Skip to contentSkip to conversation

Add native win32 mkimage support

RFC PATCH New Peter Tyser 4 participants 11 messages Feb 6, 2009 – Feb 28, 2009 Kbuild

lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Add a mkimage_win32.exe build target which can produce a native win32 mkimage executable using the MinGW toolchain. The mkimage_win32.exe binary is generated when the MINGW_COMPILE environment variable is defined. The mkimage_win32.exe binary can be used by those who use Windows as an OS build environment but don't use cygwin.

---
I'm not sure what the odds of this being picked up, but I thought others might find this useful.

 Makefile              |    5 ++-
 README                |    6 +++
 include/image.h       |    2 +
 include/libfdt_env.h  |    6 +++
 include/linux/types.h |    5 ++-
 tools/Makefile        |   13 ++++++-
 tools/mingw_support.c |   95 +++++++++++++++++++++++++++++++++++++++++++++++++
 tools/mingw_support.h |   48 +++++++++++++++++++++++++
 tools/mkimage.c       |    4 ++
 tools/mkimage.h       |    5 +++
 10 files changed, 185 insertions(+), 4 deletions(-)
 create mode 100644 tools/mingw_support.c
 create mode 100644 tools/mingw_support.h
Makefile +3 -2
index 84e6125..8fb59de 100644
@@ -3447,8 +3447,9 @@ clean:
 	       $(obj)tools/envcrc					  \
 	       $(obj)tools/gdb/{astest,gdbcont,gdbsend}			  \
 	       $(obj)tools/gen_eth_addr    $(obj)tools/img2srec		  \
-	       $(obj)tools/mkimage	   $(obj)tools/mpc86x_clk	  \
-	       $(obj)tools/ncb		   $(obj)tools/ubsha1
+	       $(obj)tools/mkimage	   $(obj)tools/mkimage_win32.exe  \
+	       $(obj)tools/mpc86x_clk	   $(obj)tools/ncb		  \
+	       $(obj)tools/ubsha1					  \
 	@rm -f $(obj)board/cray/L1/{bootscript.c,bootscript.image}	  \
 	       $(obj)board/netstar/{eeprom,crcek,crcit,*.srec,*.bin}	  \
 	       $(obj)board/trab/trab_fkt   $(obj)board/voiceblue/eeprom   \
README +6 -0
index 522471c..96b8db7 100644
@@ -2695,6 +2695,12 @@ necessary. For example using the ELDK on a 4xx CPU, please enter:
 	$ CROSS_COMPILE=ppc_4xx-
 	$ export CROSS_COMPILE
 
+Note: If you wish to generate a WIN32 version of the mkimage tool
+      using the MinGW toolchain (see http://www.mingw.org) you must
+      also set the MINGW_COMPILE environment variable, for example:
+
+	$ export MINGW_COMPILE=i586-mingw32msvc-
+
 U-Boot is intended to be simple to build. After installing the
 sources you must configure U-Boot for one specific board type. This
 is done by typing:
include/image.h +2 -0
index 4609200..ec6bb1a 100644
@@ -34,7 +34,9 @@
 #define __IMAGE_H__
 
 #if USE_HOSTCC
+#ifndef __MINGW32__
 #include <endian.h>
+#endif
 
 /* new uImage format support enabled on host */
 #define CONFIG_FIT		1
include/libfdt_env.h +6 -0
index ea474a5..bdc7a58 100644
@@ -24,8 +24,10 @@
 #ifdef USE_HOSTCC
 #include <stdint.h>
 #include <string.h>
+#ifndef __MINGW32__
 #include <endian.h>
 #include <byteswap.h>
+#endif /* __MINGW32__ */
 #else
 #include <linux/string.h>
 #include <linux/types.h>
@@ -36,6 +38,10 @@
 extern struct fdt_header *working_fdt;  /* Pointer to the working fdt */
 
 #if __BYTE_ORDER == __LITTLE_ENDIAN
+#ifdef __MINGW32__
+extern uint32_t bswap_32(uint32_t x);
+extern uint64_t bswap_64(uint64_t x);
+#endif
 #define fdt32_to_cpu(x)		bswap_32(x)
 #define cpu_to_fdt32(x)		bswap_32(x)
 #define fdt64_to_cpu(x)		bswap_64(x)
include/linux/types.h +4 -1
index 1b0b4a4..90c0695 100644
@@ -10,10 +10,13 @@
 
