From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Rapoport Date: Fri, 27 Jul 2018 20:36:42 +0000 Subject: Re: [ia64:noboot 4/4] mm/page_alloc.c:321:14: error: 'PAGES_PER_SECTION' undeclared; did you mean 'U Message-Id: <20180727203641.GA17745@rapoport-lnx> List-Id: References: <201807270703.ixR4vo9j%fengguang.wu@intel.com> In-Reply-To: <201807270703.ixR4vo9j%fengguang.wu@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: quoted-printable To: linux-ia64@vger.kernel.org On Fri, Jul 27, 2018 at 07:08:40PM +0000, Luck, Tony wrote: > >> mm/page_alloc.c:321:14: error: 'PAGES_PER_SECTION' undeclared (first u= se in this function); did you mean 'USEC_PER_SEC'? >=20 > Weird. I unpacked that config file and tried to reproduce ... but got a = clean build. >=20 > Maybe my build flipped some other options because I'm using an ancient co= mpiler? > I noticed this difference at the top of the output when I diffed my .conf= ig with the > supplied one: >=20 > < CONFIG_GCC_VERSION=80100 > --- > > CONFIG_GCC_VERSION@304 >=20 > Mike: Did you see anything like this when you did cross builds? =20 Yep, and I think it's not related to the compiler version. Selecting NO_BOOTMEM allows enabling the DEFERRED_STRUCT_PAGE_INIT option that in turn causes compilation of the code that uses PAGES_PER_SECTION. The definition of PAGES_PER_SECTION, though, depends on SPARSEMEM which is not enabled in allmconfig for ia64. I've sent a patch [1] to make DEFERRED_STRUCT_PAGE_INIT depend on SPARSEMEM, it's landed in the akmp tree. [1] https://lore.kernel.org/lkml/1530279308-24988-1-git-send-email-rppt@lin= ux.vnet.ibm.com/#t > -Tony >=20 >=20 --=20 Sincerely yours, Mike.