linux-kernel - Re: [PATCH 2/2] kconfig: Make comments look different than menus in .config

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 18 Feb 2022 20:55:49 +0200
From:   Ariel Marcovitch <arielmarcovitch@...il.com>
To:     Masahiro Yamada <masahiroy@...nel.org>
Cc:     Linux Kbuild mailing list <linux-kbuild@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 2/2] kconfig: Make comments look different than menus in
 .config

On 18/01/2022 20:25, Masahiro Yamada wrote:
> On Mon, Dec 13, 2021 at 7:01 PM Ariel Marcovitch
> <arielmarcovitch@...il.com> wrote:
>> Currently, the same code that handles menus in the write to .config
>> handles comments as well. That's why comments look exactly like menus in
>> the .config except for the 'end of menu' comments that appear only for
>> menus. This makes sense because sometimes comments are used as sort of
>> submenus. However for the other cases, it looks kinda weird because one
>> might attempt to look for the 'end of menu' for comments as well and be
>> very confused.
>>
>> Make comments look different than menus. For the following:
>> ```kconfig
>> menu "Stuff"
>>
>> config FOO
>>          def_bool y
>>
>> comment "Some comment"
>>
>> config BAR
>>          def_bool n
>>
>> endmenu
>> ```
>>
>> The .config will look like this:
>> ```
>>   #
>>   # Stuff
>>   #
>>   CONFIG_FOO=y
>>
>>   ### Some comment
>>   # CONFIG_BAR is not defined
>>   # end of Stuff
>>
>> ```
>>
>> Signed-off-by: Ariel Marcovitch <arielmarcovitch@...il.com>
>> ---
>>   scripts/kconfig/confdata.c | 14 ++++++++++----
>>   1 file changed, 10 insertions(+), 4 deletions(-)
>>
>> diff --git a/scripts/kconfig/confdata.c b/scripts/kconfig/confdata.c
>> index 9f2c22f46ee0..d3ec1ad67d92 100644
>> --- a/scripts/kconfig/confdata.c
>> +++ b/scripts/kconfig/confdata.c
>> @@ -880,10 +880,16 @@ int conf_write(const char *name)
>>
>>                          if (type == P_MENU || type == P_COMMENT) {
>>                                  str = menu_get_prompt(menu);
>> -                               fprintf(out, "\n"
>> -                                       "#\n"
>> -                                       "# %s\n"
>> -                                       "#\n", str);
>> +
>> +                               if (type == P_MENU)
>> +                                       fprintf(out, "\n"
>> +                                               "#\n"
>> +                                               "# %s\n"
>> +                                               "#\n", str);
>> +                               else
>> +                                       fprintf(out, "\n"
>> +                                               "### %s\n", str);
>> +
>>                                  need_newline = false;
>>                          }
>>                  }
>> --
>> 2.25.1
>>
>
> Since "# CONFIG... is not set" looks like a comment,
> I am not sure if this improves the visibility.

I agree that adding another '#' signs to the real comments doesn't solve 
the real
problem here, being that kconfig uses comments to save actual information

I guess this is for being able to check for a config in shell script 
with [[ -n $CONFIG_FOO ]]?

Although if that's the case, leaving the config empty has the same 
effect, no? And then
we can add a comment to the end of the definition stating that the 
config is unset.
Something like this:

CONFIG_FOO=y
CONFIG_BAR= # is not set

It may break scripts doing something like this:

: ${CONFIG_FOO=?Config FOO must be defined}

But they can be changed to use ':?' instead (which checks for non-zero 
length string
rather than whether the variable is defined or not)

Actually, now that I think of it, it might even be an improvement for 
scripts to be able to tell whether a config isn't defined or whether it 
has an 'n' value

Anyway, I'm absolutely fine with delaying this patch until we find a 
solution

>
> I will not pick up this until I find out what a really good format is.
Thanks!