 #ifndef __KERNEL_STRICT_NAMES
 
+#ifndef __MINGW32__
+/* prevent mingw overlaps for certain typedefs */
 typedef __kernel_fd_set		fd_set;
 typedef __kernel_dev_t		dev_t;
 typedef __kernel_ino_t		ino_t;
 typedef __kernel_mode_t		mode_t;
+#endif
 typedef __kernel_nlink_t	nlink_t;
 typedef __kernel_off_t		off_t;
 typedef __kernel_pid_t		pid_t;
@@ -54,7 +57,7 @@ typedef __kernel_loff_t		loff_t;
 typedef __kernel_size_t		size_t;
 #endif
 
-#ifndef _SSIZE_T
+#if !defined(_SSIZE_T) && !defined(__MINGW32__)
 #define _SSIZE_T
 typedef __kernel_ssize_t	ssize_t;
 #endif
tools/Makefile +12 -1
index 9e9ee15..64d4a3f 100644
@@ -21,7 +21,8 @@
 # MA 02111-1307 USA
 #
 
-BIN_FILES	= img2srec$(SFX) mkimage$(SFX) envcrc$(SFX) ubsha1$(SFX) gen_eth_addr$(SFX) bmp_logo$(SFX)
+BIN_FILES	= img2srec$(SFX) mkimage$(SFX) mkimage_win32.exe envcrc$(SFX) \
+		ubsha1$(SFX) gen_eth_addr$(SFX) bmp_logo$(SFX)
 
 OBJ_LINKS	= env_embedded.o crc32.o md5.o sha1.o image.o
 OBJ_FILES	= img2srec.o mkimage.o envcrc.o ubsha1.o gen_eth_addr.o bmp_logo.o
@@ -39,6 +40,9 @@ endif
 
 LIBFDT_OBJ_FILES	= $(obj)fdt.o $(obj)fdt_ro.o $(obj)fdt_rw.o $(obj)fdt_strerror.o $(obj)fdt_wip.o
 
+MKIMAGE_WIN32_SRC_FILES	= $(src)mkimage.c $(src)mingw_support.c $(obj)crc32.c \
+			$(obj)image.c $(obj)md5.c $(obj)sha1.c $(LIBFDT_OBJ_FILES:%.o=%.c)
+
 LOGO_H	= $(OBJTREE)/include/bmp_logo.h
 
 ifeq ($(LOGO_BMP),)
@@ -151,6 +155,13 @@ $(obj)mkimage$(SFX):	$(obj)mkimage.o $(obj)crc32.o $(obj)image.o $(obj)md5.o $(o
 		$(CC) $(CFLAGS) $(HOST_LDFLAGS) -o $@ $^
 		$(STRIP) $@
 
+# Build native win32 mkimage if mingw compiler is specified
+ifneq ($(MINGW_COMPILE),)
+$(obj)mkimage_win32.exe: $(MKIMAGE_WIN32_SRC_FILES)
+		$(MINGW_COMPILE)gcc $(FIT_CFLAGS) $(HOST_LDFLAGS) -o $@ $^
+		$(MINGW_COMPILE)strip $@
+endif
+
 $(obj)ncb$(SFX):	$(obj)ncb.o
 		$(CC) $(CFLAGS) $(HOST_LDFLAGS) -o $@ $^
 		$(STRIP) $@
tools/mingw_support.c new file +95 -0

95 lines not on this page · raw message

tools/mingw_support.h new file +48 -0

48 lines not on this page · raw message

tools/mkimage.c +4 -0
index 967fe9a..380b535 100644
@@ -25,6 +25,10 @@
 #include "mkimage.h"
 #include <image.h>
 
+#ifdef __MINGW32__
+#include "mingw_support.h"
+#endif
+
 extern int errno;
 
 #ifndef MAP_FAILED
tools/mkimage.h +5 -0
index a2d5248..6974fa4 100644
@@ -28,7 +28,11 @@
 #ifndef __WIN32__
 #include <netinet/in.h>		/* for host / network byte order conversions	*/
 #endif
+#ifdef __MINGW32__
+#include <stdint.h>
+#else
 #include <sys/mman.h>
+#endif
 #include <sys/stat.h>
 #include <time.h>
 #include <unistd.h>
@@ -65,6 +69,7 @@ typedef unsigned int __u32;
 typedef		unsigned char	uint8_t;
 typedef		unsigned short	uint16_t;
 typedef		unsigned int	uint32_t;
+typedef		unsigned long	ulong;
 
