On 03/31/2015 06:02 AM, Masahiro Yamada wrote:
Since the Kconfig conversion, some developers have reported that Kbuild sometimes fails completely at random. According to the error reports, it seems to occur for any target board, but only on very fast computers.
The log message for the fail case is like this:
make[1]: *** No rule to make target `../arch//cpu/u-boot.lds', needed by `u-boot.lds'. Stop.
It looks like the top config.mk has not been included for *some* reason, and $(ARCH) has been left blank.
I suspect "autoconf_is_current" is not working in some situation.
This commit moves the definition of ARCH, CPU, SOC, etc. to the top Makefile, so they are surely set.
This doesn't solve the issue for me, although it improves the chance of success and changes the error I get.
With and without this patch, I ran:
CROSS_COMPILE=arm-linux-gnueabi- ./MAKEALL rpi_2
... until whichever of success/fail had the fewest "hits" had at least 5 "hits".
Without this patch, the build failed 6 out of 11 runs with:
====================================================================== boards.cfg is up to date. Nothing to do. Building rpi_2 board... arm-linux-gnueabi-size: './u-boot': No such file make: *** No rule to make target `arch//cpu/u-boot.lds', needed by `u-boot.lds'. Stop. make: *** Waiting for unfinished jobs....
--------------------- SUMMARY ---------------------------- Boards compiled: 1 Boards with warnings but no errors: 1 ( rpi_2 ) ---------------------------------------------------------- ======================================================================
With this patch, the build failed 5 out of 28 runs with:
====================================================================== boards.cfg is up to date. Nothing to do. Building rpi_2 board... arm-linux-gnueabi-size: './u-boot': No such file make: *** [prepare1] Error 1 Your architecture does not support generic board. Please undefine CONFIG_SYS_GENERIC_BOARD in your board config file. make: *** [prepare1] Error 1
--------------------- SUMMARY ---------------------------- Boards compiled: 1 Boards with errors: 1 ( rpi_2 ) ---------------------------------------------------------- ======================================================================
This feels more like a missing dependency or similar timing issue than an issue with the order of variable setup in the makefile.
I tried to repro this on a couple of VM services (Linode 8192, and EC2 c3.2xlarge) with the idea you could debug on those systems, but could not repro on those. Unfortunately the machine I have which repros the issue is on my work network, so I can't give you access.
One comment on the patch below:
diff --git a/Makefile b/Makefile
+ARCH = $(CONFIG_SYS_ARCH:"%"=%) +CPU = $(CONFIG_SYS_CPU:"%"=%) +SOC = $(CONFIG_SYS_SOC:"%"=%) +VENDOR = $(CONFIG_SYS_VENDOR:"%"=%) +BOARD = $(CONFIG_SYS_BOARD:"%"=%) +CPUDIR = arch/$(ARCH)/cpu$(if $(CPU),/$(CPU),) +BOARDDIR = $(if $(BOARD),$(if $(VENDOR),$(VENDOR)/)$(BOARD))
diff --git a/scripts/Makefile.autoconf b/scripts/Makefile.autoconf
+ARCH = $(CONFIG_SYS_ARCH:"%"=%) +CPU = $(CONFIG_SYS_CPU:"%"=%) +SOC = $(CONFIG_SYS_SOC:"%"=%) +VENDOR = $(CONFIG_SYS_VENDOR:"%"=%) +BOARD = $(CONFIG_SYS_BOARD:"%"=%) +CPUDIR = arch/$(ARCH)/cpu$(if $(CPU),/$(CPU),) +BOARDDIR = $(if $(BOARD),$(if $(VENDOR),$(VENDOR)))
Duplicating that doesn't seem like a good idea. Could it at least be included from a shared foo.mk file?