2.6: spurious recompiles
Subsystems (2)
When doing a "make" inside an already compiled kernel source there shouldn't be anything rebuilt. I've identified three places where this isn't the case in recent 2.6 kernels:
1. ikconfig
CC kernel/configs.o
even when the .config wasn't changed
2. pnmtologo
The following happens again once, but not when doing a third "make":
./scripts/pnmtologo -t mono -n logo_linux_mono -o drivers/video/logo/logo_linux_mono.c drivers/video/logo/logo_linux_mono.pbm
CC drivers/video/logo/logo_linux_mono.o
./scripts/pnmtologo -t vga16 -n logo_linux_vga16 -o drivers/video/logo/logo_linux_vga16.c drivers/video/logo/logo_linux_vga16.ppm
CC drivers/video/logo/logo_linux_vga16.o
./scripts/pnmtologo -t clut224 -n logo_linux_clut224 -o drivers/video/logo/logo_linux_clut224.c drivers/video/logo/logo_linux_clut224.ppm
CC drivers/video/logo/logo_linux_clut224.o LD drivers/video/logo/built-in.o LD drivers/video/built-in.o
3. aic7xxx
drivers/scsi/aic7xxx/aicasm/aicasm -Idrivers/scsi/aic7xxx -r drivers/scsi/aic7xxx/aic79xx_reg.h \
-p drivers/scsi/aic7xxx/aic79xx_reg_print.c -i
aic79xx_osm.h -o drivers/scsi/aic7xxx/aic79xx_seq.h \
drivers/scsi/aic7xxx/aic79xx.seq
drivers/scsi/aic7xxx/aicasm/aicasm: 785 instructions used
CC drivers/scsi/aic7xxx/aic79xx_core.o
CC drivers/scsi/aic7xxx/aic79xx_pci.o
CC drivers/scsi/aic7xxx/aic79xx_reg_print.o
CC drivers/scsi/aic7xxx/aic79xx_osm.o
CC drivers/scsi/aic7xxx/aic79xx_proc.o
CC drivers/scsi/aic7xxx/aic79xx_osm_pci.o
drivers/scsi/aic7xxx/aicasm/aicasm -Idrivers/scsi/aic7xxx -r
drivers/scsi/aic7xxx/aic7xxx_reg.h \
-p drivers/scsi/aic7xxx/aic7xxx_reg_print.c -i
aic7xxx_osm.h -o drivers/scsi/aic7xxx/aic7xxx_seq.h \
drivers/scsi/aic7xxx/aic7xxx.seq
drivers/scsi/aic7xxx/aicasm/aicasm: 879 instructions used
CC drivers/scsi/aic7xxx/aic7xxx_core.o
CC drivers/scsi/aic7xxx/aic7xxx_93cx6.o
CC drivers/scsi/aic7xxx/aic7770.o
CC drivers/scsi/aic7xxx/aic7xxx_pci.o
CC drivers/scsi/aic7xxx/aic7xxx_reg_print.o
CC drivers/scsi/aic7xxx/aic7xxx_osm.o
CC drivers/scsi/aic7xxx/aic7xxx_proc.o
CC drivers/scsi/aic7xxx/aic7770_osm.o
CC drivers/scsi/aic7xxx/aic7xxx_osm_pci.o
LD drivers/scsi/aic7xxx/aic7xxx.o
LD drivers/scsi/aic7xxx/aic79xx.o
LD drivers/scsi/aic7xxx/built-in.ocu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 7 lines
When doing a "make" inside an already compiled kernel source there shouldn't be anything rebuilt. I've identified three places where this isn't the case in recent 2.6 kernels:
1. ikconfig
CCkernel/configs.o
even when the .config wasn't changed
This is probably the same thing that Steve Hemminger posted about yesterday:
http://marc.theaimsgroup.com/?l=linux-kernel&m=106270067411137&w=2
I posted a patch based on Sam Ravnborg's comments that might fix it, but I haven't verified it yet... The patch is in this message:
http://marc.theaimsgroup.com/?l=linux-kernel&m=106272687506379&w=2
or it may be some other dependency. I'll look into it.