 #define     ntohl(a)	SWAP_LONG(a)
 #define     htonl(a)	SWAP_LONG(a)
Signature · 1 line
1.6.0.2.GIT
#1 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Dear Peter Tyser,

Quoted 6 lines · In message <[email protected]> you wrote:

Add a mkimage_win32.exe build target which can produce a native win32 mkimage executable using the MinGW toolchain. The mkimage_win32.exe binary is generated when the MINGW_COMPILE environment variable is defined. The mkimage_win32.exe binary can be used by those who use Windows as an OS build environment but don't use cygwin.

Why do we need a new envrironment variable? Can we not follow the standard pattern ansd use CROSS_COMPILE instead?

README necessary. For example using the ELDK on a 4xx CPU, please enter: · line 2702

Quoted 5 lines
README +5 -0
+Note: If you wish to generate a WIN32 version of the mkimage tool
+      using the MinGW toolchain (see http://www.mingw.org) you must
+      also set the MINGW_COMPILE environment variable, for example:
+
+	$ export MINGW_COMPILE=i586-mingw32msvc-

Why cannot we use CROSS_COMPILE=i586-mingw32msvc- ? YOu can test for *mingw* easily, I think.

include/image.h line 39

Quoted 4 lines

#if USE_HOSTCC

include/image.h +2 -0
+#ifndef __MINGW32__
 #include <endian.h>
+#endif

How does MinGW handle byte order issues, then?

include/libfdt_env.h line 30

Quoted 10 lines
include/libfdt_env.h +2 -0
@@ -24,8 +24,10 @@
 #ifdef USE_HOSTCC
 #include <stdint.h>
 #include <string.h>
+#ifndef __MINGW32__
 #include <endian.h>
 #include <byteswap.h>
+#endif /* __MINGW32__ */

Ditto - how is byte order handled?

include/libfdt_env.h line 32

Quoted 2 lines
include/libfdt_env.h +2 -0
+# Build native win32 mkimage if mingw compiler is specified
+ifneq ($(MINGW_COMPILE),)

You can test the content of CROSS_COMPILE here, too?

include/libfdt_env.h

Quoted 2 lines
include/libfdt_env.h +2 -0
+/*
+ * Copyright 2008 Extreme Engrineering Solutions, Inc.

Is this really the name of that company?

tools/mingw_support.h line 2

Quoted 8 lines
tools/mingw_support.h new file +2 -0
new file mode 100644
index 0000000..568d54d
@@ -0,0 +1,48 @@
+/*
+ * Copyright 2008 Extreme Engrineering Solutions, Inc.

Ditto?

Best regards,

Wolfgang Denk

Signature · 6 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Extended Epstein-Heisenberg Principle: In an R & D orbit, only  2  of
the  existing 3 parameters can be defined simultaneously. The parame-
ters are: task, time and resources ($).
#2 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Quoted 10 lines · Wolfgang Denk

Dear Peter Tyser,

Quoted 6 lines · In message <[email protected]> you wrote:

Add a mkimage_win32.exe build target which can produce a native win32 mkimage executable using the MinGW toolchain. The mkimage_win32.exe binary is generated when the MINGW_COMPILE environment variable is defined. The mkimage_win32.exe binary can be used by those who use Windows as an OS build environment but don't use cygwin.

Quoted 2 lines

Why do we need a new envrironment variable? Can we not follow the standard pattern ansd use CROSS_COMPILE instead?

