From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from conssluserg-01.nifty.com ([210.131.2.80]:20187 "EHLO conssluserg-01.nifty.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751681AbdIBHcf (ORCPT ); Sat, 2 Sep 2017 03:32:35 -0400 Received: from mail-yw0-f179.google.com (mail-yw0-f179.google.com [209.85.161.179]) (authenticated) by conssluserg-01.nifty.com with ESMTP id v827WPbB008670 for ; Sat, 2 Sep 2017 16:32:25 +0900 Received: by mail-yw0-f179.google.com with SMTP id c85so3404929ywa.0 for ; Sat, 02 Sep 2017 00:32:25 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <20170818222052.22375-3-nicolasporcel06@gmail.com> References: <20170818222052.22375-1-nicolasporcel06@gmail.com> <20170818222052.22375-3-nicolasporcel06@gmail.com> From: Masahiro Yamada Date: Sat, 2 Sep 2017 16:31:44 +0900 Message-ID: Subject: Re: [PATCH 2/2] kconfig: support out of tree KCONFIG_CONFIG Content-Type: text/plain; charset="UTF-8" Sender: linux-kbuild-owner@vger.kernel.org List-ID: To: Nicolas Porcel Cc: Linux Kbuild mailing list 2017-08-19 7:20 GMT+09:00 Nicolas Porcel : > If KCONFIG_CONFIG is on a different file system than the kernel source, > oldconfig/defconfig will fail due to the use of rename in conf_write. It > will also fail if the kernel source directory is not writable. If the kernel source is not writable, O= should be used. > This patch changes the behavior of conf_write to be more intuitive. > > When no argument is passed to conf_write, KCONFIG_CONFIG is used instead > for the config path. The consequence is that the .tmpconfig.%(pid) is > written in the same directory as KCONFIG_CONFIG. > > Some defaulting logic has been added, allowing the use of a directory > for KCONFIG_CONFIG. In that case, the .config file will be written in > this directory. I think KCONFIG_CONFIG only contains a file name and this restriction is intentional because conf_write() might be called with a directory path (for example from gconf.c) If we change the behavior, I think conf_write() should just error-out when directory path is given. > Signed-off-by: Nicolas Porcel > --- > scripts/kconfig/confdata.c | 35 ++++++++++++++++++++--------------- > 1 file changed, 20 insertions(+), 15 deletions(-) > > diff --git a/scripts/kconfig/confdata.c b/scripts/kconfig/confdata.c > index 297b079ae4d9..3c2b7155a385 100644 > --- a/scripts/kconfig/confdata.c > +++ b/scripts/kconfig/confdata.c > @@ -743,32 +743,37 @@ int conf_write(const char *name) > FILE *out; > struct symbol *sym; > struct menu *menu; > + struct stat st; > const char *basename; > const char *str; > char dirname[PATH_MAX+1], tmpname[PATH_MAX+1], newname[PATH_MAX+1]; > char *env; > + char *slash; > + > + const char *configname = conf_get_configname(); > > dirname[0] = 0; > - if (name && name[0]) { > - struct stat st; > - char *slash; > - > - if (!stat(name, &st) && S_ISDIR(st.st_mode)) { > - strcpy(dirname, name); > - strcat(dirname, "/"); > - basename = conf_get_configname(); > - } else if ((slash = strrchr(name, '/'))) { > + if (!name || !name[0]) > + name = configname; > + > + if (!stat(name, &st) && S_ISDIR(st.st_mode)) { > + strcpy(dirname, name); > + strcat(dirname, "/"); > + if (name == configname || strchr(configname, '/')) > + basename = ".config"; Hmm, I think the intention of conf_get_configname() is to avoid the bare use of ".config" -- Best Regards Masahiro Yamada