Thanks,
~Randy
Quoted 4 lines · Randy.Dunlap
Quoted 4 lines
When doing a "make" inside an already compiled kernel source there shouldn't be anything rebuilt. I've identified three places where this isn't the case in recent 2.6 kernels:
Quoted 8 lines
Quoted 3 lines
1. ikconfig
CCkernel/configs.o
even when the .config wasn't changed...
I posted a patch based on Sam Ravnborg's comments that might fix it, but I haven't verified it yet... The patch is in this message:
http://marc.theaimsgroup.com/?l=linux-kernel&m=106272687506379&w=2
...
It seems I missed this mail.
The patch in this mail seems to fix the recompiles.
Quoted 2 lines
Thanks,
~Randy
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 7 lines · Adrian Bunk
When doing a "make" inside an already compiled kernel source there shouldn't be anything rebuilt. I've identified three places where this isn't the case in recent 2.6 kernels:
1. ikconfig
CCkernel/configs.o
even when the .config wasn't changed
configs.o included compile.h. Compile.h contains date when kernel was compiled, and gets updated each time there is new .o files. That is fixed in patch sent to Randy.
Quoted 10 lines
2. pnmtologo
The following happens again once, but not when doing a third "make":
./scripts/pnmtologo-t mono -nlogo_linux_mono-odrivers/video/logo/logo_linux_mono.cdrivers/video/logo/logo_linux_mono.pbm
CCdrivers/video/logo/logo_linux_mono.o
./scripts/pnmtologo-t vga16 -nlogo_linux_vga16-odrivers/video/logo/logo_linux_vga16.cdrivers/video/logo/logo_linux_vga16.ppm
CCdrivers/video/logo/logo_linux_vga16.o
./scripts/pnmtologo-t clut224 -nlogo_linux_clut224-odrivers/video/logo/logo_linux_clut224.cdrivers/video/logo/logo_linux_clut224.ppmCC drivers/video/logo/logo_linux_clut224.o LD drivers/video/logo/built-in.o LD drivers/video/built-in.o
I have sent a patch to James Simmons some time ago. I will try to dig it up and see if it still applies, and fixes the problem.
Quoted 8 lines
3. aic7xxx
drivers/scsi/aic7xxx/aicasm/aicasm-Idrivers/scsi/aic7xxx -rdrivers/scsi/aic7xxx/aic79xx_reg.h\
-pdrivers/scsi/aic7xxx/aic79xx_reg_print.c-i
aic79xx_osm.h-odrivers/scsi/aic7xxx/aic79xx_seq.h\
drivers/scsi/aic7xxx/aic79xx.seq
drivers/scsi/aic7xxx/aicasm/aicasm: 785 instructions used
CCdrivers/scsi/aic7xxx/aic79xx_core.o
New to me, I will take a look.
Sam
Quoted 12 lines
Quoted 11 lines
2. pnmtologo
The following happens again once, but not when doing a third "make":
./scripts/pnmtologo-t mono -nlogo_linux_mono-odrivers/video/logo/logo_linux_mono.cdrivers/video/logo/logo_linux_mono.pbm
CCdrivers/video/logo/logo_linux_mono.o
./scripts/pnmtologo-t vga16 -nlogo_linux_vga16-odrivers/video/logo/logo_linux_vga16.cdrivers/video/logo/logo_linux_vga16.ppm
CCdrivers/video/logo/logo_linux_vga16.o
./scripts/pnmtologo-t clut224 -nlogo_linux_clut224-odrivers/video/logo/logo_linux_clut224.cdrivers/video/logo/logo_linux_clut224.ppmCC drivers/video/logo/logo_linux_clut224.o LD drivers/video/logo/built-in.o LD drivers/video/built-in.o
Quoted 2 lines
I have sent a patch to James Simmons some time ago. I will try to dig it up and see if it still applies, and fixes the problem.
I have the patch in the fbdev BK tree.