Most files can't be compiled with MinGW so if CROSS_COMPILE is set to a MinGW compile errors quickly occur in the build process. Also, it would be a bigger pain to compile u-boot + mkimage_win32.exe. Eg:

export CROSS_COMPILE=powerpc-linux-
export MINGW_COMPILE=i586-mingw32msvc-
make

vs

export CROSS_COMPILE=i586-mingw32msvc-
make
cp tools/mkimage_win32.exe ~/
make clean
export CROSS_COMPILE=powerpc-linux-
make

Quoted 9 lines

README necessary. For example using the ELDK on a 4xx CPU, please enter: · line 2702

Quoted 5 lines
README +5 -0
+Note: If you wish to generate a WIN32 version of the mkimage tool
+      using the MinGW toolchain (see http://www.mingw.org) you must
+      also set the MINGW_COMPILE environment variable, for example:
+
+	$ export MINGW_COMPILE=i586-mingw32msvc-

Why cannot we use CROSS_COMPILE=i586-mingw32msvc- ? YOu can test for *mingw* easily, I think.

Quoted 6 lines

include/image.h line 39

Quoted 4 lines

#if USE_HOSTCC

include/image.h +2 -0
+#ifndef __MINGW32__
 #include <endian.h>
+#endif

How does MinGW handle byte order issues, then?

bswap_XX functions were added in mingw_support.c.

Quoted 18 lines

include/libfdt_env.h line 30

Quoted 10 lines
include/libfdt_env.h +2 -0
@@ -24,8 +24,10 @@
 #ifdef USE_HOSTCC
 #include <stdint.h>
 #include <string.h>
+#ifndef __MINGW32__
 #include <endian.h>
 #include <byteswap.h>
+#endif /* __MINGW32__ */

Ditto - how is byte order handled?

include/libfdt_env.h line 32

Quoted 2 lines
include/libfdt_env.h +2 -0
+# Build native win32 mkimage if mingw compiler is specified
+ifneq ($(MINGW_COMPILE),)

You can test the content of CROSS_COMPILE here, too?

Quoted 4 lines

include/libfdt_env.h

Quoted 2 lines
include/libfdt_env.h +2 -0
+/*
+ * Copyright 2008 Extreme Engrineering Solutions, Inc.

Is this really the name of that company?

Damn fingers...

Thanks,
Peter

#3 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Quoted 20 lines · On Friday 06 February 2009 14:59:10 Peter Tyser wrote:
Quoted 10 lines · Wolfgang Denk
Quoted 6 lines · In message Peter Tyser

Add a mkimage_win32.exe build target which can produce a native win32 mkimage executable using the MinGW toolchain. The mkimage_win32.exe binary is generated when the MINGW_COMPILE environment variable is defined. The mkimage_win32.exe binary can be used by those who use Windows as an OS build environment but don't use cygwin.

Why do we need a new envrironment variable? Can we not follow the standard pattern ansd use CROSS_COMPILE instead?

Most files can't be compiled with MinGW so if CROSS_COMPILE is set to a MinGW compile errors quickly occur in the build process. Also, it would be a bigger pain to compile u-boot + mkimage_win32.exe. Eg:

export CROSS_COMPILE=powerpc-linux-
export MINGW_COMPILE=i586-mingw32msvc-
make

Quoted 8 lines

vs

export CROSS_COMPILE=i586-mingw32msvc-
make
cp tools/mkimage_win32.exe ~/
make clean
export CROSS_COMPILE=powerpc-linux-
make

i think you're trying too hard. you're trying to combine three targets at
once: the host system (ppc), the build system (i686 prob), and canadian cross for some other random build system (mingw).

the logical step would be to only build the tools for your random other build system instead of trying to combine it all together. in other words, set the HOSTCC to your mingw compiler and only process the tools target. this is how every other package out there works and it doesnt really make sense to have u- boot be different in this regard.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20090206/810accee/attachment.pgp

#4 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Quoted 34 lines · Mike Frysinger
Quoted 20 lines · On Friday 06 February 2009 14:59:10 Peter Tyser wrote:
Quoted 10 lines · Wolfgang Denk
Quoted 6 lines · In message Peter Tyser

Add a mkimage_win32.exe build target which can produce a native win32 mkimage executable using the MinGW toolchain. The mkimage_win32.exe binary is generated when the MINGW_COMPILE environment variable is defined. The mkimage_win32.exe binary can be used by those who use Windows as an OS build environment but don't use cygwin.

Why do we need a new envrironment variable? Can we not follow the standard pattern ansd use CROSS_COMPILE instead?

