Re: [PATCH v3] checkkconfigsymbols.sh: reimplementation in python

[Valentin Rothberg <valentinrothberg@gmail.com>]

On lun., 2014-09-22 at 10:24 +0200, Paul Bolle wrote:
> Hi Valentin,
> 
> On Mon, 2014-09-22 at 09:43 +0200, Valentin Rothberg wrote:
> > On dim., 2014-09-21 at 23:28 +0200, Paul Bolle wrote:
> > > Valentin Rothberg schreef op zo 21-09-2014 om 21:53 [+0200]:
> > > > Furthermore, it generates false positives (4 of 526 in v3.17-rc1).
> > > 
> > > Curiosity: what are those four false positives?
> > 
> > 1) /arch/cris/kernel/module.c: 	ETRAX_KMALLOCED_MODULES (defined in
> > arch/cris/Kconfig)
> 
> This probably because
>     symb_bare=`echo $symb | sed -e 's/_MODULE//'`
> 
> in the shell script you removed should read (something untested like):
>     symb_bare=`echo $symb | sed -e 's/_MODULE$//'`
> 
> > 2) ./lib/Makefile: TEST_MODULE (defined in lib/Kconfig.debug)
> 
> TEST_MODULE is an awkward name for a Kconfig symbol. My local script has
> it special cased.

I plan to rename this feature in a future patch, since imho it violates
the _MODULE suffix for kernel modules.

> 
> > 3,4) ./include/linux/module.h, ./kernel/module.c: DEBUG_SET_MODULE_RONX
> > (defined in arch/{s390,arm,x86}/Kconfig.debug)
> 
> See above.
> 
> > > > This patch replaces the shell script with an implementation in Python,
> > > > which:
> > > >     (a) detects the same bugs, but does not report false positives
> > > 
> > > Depends a bit on the definition of false positives. Ie, the hit for
> > >     ./arch/sh/kernel/head_64.S:	CACHE_
> > > 
> > > is caused by
> > >      #error preprocessor flag CONFIG_CACHE_... not recognized!
> > > 
> > > Should that line, and similar lines, be changed?
> > 
> > I consider a false positive to actually be defined in Kconfig. The
> > feature in your example does not really apply to the naming convention
> > of Kconfig features ("..."), so that our regex does not match it.
> 
> But your python script does report it, doesn't it?

It does. Regexes are hell :) I will check that and fix it in the next
version of this patch.

> 
> > However, the regex matches "CONFIG_X86_". I shall change the regex to
> > not accept strings ending with "_", so that such cases are not reported.
> 
> > > > +# REGEX EXPRESSIONS
> > > > +OPERATORS = r"&|\(|\)|\||\!"
> > > > +FEATURE = r"\w*[A-Z]{1}\w*"
> > > > +FEATURE_DEF = r"^\s*(menu){,1}config\s+" + FEATURE + r"\s*"
> > > > +EXPR = r"(" + OPERATORS + r"|\s|" + FEATURE + r")*"
> > > > +STMT = r"^\s*(if|select|depends\s+on)\s+" + EXPR
> > > 
> > >                           "depends on" with multiple spaces?
> > > > +
> > > > +# REGEX OBJECTS
> > > > +REGEX_FILE_KCONFIG = re.compile(r"Kconfig[\.\w+\-]*$")
> > > > +REGEX_FILE_SOURCE = re.compile(r"\.[cSh]$")
> 
> New observation: this causes the script to skip text files, shell
> scripts, etc, doesn't it? 

Yes. Do you prefer to cover such files? I just grepped CONFIG_ in
Documentation and think that covering such could improve the quality
there too. I will put this into the next version of the patch.

> > > > +REGEX_FILE_MAKE = re.compile(r"Makefile|Kbuild[\.\w+]*$")
> > > > +REGEX_FEATURE = re.compile(r"(" + FEATURE + r")")
> > > > +REGEX_FEATURE_DEF = re.compile(FEATURE_DEF)
> > > > +REGEX_CPP_FEATURE = re.compile(r"\W+CONFIG_(" + FEATURE + r")[.]*")
> > > 
> > > There are a few uses of "-DCONFIG_[...]" in Makefiles. This will miss
> > > those, won't it? That's not bad, per se, but a comment why you're
> > > skipping those might be nice. Or are those caught too, somewhere else?
> > 
> > I was not aware of such uses, thanks. It seems important to cover them
> > too. Does this prefix has a certain purpose?
> 
> It is, in short, a way to define preprocessor macros from the GCC
> command line (see info gcc).
> 
> > > > +REGEX_KCONFIG_EXPR = re.compile(EXPR)
> > > > +REGEX_KCONFIG_STMT = re.compile(STMT)
> > > > +REGEX_KCONFIG_HELP = re.compile(r"^[\s|-]*help[\s|-]*")
> > > 
> > > Won't "^\s\+(---help---|help)$" do? Might help catch creative variants
> > > of the help statement (we had a few in the past).
> > 
> > Yes, your regex is more robust. Thanks!
> 
> But it seems I should not have escaped the plus. Please check.

Thank you,
 Valentin

> 
> 
> Paul Bolle
>