Quoted 3 lines · Adrian Bunk
2. pnmtologo
The following happens again once, but not when doing a third "make":
./scripts/pnmtologo-t mono -nlogo_linux_mono-odrivers/video/logo/logo_linux_mono.cdrivers/video/logo/logo_linux_mono.pbm
Would you mind to give this patch a spin. Only lightly tested here.
===== drivers/video/aty/Makefile 1.10 vs edited =====
drivers/video/aty/Makefile +0 -1
@@ -4,4 +4,3 @@ atyfb-y := atyfb_base.o mach64_accel.o atyfb-$(CONFIG_FB_ATY_GX) += mach64_gx.o atyfb-$(CONFIG_FB_ATY_CT) += mach64_ct.o mach64_cursor.o -atyfb-objs := $(atyfb-y)
===== drivers/video/console/Makefile 1.15 vs edited =====
drivers/video/console/Makefile +14 -13
@@ -3,18 +3,16 @@ # Rewritten to use lists instead of if-statements. # Font handling -font-objs := fonts.o +font-y := fonts.o -font-objs-$(CONFIG_FONT_SUN8x16) += font_sun8x16.o -font-objs-$(CONFIG_FONT_SUN12x22) += font_sun12x22.o -font-objs-$(CONFIG_FONT_8x8) += font_8x8.o -font-objs-$(CONFIG_FONT_8x16) += font_8x16.o -font-objs-$(CONFIG_FONT_6x11) += font_6x11.o -font-objs-$(CONFIG_FONT_PEARL_8x8) += font_pearl_8x8.o -font-objs-$(CONFIG_FONT_ACORN_8x8) += font_acorn_8x8.o -font-objs-$(CONFIG_FONT_MINI_4x6) += font_mini_4x6.o - -font-objs += $(font-objs-y) +font-$(CONFIG_FONT_SUN8x16) += font_sun8x16.o +font-$(CONFIG_FONT_SUN12x22) += font_sun12x22.o +font-$(CONFIG_FONT_8x8) += font_8x8.o +font-$(CONFIG_FONT_8x16) += font_8x16.o +font-$(CONFIG_FONT_6x11) += font_6x11.o +font-$(CONFIG_FONT_PEARL_8x8) += font_pearl_8x8.o +font-$(CONFIG_FONT_ACORN_8x8) += font_acorn_8x8.o +font-$(CONFIG_FONT_MINI_4x6) += font_mini_4x6.o # Each configuration option enables a list of files.
@@ -31,8 +29,11 @@ # Files generated that shall be removed upon make clean clean-files := promcon_tbl.c -$(obj)/promcon_tbl.c: $(src)/prom.uni - $(objtree)/scripts/conmakehash $< | \ + +quiet_cmd_promtbl = GEN $@ + cmd_promtbl = scripts/conmakehash $< | \ sed -e '/#include <[^>]*>/p' -e 's/types/init/' \ -e 's/dfont\(_uni.*\]\)/promfont\1 __initdata/' > $@ +$(obj)/promcon_tbl.c: $(src)/prom.uni + $(call cmd,promtbl)
===== drivers/video/i810/Makefile 1.2 vs edited =====
drivers/video/i810/Makefile +0 -7
@@ -1,15 +1,8 @@ # # Makefile for the Intel 810/815 framebuffer driver # -# Note! Dependencies are done automagically by 'make dep', which also -# removes any old dependencies. DON'T put your own dependencies here -# unless it's something special (ie not a .c file). -# -# Note 2! The CFLAGS definitions are now in the main makefile... - obj-$(CONFIG_FB_I810) += i810fb.o - i810fb-objs := i810_main.o i810_accel.o
===== drivers/video/logo/Makefile 1.3 vs edited =====
drivers/video/logo/Makefile +10 -5