Most files can't be compiled with MinGW so if CROSS_COMPILE is set to a MinGW compile errors quickly occur in the build process. Also, it would be a bigger pain to compile u-boot + mkimage_win32.exe. Eg:

export CROSS_COMPILE=powerpc-linux-
export MINGW_COMPILE=i586-mingw32msvc-
make

Quoted 8 lines

vs

export CROSS_COMPILE=i586-mingw32msvc-
make
cp tools/mkimage_win32.exe ~/
make clean
export CROSS_COMPILE=powerpc-linux-
make

i think you're trying too hard. you're trying to combine three targets at
once: the host system (ppc), the build system (i686 prob), and canadian cross for some other random build system (mingw).

Quoted 5 lines

the logical step would be to only build the tools for your random other build system instead of trying to combine it all together. in other words, set the HOSTCC to your mingw compiler and only process the tools target. this is how every other package out there works and it doesnt really make sense to have u- boot be different in this regard.

But Windows isn't really the HOSTCC, its a cross-compile target. Setting my HOSTCC to a cross compiler seems just as confusing as setting up 2 cross-compile targets to me. The process of building would also still be slightly more complex to me:

export CROSS_COMPILE=powerpc-linux-
export MINGW_COMPILE=i586-mingw32msvc-
make

vs

export HOSTCC=i586-mingw32msvc-
make mkimage_win32
export HOSTCC=gcc
make

Best,
Peter

#5 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Quoted 41 lines · On Friday 06 February 2009 17:01:43 Peter Tyser wrote:
Quoted 34 lines · Mike Frysinger
Quoted 20 lines · On Friday 06 February 2009 14:59:10 Peter Tyser wrote:
Quoted 10 lines · Wolfgang Denk
Quoted 6 lines · In message Peter Tyser

Add a mkimage_win32.exe build target which can produce a native win32 mkimage executable using the MinGW toolchain. The mkimage_win32.exe binary is generated when the MINGW_COMPILE environment variable is defined. The mkimage_win32.exe binary can be used by those who use Windows as an OS build environment but don't use cygwin.

Why do we need a new envrironment variable? Can we not follow the standard pattern ansd use CROSS_COMPILE instead?

Most files can't be compiled with MinGW so if CROSS_COMPILE is set to a MinGW compile errors quickly occur in the build process. Also, it would be a bigger pain to compile u-boot + mkimage_win32.exe. Eg:

export CROSS_COMPILE=powerpc-linux-
export MINGW_COMPILE=i586-mingw32msvc-
make

Quoted 8 lines

vs

export CROSS_COMPILE=i586-mingw32msvc-
make
cp tools/mkimage_win32.exe ~/
make clean
export CROSS_COMPILE=powerpc-linux-
make

i think you're trying too hard. you're trying to combine three targets at once: the host system (ppc), the build system (i686 prob), and canadian cross for some other random build system (mingw).

Quoted 5 lines

the logical step would be to only build the tools for your random other build system instead of trying to combine it all together. in other words, set the HOSTCC to your mingw compiler and only process the tools target. this is how every other package out there works and it doesnt really make sense to have u- boot be different in this regard.

Quoted 3 lines

But Windows isn't really the HOSTCC, its a cross-compile target. Setting my HOSTCC to a cross compiler seems just as confusing as setting up 2 cross-compile targets to me.

when you're only building the tools, there is no native code execution nor building for the target. so the HOSTCC represents the system where the tools compiled will be run.

Quoted 1 line

The process of building would also still be slightly more complex to me:

that's just too bad then imo. what you're prosing is carrying system-specific cruft in the build system so you can avoid running make once more. that's not a scalable system and that's crap we dont want to see. we'll have to add all kind of weird stuff for every new random system someone wants to support.

what i'm proposing is something that works for any system. to build u-boot:
make CROSS_COMPILE=bfin-uclinux- ...
to cross-compile the helper tools for some random system:
make HOSTCC=mingw32-gcc tools

if you want to avoid object file conflicts, then just use sep build trees with O=... like normal
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20090206/bc224919/attachment.pgp

#6 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Quoted 57 lines · Mike Frysinger
Quoted 41 lines · On Friday 06 February 2009 17:01:43 Peter Tyser wrote:
Quoted 34 lines · Mike Frysinger
Quoted 20 lines · On Friday 06 February 2009 14:59:10 Peter Tyser wrote:
Quoted 10 lines · Wolfgang Denk
Quoted 6 lines · In message Peter Tyser

Add a mkimage_win32.exe build target which can produce a native win32 mkimage executable using the MinGW toolchain. The mkimage_win32.exe binary is generated when the MINGW_COMPILE environment variable is defined. The mkimage_win32.exe binary can be used by those who use Windows as an OS build environment but don't use cygwin.

