[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAFkk2KR7SndoEGbNvtDGjx71fNkoia7_HEYuHoaQL75Ukp9vmQ@mail.gmail.com>
Date: Fri, 12 Jan 2018 09:47:38 +0100
From: Ulf Magnusson <ulfalizer@...il.com>
To: Masahiro Yamada <yamada.masahiro@...ionext.com>
Cc: "Yann E. MORIN" <yann.morin.1998@...e.fr>,
Linux Kbuild mailing list <linux-kbuild@...r.kernel.org>,
Sam Ravnborg <sam@...nborg.org>, zippel@...ux-m68k.org,
Nicolas Pitre <nicolas.pitre@...aro.org>,
Michal Marek <michal.lkml@...kovi.net>, dirk@...ders.net,
Arnaud Lacombe <lacombar@...il.com>, walch.martin@....de,
Jan Beulich <JBeulich@...e.com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 5/6] kconfig: Don't leak help strings during parsing
On Wed, Jan 10, 2018 at 5:12 PM, Masahiro Yamada
<yamada.masahiro@...ionext.com> wrote:
> 2017-10-09 2:11 GMT+09:00 Ulf Magnusson <ulfalizer@...il.com>:
>> This is just for completeness to get rid of the last memory leak
>> currently generated during parsing for ARCH=x86. The symbol
>> DVB_NETUP_UNIDVB in drivers/media/pci/netup_unidvb/Kconfig currently has
>> two help strings, and we leak the first one.
>
> Looks like it was fixed by
> 561b29e4ec8d0aac7e094f70d649ee4abccdda03
>
>
>
>> Summary from Valgrind on 'menuconfig' (ARCH=x86) before the fix:
>>
>> LEAK SUMMARY:
>> definitely lost: 344,440 bytes in 14,350 blocks
>> ...
>>
>> Summary after the fix:
>>
>> LEAK SUMMARY:
>> definitely lost: 344,376 bytes in 14,349 blocks
>> ...
>>
>> Signed-off-by: Ulf Magnusson <ulfalizer@...il.com>
>> ---
>> scripts/kconfig/zconf.y | 3 +++
>> 1 file changed, 3 insertions(+)
>>
>> diff --git a/scripts/kconfig/zconf.y b/scripts/kconfig/zconf.y
>> index 468ab03..3c9f436 100644
>> --- a/scripts/kconfig/zconf.y
>> +++ b/scripts/kconfig/zconf.y
>> @@ -435,6 +435,9 @@ help_start: T_HELP T_EOL
>>
>> help: help_start T_HELPTEXT
>> {
>> + if (current_entry->help)
>> + /* Weird menu node with two help strings */
>> + free(current_entry->help);
>> current_entry->help = $2;
>> };
>
>
> I applied this, but maybe is it better to warning message?
> zconfprint("warning: ...") or zconf_error()?
>
>
> If you send v2, I will replace this.
Yeah, that's a good idea.
New version at https://marc.info/?l=linux-kernel&m=151573969713437&w=2.
The warning seems more important than the leak, so I rephrased it a
bit.
Cheers,
Ulf
Powered by blists - more mailing lists