@@ -25,18 +25,23 @@ # How to generate them +# Create commands like "pnmtologo -t mono -n logo_mac_mono -o ..." +quiet_cmd_logo = LOGO $@ + cmd_logo = scripts/pnmtologo \ + -t $(patsubst $*_%,%,$(notdir $(basename $<))) \ + -n $(notdir $(basename $<)) -o $@ $< + $(obj)/%_mono.c: $(src)/%_mono.pbm - $(objtree)/scripts/pnmtologo -t mono -n $*_mono -o $@ $< + $(call cmd,logo) $(obj)/%_vga16.c: $(src)/%_vga16.ppm - $(objtree)/scripts/pnmtologo -t vga16 -n $*_vga16 -o $@ $< + $(call cmd,logo) $(obj)/%_clut224.c: $(src)/%_clut224.ppm - $(objtree)/scripts/pnmtologo -t clut224 -n $*_clut224 -o $@ $< + $(call cmd,logo) $(obj)/%_gray256.c: $(src)/%_gray256.pgm - $(objtree)/scripts/pnmtologo -t gray256 -n $*_gray256 -o $@ $< - + $(call cmd,logo) # Files generated that shall be removed upon make clean clean-files := *_mono.c *_vga16.c *_clut224.c *_gray256.c
Quoted 6 lines · Sam Ravnborg
Quoted 3 lines · Adrian Bunk
2. pnmtologo
The following happens again once, but not when doing a third "make":
./scripts/pnmtologo-t mono -nlogo_linux_mono-odrivers/video/logo/logo_linux_mono.cdrivers/video/logo/logo_linux_mono.pbmWould you mind to give this patch a spin. Only lightly tested here.
Tested it a bit more, it was not good. Corrected patch follows. I had to spell out the dependencies, otherwise make saw them as implicit targets, and deleted the .c files afterwards.
Btw the patch contains some general clean-up as well, unrelated to the 'spurious recompile' issue.
Sam
===== drivers/video/aty/Makefile 1.10 vs edited =====
drivers/video/aty/Makefile +0 -1
@@ -4,4 +4,3 @@ atyfb-y := atyfb_base.o mach64_accel.o atyfb-$(CONFIG_FB_ATY_GX) += mach64_gx.o atyfb-$(CONFIG_FB_ATY_CT) += mach64_ct.o mach64_cursor.o -atyfb-objs := $(atyfb-y)
===== drivers/video/console/Makefile 1.15 vs edited =====
drivers/video/console/Makefile +14 -13
@@ -3,18 +3,16 @@ # Rewritten to use lists instead of if-statements. # Font handling -font-objs := fonts.o +font-y := fonts.o -font-objs-$(CONFIG_FONT_SUN8x16) += font_sun8x16.o -font-objs-$(CONFIG_FONT_SUN12x22) += font_sun12x22.o -font-objs-$(CONFIG_FONT_8x8) += font_8x8.o -font-objs-$(CONFIG_FONT_8x16) += font_8x16.o -font-objs-$(CONFIG_FONT_6x11) += font_6x11.o -font-objs-$(CONFIG_FONT_PEARL_8x8) += font_pearl_8x8.o -font-objs-$(CONFIG_FONT_ACORN_8x8) += font_acorn_8x8.o -font-objs-$(CONFIG_FONT_MINI_4x6) += font_mini_4x6.o - -font-objs += $(font-objs-y) +font-$(CONFIG_FONT_SUN8x16) += font_sun8x16.o +font-$(CONFIG_FONT_SUN12x22) += font_sun12x22.o +font-$(CONFIG_FONT_8x8) += font_8x8.o +font-$(CONFIG_FONT_8x16) += font_8x16.o +font-$(CONFIG_FONT_6x11) += font_6x11.o +font-$(CONFIG_FONT_PEARL_8x8) += font_pearl_8x8.o +font-$(CONFIG_FONT_ACORN_8x8) += font_acorn_8x8.o +font-$(CONFIG_FONT_MINI_4x6) += font_mini_4x6.o # Each configuration option enables a list of files.