Why do we need a new envrironment variable? Can we not follow the standard pattern ansd use CROSS_COMPILE instead?

Most files can't be compiled with MinGW so if CROSS_COMPILE is set to a MinGW compile errors quickly occur in the build process. Also, it would be a bigger pain to compile u-boot + mkimage_win32.exe. Eg:

export CROSS_COMPILE=powerpc-linux-
export MINGW_COMPILE=i586-mingw32msvc-
make

Quoted 8 lines

vs

export CROSS_COMPILE=i586-mingw32msvc-
make
cp tools/mkimage_win32.exe ~/
make clean
export CROSS_COMPILE=powerpc-linux-
make

i think you're trying too hard. you're trying to combine three targets at once: the host system (ppc), the build system (i686 prob), and canadian cross for some other random build system (mingw).

Quoted 5 lines

the logical step would be to only build the tools for your random other build system instead of trying to combine it all together. in other words, set the HOSTCC to your mingw compiler and only process the tools target. this is how every other package out there works and it doesnt really make sense to have u- boot be different in this regard.

Quoted 3 lines

But Windows isn't really the HOSTCC, its a cross-compile target. Setting my HOSTCC to a cross compiler seems just as confusing as setting up 2 cross-compile targets to me.

when you're only building the tools, there is no native code execution nor building for the target. so the HOSTCC represents the system where the tools compiled will be run.

Quoted 1 line

The process of building would also still be slightly more complex to me:

that's just too bad then imo. what you're prosing is carrying system-specific cruft in the build system so you can avoid running make once more. that's not a scalable system and that's crap we dont want to see. we'll have to add all kind of weird stuff for every new random system someone wants to support.

Quoted 7 lines

what i'm proposing is something that works for any system. to build u-boot:
make CROSS_COMPILE=bfin-uclinux- ...
to cross-compile the helper tools for some random system:
make HOSTCC=mingw32-gcc tools

if you want to avoid object file conflicts, then just use sep build trees with O=... like normal

Is there any interest in getting win32 mkimage support into U-Boot? Would it be accepted if I make any suggested changes?

I don't want to spend any time discussing/cleaning/testing if its not going to be accepted:) If there is interest, I'll gladly cleanup and re-submit.

Thanks,
Peter

#7 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Quoted 70 lines · On Wednesday 25 February 2009 15:48:29 Peter Tyser wrote:
Quoted 61 lines · Mike Frysinger
Quoted 43 lines · On Friday 06 February 2009 17:01:43 Peter Tyser wrote:
Quoted 35 lines · Mike Frysinger
Quoted 21 lines · On Friday 06 February 2009 14:59:10 Peter Tyser wrote:
Quoted 10 lines · Wolfgang Denk
Quoted 6 lines · In message Peter Tyser

Add a mkimage_win32.exe build target which can produce a native win32 mkimage executable using the MinGW toolchain. The mkimage_win32.exe binary is generated when the MINGW_COMPILE environment variable is defined. The mkimage_win32.exe binary can be used by those who use Windows as an OS build environment but don't use cygwin.

Why do we need a new envrironment variable? Can we not follow the standard pattern ansd use CROSS_COMPILE instead?

Most files can't be compiled with MinGW so if CROSS_COMPILE is set to a MinGW compile errors quickly occur in the build process. Also, it would be a bigger pain to compile u-boot + mkimage_win32.exe. Eg:

export CROSS_COMPILE=powerpc-linux-
export MINGW_COMPILE=i586-mingw32msvc-
make

Quoted 8 lines

vs

export CROSS_COMPILE=i586-mingw32msvc-
make
cp tools/mkimage_win32.exe ~/
make clean
export CROSS_COMPILE=powerpc-linux-
make

i think you're trying too hard. you're trying to combine three targets at once: the host system (ppc), the build system (i686 prob), and canadian cross for some other random build system (mingw).

