From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sebastien Carlier Date: Sun, 07 Nov 2010 16:11:14 +0100 Subject: [U-Boot] [PATCH] Switch from library archives to partial linking In-Reply-To: <1289139379.13696.12.camel@ptyser-laptop> References: <4CD3DEFC.7010104@gmail.com> <4CD3E7A4.4000109@emk-elektronik.de> <4CD3F58F.8090302@gmail.com> <20101105122336.C9C1A12A81E2@gemini.denx.de> <4CD3FF43.8030009@gmail.com> <4CD56618.4010603@gmail.com> <4CD58E99.6060705@free.fr> <1289139379.13696.12.camel@ptyser-laptop> Message-ID: <4CD6C192.20008@gmail.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Dear Peter, On 11/07/2010 03:16 PM, Peter Tyser wrote: > You shouldn't need to send the patch using "git send-email". The patch > is greater than U-Boot's mailing list limit (100k) and posting the patch > on a website is perfectly acceptable. Also, it shouldn't be necessary > to split the patch into each separate patch's to address each lib. It'd > be a lot of work on Sebastien's part to do this and not break bisection, > and most maintainers can either ack this patch, or probably don't need > to since its more of a build change, not low-level change that a > maintainer has insight into. > That makes a lot of sense. > I had a couple of comments though: > - You need to add your "Signed-of-by: " line to the patch. > Okay. As you may have guessed, I am a first-time git user. > - A patch description illustrating why this approach is better than the > current approach would be appreciated. > Will do. > - You shouldn't be making changes to stuff like CONFIG_CMD_NFS in this > patch. Its unrelated, and should be dealt with in another patch. eg > your patches could be: > 1/2: Fix boards with CONFIG_CMD_NFS but !CONFIG_CMD_NET > 2/2: Switch from library archives to partial linking > Sounds good, will do. > I'm guessing lots of boards will have this same issue. I imagine its > due to include/config_cmd_defaults.h, so maybe if you fix the issue in > that one place all the compile issues will go away. > The generic fix is to include the following lines somewhere at the end of the config.h generated in the mkconfig script: #ifndef CONFIG_CMD_NET # undef CONFIG_CMD_NFS #endif These lines should probable be put in a new header file; would config_checks.h be an ok name for it? I suppose there might be other cases where a module (that is included by default) needs to be excluded when one of its dependencies is disabled. Regards, Sebastien Carlier