@@ -31,8 +29,11 @@ # Files generated that shall be removed upon make clean clean-files := promcon_tbl.c -$(obj)/promcon_tbl.c: $(src)/prom.uni - $(objtree)/scripts/conmakehash $< | \ + +quiet_cmd_promtbl = GEN $@ + cmd_promtbl = scripts/conmakehash $< | \ sed -e '/#include <[^>]*>/p' -e 's/types/init/' \ -e 's/dfont\(_uni.*\]\)/promfont\1 __initdata/' > $@ +$(obj)/promcon_tbl.c: $(src)/prom.uni + $(call cmd,promtbl)
===== drivers/video/i810/Makefile 1.2 vs edited =====
drivers/video/i810/Makefile +0 -7
@@ -1,15 +1,8 @@ # # Makefile for the Intel 810/815 framebuffer driver # -# Note! Dependencies are done automagically by 'make dep', which also -# removes any old dependencies. DON'T put your own dependencies here -# unless it's something special (ie not a .c file). -# -# Note 2! The CFLAGS definitions are now in the main makefile... - obj-$(CONFIG_FB_I810) += i810fb.o - i810fb-objs := i810_main.o i810_accel.o
===== drivers/video/logo/Makefile 1.3 vs edited =====
drivers/video/logo/Makefile +25 -16
54 lines not on this page · raw message
Quoted 10 lines · Sam Ravnborg
Quoted 6 lines · Sam Ravnborg
Quoted 3 lines · Adrian Bunk
2. pnmtologo
The following happens again once, but not when doing a third "make":
./scripts/pnmtologo-t mono -nlogo_linux_mono-odrivers/video/logo/logo_linux_mono.cdrivers/video/logo/logo_linux_mono.pbmWould you mind to give this patch a spin. Only lightly tested here.
Tested it a bit more, it was not good. Corrected patch follows. I had to spell out the dependencies, otherwise make saw them as implicit targets, and deleted the .c files afterwards.
Thanks, this patch seems to fix it.
Quoted 2 lines
Btw the patch contains some general clean-up as well, unrelated to the 'spurious recompile' issue.
These are the places where I got rejects in 2.6.0-test4-mm6 (these changes are already in the -mm tree).
Quoted 2 lines
Sam
...
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedI have most of this patch already. I applied your logo Makefile changes. I will test them tonight.
Quoted 131 lines
=====
drivers/video/aty/Makefile1.10 vs edited =====drivers/video/aty/Makefile +0
-1@@ -4,4 +4,3 @@ atyfb-y := atyfb_base.o mach64_accel.o atyfb-$(CONFIG_FB_ATY_GX) += mach64_gx.o atyfb-$(CONFIG_FB_ATY_CT) += mach64_ct.o mach64_cursor.o -atyfb-objs := $(atyfb-y)=====
drivers/video/console/Makefile1.15 vs edited =====drivers/video/console/Makefile +14
-13@@ -3,18 +3,16 @@ # Rewritten to use lists instead of if-statements. # Font handling -font-objs := fonts.o +font-y := fonts.o -font-objs-$(CONFIG_FONT_SUN8x16) += font_sun8x16.o -font-objs-$(CONFIG_FONT_SUN12x22) += font_sun12x22.o -font-objs-$(CONFIG_FONT_8x8) += font_8x8.o -font-objs-$(CONFIG_FONT_8x16) += font_8x16.o -font-objs-$(CONFIG_FONT_6x11) += font_6x11.o -font-objs-$(CONFIG_FONT_PEARL_8x8) += font_pearl_8x8.o -font-objs-$(CONFIG_FONT_ACORN_8x8) += font_acorn_8x8.o -font-objs-$(CONFIG_FONT_MINI_4x6) += font_mini_4x6.o - -font-objs += $(font-objs-y) +font-$(CONFIG_FONT_SUN8x16) += font_sun8x16.o +font-$(CONFIG_FONT_SUN12x22) += font_sun12x22.o +font-$(CONFIG_FONT_8x8) += font_8x8.o +font-$(CONFIG_FONT_8x16) += font_8x16.o +font-$(CONFIG_FONT_6x11) += font_6x11.o +font-$(CONFIG_FONT_PEARL_8x8) += font_pearl_8x8.o +font-$(CONFIG_FONT_ACORN_8x8) += font_acorn_8x8.o +font-$(CONFIG_FONT_MINI_4x6) += font_mini_4x6.o # Each configuration option enables a list of files.