From: gang.chen@asianux.com (Chen Gang) Date: Wed, 29 May 2013 10:06:14 +0800 Subject: [PATCH] arm64: compiling issue for allmodconfig, need "asm/parport.h" by "drivers/parport/parport_pc.c" In-Reply-To: <20130528170219.GI17912@arm.com> References: <5177551D.8090007@asianux.com> <20130424082237.GA23397@arm.com> <5177B1F0.10106@asianux.com> <20130424102327.GF21850@mudshark.cambridge.arm.com> <5177BAD0.9000500@asianux.com> <20130424110511.GK23397@arm.com> <5199A453.4080402@asianux.com> <20130521131734.GD27002@arm.com> <519C1539.10908@asianux.com> <519C7484.9020709@asianux.com> <20130528170219.GI17912@arm.com> Message-ID: <51A56296.1090404@asianux.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 05/29/2013 01:02 AM, Catalin Marinas wrote: >> > >> > After check the Kconfig in drivers, we can find quite a few using long >> > list args which related with ARCH. >> > >> > And they are not only depend on ARCHs, but also depend on another >> > things (e.g. VGA_CONSOL also depend on !4xx && !8xx, and another >> > combinations). > Given this dependency is on the negation of that symbol, you just remove > it and have a "depends on HAVE_VGA_CONSOLE". The few architectures that > need it will have to define it (X86, ARM with a few configurations). > it is "depends on !4xx && !8xx && !SPARC && !M68K && !PARISC && !FRV && !SUPERH && !BLACKFIN && !AVR32 && !MN10300 && (!ARM || ARCH_FOOTBRIDGE || ARCH_INTEGRATOR || ARCH_NETWINDER)" So, not only x86, ARM, but also s390, xtensa, alpha, cris, h8300, ia64, mips... (at least, we do not know, they whether need it or not). >> > We also can find another long list for depend on, but not related with >> > ARCH. >> > >> > They seems not quit difficult for reading (in fact, if we separate them >> > into pieces and let them spread into another areas, the reader will not >> > see the whole dependency clearly) >> > >> > For fixing bug, better as local as possible, as independent as possible, >> > it will be OK if only add !ARM64 for PARPORT_PC depend on, and it has >> > no any negative effect. But if we define CONFIG_HAVE_*, we have to let >> > it related with multiple platforms. >> > >> > For design: if can bear current design, better not to touch it, I think >> > current design is still sustainable, and not complex for understanding, >> > so not need redesign. > We usually clean some of these as we hit them, otherwise people forget > about the clean-up later ;). In fact, I have sent another configuration modification about disabling CONFIG_BUG. it is in discussing... We are sure most of architectures do not need CONFIG_BUG, but for VGA_CONSOLE, we can not be sure most of architectures do not need it. For CONFIG_BUG issue, it will make noisy for most of architectures, but for VGA_CONSOLE issue, at least now, it is only effect with arm64. For CONFIG_BUG issue, it seems we can not provide a satisfied fixing if not get rid of it, but for VGA_CONSOLE issue, at least we have a complete fixing way for it. And now for CONFIG_BUG issue, it is still continuing discussing, and it seems that can not get a result quickly. So for VGA_CONSOLE... ;-) Thanks. -- Chen Gang Asianux Corporation