Andrey Borzenkov [Sat, 18 Jan 2014 18:48:04 +0000 (22:48 +0400)]
fix removal of {cpu,machine} links on mingw/msys
At least on Windows 2003 using "ln -s dir1 dir2" in msys shell succeeds,
but results in what looks like hard link. Subsequent "rm -f dir2" (e.g.
during second config.status invocation) fails. Check that we also can
remove link to directory.
Make it more clear in message that we are checking "ln -s".
Andrey Borzenkov [Sat, 18 Jan 2014 17:22:57 +0000 (21:22 +0400)]
fix include loop on MinGW due to libintl.h pulling stdio.h
In file included from ./include/grub/dl.h:23:0,
from grub-core/lib/libgcrypt-grub/cipher/rfc2268.c:3:
./include/grub/list.h:34:18: warning: conflicting types for 'grub_list_push' [en
abled by default]
void EXPORT_FUNC(grub_list_push) (grub_list_t *head, grub_list_t item);
^
./include/grub/symbol.h:68:25: note: in definition of macro 'EXPORT_FUNC'
# define EXPORT_FUNC(x) x
^
In file included from ./include/grub/fs.h:30:0,
from ./include/grub/file.h:25,
from ./grub-core/lib/posix_wrap/stdio.h:23,
from c:\mingw\include\libintl.h:314,
from ./include/grub/i18n.h:33,
from ./include/grub/misc.h:27,
from ./include/grub/list.h:25,
from ./include/grub/dl.h:28,
from grub-core/lib/libgcrypt-grub/cipher/rfc2268.c:3:
./include/grub/partition.h:106:3: note: previous implicit declaration of 'grub_l
ist_push' was here
grub_list_push (GRUB_AS_LIST_P (&grub_partition_map_list),
^
list.h needs just ATTRIBUTE_ERROR from misc.h; split compiler features
into separate file grub/compiler.h and include it instead.
Andrey Borzenkov [Sat, 18 Jan 2014 16:04:11 +0000 (20:04 +0400)]
Use _W64 to detect MinGW W64-32 instead of _FILE_OFFSET_BITS
In 94cee4a4c201bb506377b2c26e072eee8cb19d6f I overlooked that config.h
unconditionally sets _FILE_OFFSET_BITS, so it cannot be used to detect
MinGW W64 environment. It looks like Emacs folks already found
solution; instead of _FILE_OFFSET_BITS use _W64 as suggested in
http://lists.gnu.org/archive/html/emacs-devel/2013-03/msg00723.html
Andrey Borzenkov [Sat, 18 Jan 2014 15:50:54 +0000 (19:50 +0400)]
reintroduce BUILD_LDFLAGS for the cross-compile case
This allows providing separate LDFLAGS for build and host environments, which
are not necessary the same for cross-compile case. In particular, it allows
building host programs statically to not depend on presence of libraries at
run-time (e.g. MinGW DLLs on Windows) while continue to use default dynamic
linking at build time.
Also fix obsolete comments in confgure.ac - we do use different environment
for build and host now.
Colin Watson [Fri, 17 Jan 2014 15:24:50 +0000 (15:24 +0000)]
Prefer more portable test(1) constructs
* util/grub.d/00_header.in (make_timeout): Use && rather than test
-a.
* util/grub.d/10_windows.in: Likewise.
* util/grub.d/10_netbsd.in (netbsd_load_fs_module): Use || rather
than test -o.
* util/grub.d/30_os-prober.in: Use && rather than test -a, and ||
rather than test -o.
Andrey Borzenkov [Sun, 12 Jan 2014 11:29:21 +0000 (15:29 +0400)]
add part_apple to EFI rescue image to fix missing prefix
On Mac rescue image is booted from HFS+ partition, so bootpath looks like
/ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/HD(3,5d1,ca3,0000000000000000,20,0)/EndEntire
grub fails to find device for this path because it cannot scan partition
table. The simplest fix is to add part_apple by default.
Leif Lindholm [Tue, 7 Jan 2014 17:52:50 +0000 (17:52 +0000)]
arm64: set correct length of device path end entry
The length of the Device Path End entry in the grub_linux_boot()
function was incorrectly set to 0. This triggers an assert failure
in debug builds of Tianocore.
Use _FILE_OFFSET_BITS macro to distinguish between native MinGW and
32 bit under MinGW-64. The latter does not require fseeko/ftello
redefinition which it already does in case of _FILE_OFFSET_BITS=64.
use {grub,boot}aa64.efi for boot images on AArch64
According to UEFI 2.4 specification, default boot file name on AArch64
is BOOTAA64.EFI (3.4.1.1 Removable Media Boot Behavior). Also set default
GRUB image name to grubaa64.efi to match it.
Make grub_util_devmapper_part_to_disk and grub_util_find_partition_start
follow the same algorithm to avoid method mismatch. Don't assume
DMRAID- UUID to mean full disk but instead check that mapping is linear.
Colin Watson [Mon, 23 Dec 2013 14:30:35 +0000 (14:30 +0000)]
Don't distribute config.h.
* Makefile.am (platform_HEADERS): Move to ...
(nodist_platform_HEADERS): ... here. Fixes gettext_strings_test
failure when building from a distributed tarball.