@@ -31,8 +29,11 @@ # Files generated that shall be removed upon make clean clean-files := promcon_tbl.c -$(obj)/promcon_tbl.c: $(src)/prom.uni - $(objtree)/scripts/conmakehash $< | \ + +quiet_cmd_promtbl = GEN $@ + cmd_promtbl = scripts/conmakehash $< | \ sed -e '/#include <[^>]*>/p' -e 's/types/init/' \ -e 's/dfont\(_uni.*\]\)/promfont\1 __initdata/' > $@ +$(obj)/promcon_tbl.c: $(src)/prom.uni + $(call cmd,promtbl)=====
drivers/video/i810/Makefile1.2 vs edited =====drivers/video/i810/Makefile +0
-7@@ -1,15 +1,8 @@ # # Makefile for the Intel 810/815 framebuffer driver # -# Note! Dependencies are done automagically by 'make dep', which also -# removes any old dependencies. DON'T put your own dependencies here -# unless it's something special (ie not a .c file). -# -# Note 2! The CFLAGS definitions are now in the main makefile... - obj-$(CONFIG_FB_I810) += i810fb.o - i810fb-objs := i810_main.o i810_accel.o=====
drivers/video/logo/Makefile1.3 vs edited =====drivers/video/logo/Makefile +25
-16@@ -14,29 +14,38 @@ obj-$(CONFIG_LOGO_SUPERH_CLUT224) += logo_superh_clut224.o # Dependencies on generated files need to be listed explicitly - -$(obj)/%_mono.o: $(src)/%_mono.c - -$(obj)/%_vga16.o: $(src)/%_vga16.c - -$(obj)/%_clut224.o: $(src)/%_clut224.c - -$(obj)/%_gray256.o: $(src)/%_gray256.c +# mono logo's +$(obj)/logo_linux_mono.o: $(obj)/logo_linux_mono.c +$(obj)/logo_superh_mono.o: $(obj)/logo_superh_mono.c + +# vga16 logo's +$(obj)/logo_linux_vga16.o: $(obj)/logo_linux_vga16.c +$(obj)/logo_superh_vga16.o: $(obj)/logo_superh_vga16.c + +# clut224 logo's +$(obj)/logo_linux_clut224.o: $(obj)/logo_linux_clut224.c +$(obj)/logo_dec_clut224.o: $(obj)/logo_dec_clut224.c +$(obj)/logo_mac_clut224.o: $(obj)/logo_mac_clut224.c +$(obj)/logo_sgi_clut224.o: $(obj)/logo_sgi_clut224.c +$(obj)/logo_sun_clut224.o: $(obj)/logo_sun_clut224.c +$(obj)/logo_superh_clut224.o: $(obj)/logo_superh_clut224.c # How to generate them +# Create commands like "pnmtologo -t mono -n logo_mac_mono -o ..." +quiet_cmd_logo = LOGO $@ + cmd_logo = scripts/pnmtologo \ + -t $(patsubst $*_%,%,$(notdir $(basename $<))) \ + -n $(notdir $(basename $<)) -o $@ $< + $(obj)/%_mono.c: $(src)/%_mono.pbm - $(objtree)/scripts/pnmtologo -t mono -n $*_mono -o $@ $< + $(call cmd,logo) $(obj)/%_vga16.c: $(src)/%_vga16.ppm - $(objtree)/scripts/pnmtologo -t vga16 -n $*_vga16 -o $@ $< + $(call cmd,logo) $(obj)/%_clut224.c: $(src)/%_clut224.ppm - $(objtree)/scripts/pnmtologo -t clut224 -n $*_clut224 -o $@ $< - -$(obj)/%_gray256.c: $(src)/%_gray256.pgm - $(objtree)/scripts/pnmtologo -t gray256 -n $*_gray256 -o $@ $< - + $(call cmd,logo) # Files generated that shall be removed upon make clean -clean-files := *_mono.c *_vga16.c *_clut224.c *_gray256.c +clean-files := *_mono.c *_vga16.c *_clut224.c