
>>> On 07/03/2015 12:46, Ulf Magnusson wrote:
>>>>
>>>> Except for scattered accidents like in the original message, which are
>>>> hopefully pretty rare and easy to fix, the only documented thing that depends
>>>> on that lexer sloppiness is the ---help--- "token".
>>>>
>>>> I'd just add "---help---" as another T_HELP alias (or get rid of it altogether,
>>>> but that's probably more work than it's worth). Tightening things up should be
>>>> safe after that.
>>>>
>>>> /Ulf
>>>>
>>>
>>> So we might want to do something like the attached patch, right?
>>>
>>> Unfortunately, when I generate the zconf.{hash,lex,tab}.c files on my
>>> machine, they have some notable differences to the _shipped versions:
>>>
>>> ruprecht@box:linux-next$ diff -u zconf.tab.c zconf.tab.c_shipped
>>> - return yyresult;
>>> + /* Make sure YYID is used. */
>>> + return YYID (yyresult);
>>>
>>> and I can't find any Documentation on how to properly rebuild the
>>> _shipped files... Does anybody have a hint on that?
>>>
>>> Regards,
>>>
>>> Andreas
>>
>> Yup, something like that. Been too long, so I don't remember either,
>> though I don't remember having to do anything special.
>>
>> I'm heading off to bed soon, so probably shouldn't be thinking. Could
>> take a look tomorrow if there's still a problem. :)
>>
>> /Ulf
>
> Or add '---help---' to zconf.gperf I guess, to avoid special-casing it. Would
> need to add '-' to the
>
> n [A-Za-z0-9_]
>
> definition in zconf.l in that case too.
>
> That should be safe provided the gperf part works out, since zconf.y should
> verify that the T_WORD has the expected form. (As a side note, the first token
> on a line is always a keyword, except for within help texts, which are handled
> specially.)
...or that it's not a T_WORD rather, but a command. :P
>
> I guess that might change the meaning of lines like "foo-bar" if '-' is
> introduced as an operator, but that seems kinda academic as 'foo' would
> probably still be a keyword. It's already broken due to the <PARAM> regex
> ({n}|[-/.])+ having '-' in it too.
>
> /Ulf
/Ulf
--
To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