Quoted 6 lines

the logical step would be to only build the tools for your random other build system instead of trying to combine it all together. in other words, set the HOSTCC to your mingw compiler and only process the tools target. this is how every other package out there works and it doesnt really make sense to have u- boot be different in this regard.

Quoted 3 lines

But Windows isn't really the HOSTCC, its a cross-compile target. Setting my HOSTCC to a cross compiler seems just as confusing as setting up 2 cross-compile targets to me.

when you're only building the tools, there is no native code execution nor building for the target. so the HOSTCC represents the system where the tools compiled will be run.

Quoted 2 lines

The process of building would also still be slightly more complex to me:

that's just too bad then imo. what you're prosing is carrying system-specific cruft in the build system so you can avoid running make once more. that's not a scalable system and that's crap we dont want to see. we'll have to add all kind of weird stuff for every new random system someone wants to support.

Quoted 7 lines

what i'm proposing is something that works for any system. to build
u-boot: make CROSS_COMPILE=bfin-uclinux- ...
to cross-compile the helper tools for some random system:
make HOSTCC=mingw32-gcc tools

if you want to avoid object file conflicts, then just use sep build trees with O=... like normal

Quoted 6 lines

Is there any interest in getting win32 mkimage support into U-Boot? Would it be accepted if I make any suggested changes?

I don't want to spend any time discussing/cleaning/testing if its not going to be accepted:) If there is interest, I'll gladly cleanup and re-submit.

i'm interested in having it work, but i'm not the main u-boot maintainer, so i wouldnt be the one accepting/rejecting ...

i would however carry it in my tree ...
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20090226/88621925/attachment.pgp

#8 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

In reply to Peter Tyser

Hello Peter,

Quoted 6 lines

Is there any interest in getting win32 mkimage support into U-Boot? Would it be accepted if I make any suggested changes?

I don't want to spend any time discussing/cleaning/testing if its not going to be accepted:) ?If there is interest, I'll gladly cleanup and re-submit.

I have a similar patch in my tree for quite some time now. I did not mainline it because it is somewhat dirty and it requires the DTC to run on Windows too. (for which I have patches also)
So, yes, I am also interested in having this in mainline, but just as Mike said, it is up to the main u-boot maintainer for accepting/rejecting.

Kind Regards,

Remy

#9 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Dear Remy,

Quoted 9 lines · In message <[email protected]> you…

Hello Peter,

Quoted 6 lines

Is there any interest in getting win32 mkimage support into U-Boot? Would it be accepted if I make any suggested changes?

I don't want to spend any time discussing/cleaning/testing if its not going to be accepted:) =A0If there is interest, I'll gladly cleanup and re-submit.

Quoted 6 lines

I have a similar patch in my tree for quite some time now. I did not mainline it because it is somewhat dirty and it requires the DTC to run on Windows too. (for which I have patches also)
So, yes, I am also interested in having this in mainline, but just as Mike said, it is up to the main u-boot maintainer for accepting/rejecting.

Please post the patches, so these can be reviewd.

Our policy is that we accept stuff that looks technically clean and helps at least some of the people while at the same time not (obviously) hurting anybody.

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
For those who like this sort of thing, this is the sort of thing they
like.                                               - Abraham Lincoln
#10 lore raw

RFC PATCH 1/1 Add native win32 mkimage support

Hello Wolfgang,

Quoted 8 lines
Quoted 6 lines

I have a similar patch in my tree for quite some time now. I did not mainline it because it is somewhat dirty and it requires the DTC to run on Windows too. (for which I have patches also)
So, yes, I am also interested in having this in mainline, but just as Mike said, it is up to the main u-boot maintainer for accepting/rejecting.

Please post the patches, so these can be reviewd.

I had my mkimage patch cleanup on my todo list for some time, but I did not gave it any priority to work it up.

The patches for mkimage on Windows are in this thread of Peter Tyser, and I do not see a reason why my mkimage-mingw patches would be better, although it did it slightly different, and the patches I have for DTC are Off topic on this list.
The DTC is maintained elsewhere, so I do not see what I need to post here, unless it is for archiving/informational only.

Quoted 3 lines

Our policy is that we accept stuff that looks technically ?clean ?and helps ?at ?least ?some ?of ?the ?people ?while ?at ?the same time not (obviously) hurting anybody.

Not every hack I have here will be ever suitable for mainline ;-)

Kind Regards,

Remy